seed
int64 42
46
| dataset
stringclasses 6
values | model
stringclasses 13
values | n_topics
int64 10
50
| topic_descriptions
sequencelengths 1
50
| runtime_s
float64 0.93
52.3k
| encoder
stringclasses 5
values | diversity
float64 0.09
1
| c_npmi
float64 -0.38
0.21
| wec_ex
float64 0.11
0.49
| wec_in
float64 0.07
0.94
|
---|---|---|---|---|---|---|---|---|---|---|
42 | ArXiv ML Papers | BERTopic | 10 | [
[
"the",
"to",
"and",
"of",
"in",
"with",
"network",
"neural",
"is",
"on"
],
[
"of",
"and",
"the",
"we",
"algorithm",
"to",
"is",
"for",
"in",
"that"
],
[
"of",
"and",
"learning",
"the",
"to",
"in",
"we",
"that",
"this",
"for"
],
[
"of",
"the",
"bound",
"is",
"we",
"algorithm",
"in",
"delta",
"for",
"lower"
],
[
"of",
"the",
"forecast",
"day",
"forecasting",
"for",
"demand",
"in",
"and",
"forecasts"
],
[
"adversarial",
"attacks",
"the",
"and",
"attack",
"to",
"of",
"we",
"in",
"model"
],
[
"in",
"we",
"is",
"of",
"and",
"to",
"the",
"on",
"for",
"that"
],
[
"the",
"of",
"graph",
"graphs",
"and",
"to",
"node",
"we",
"on",
"in"
],
[
"of",
"and",
"the",
"we",
"word",
"visual",
"language",
"to",
"in",
"that"
],
[
"in",
"image",
"with",
"the",
"of",
"for",
"and",
"to",
"images",
"we"
]
] | 12.458354 | average_word_embeddings_glove.6B.300d | 0.36 | -0.017372 | 0.226473 | 0.56269 |
42 | ArXiv ML Papers | BERTopic | 20 | [
[
"devices",
"to",
"and",
"with",
"on",
"in",
"the",
"of",
"for",
"accuracy"
],
[
"neural",
"networks",
"network",
"to",
"the",
"and",
"of",
"with",
"in",
"is"
],
[
"of",
"to",
"adversarial",
"and",
"attacks",
"the",
"we",
"attack",
"model",
"against"
],
[
"to",
"language",
"the",
"we",
"of",
"in",
"and",
"word",
"that",
"visual"
],
[
"of",
"to",
"the",
"in",
"and",
"we",
"that",
"representations",
"as",
"data"
],
[
"in",
"learning",
"the",
"this",
"that",
"tasks",
"and",
"of",
"we",
"to"
],
[
"and",
"the",
"research",
"in",
"of",
"we",
"user",
"users",
"to",
"learning"
],
[
"to",
"node",
"the",
"of",
"graph",
"graphs",
"we",
"and",
"in",
"on"
],
[
"image",
"in",
"images",
"of",
"and",
"the",
"for",
"with",
"to",
"we"
],
[
"the",
"for",
"forecast",
"day",
"forecasting",
"forecasts",
"demand",
"and",
"of",
"weather"
],
[
"and",
"of",
"data",
"to",
"the",
"in",
"traffic",
"is",
"network",
"deep"
],
[
"the",
"to",
"in",
"is",
"model",
"for",
"and",
"of",
"pooling",
"methods"
],
[
"algorithms",
"and",
"we",
"in",
"structured",
"the",
"clustering",
"to",
"of",
"on"
],
[
"we",
"of",
"bound",
"delta",
"for",
"the",
"lower",
"is",
"algorithm",
"in"
],
[
"we",
"algorithm",
"the",
"that",
"and",
"for",
"to",
"of",
"in",
"is"
]
] | 9.810699 | average_word_embeddings_glove.6B.300d | 0.366667 | -0.016607 | 0.19747 | 0.567104 |
42 | ArXiv ML Papers | BERTopic | 30 | [
[
"devices",
"to",
"and",
"with",
"on",
"in",
"the",
"of",
"for",
"accuracy"
],
[
"neural",
"networks",
"network",
"to",
"the",
"and",
"of",
"with",
"in",
"is"
],
[
"of",
"to",
"adversarial",
"and",
"attacks",
"the",
"we",
"attack",
"model",
"against"
],
[
"to",
"language",
"the",
"we",
"of",
"in",
"and",
"word",
"that",
"visual"
],
[
"of",
"to",
"the",
"in",
"and",
"we",
"that",
"representations",
"as",
"data"
],
[
"in",
"learning",
"the",
"this",
"that",
"tasks",
"and",
"of",
"we",
"to"
],
[
"and",
"the",
"research",
"in",
"of",
"we",
"user",
"users",
"to",
"learning"
],
[
"to",
"node",
"the",
"of",
"graph",
"graphs",
"we",
"and",
"in",
"on"
],
[
"image",
"in",
"images",
"of",
"and",
"the",
"for",
"with",
"to",
"we"
],
[
"the",
"for",
"forecast",
"day",
"forecasting",
"forecasts",
"demand",
"and",
"of",
"weather"
],
[
"and",
"of",
"data",
"to",
"the",
"in",
"traffic",
"is",
"network",
"deep"
],
[
"the",
"to",
"in",
"is",
"model",
"for",
"and",
"of",
"pooling",
"methods"
],
[
"algorithms",
"and",
"we",
"in",
"structured",
"the",
"clustering",
"to",
"of",
"on"
],
[
"we",
"of",
"bound",
"delta",
"for",
"the",
"lower",
"is",
"algorithm",
"in"
],
[
"we",
"algorithm",
"the",
"that",
"and",
"for",
"to",
"of",
"in",
"is"
]
] | 8.606691 | average_word_embeddings_glove.6B.300d | 0.366667 | -0.016607 | 0.19747 | 0.564008 |
42 | ArXiv ML Papers | BERTopic | 40 | [
[
"devices",
"to",
"and",
"with",
"on",
"in",
"the",
"of",
"for",
"accuracy"
],
[
"neural",
"networks",
"network",
"to",
"the",
"and",
"of",
"with",
"in",
"is"
],
[
"of",
"to",
"adversarial",
"and",
"attacks",
"the",
"we",
"attack",
"model",
"against"
],
[
"to",
"language",
"the",
"we",
"of",
"in",
"and",
"word",
"that",
"visual"
],
[
"of",
"to",
"the",
"in",
"and",
"we",
"that",
"representations",
"as",
"data"
],
[
"in",
"learning",
"the",
"this",
"that",
"tasks",
"and",
"of",
"we",
"to"
],
[
"and",
"the",
"research",
"in",
"of",
"we",
"user",
"users",
"to",
"learning"
],
[
"to",
"node",
"the",
"of",
"graph",
"graphs",
"we",
"and",
"in",
"on"
],
[
"image",
"in",
"images",
"of",
"and",
"the",
"for",
"with",
"to",
"we"
],
[
"the",
"for",
"forecast",
"day",
"forecasting",
"forecasts",
"demand",
"and",
"of",
"weather"
],
[
"and",
"of",
"data",
"to",
"the",
"in",
"traffic",
"is",
"network",
"deep"
],
[
"the",
"to",
"in",
"is",
"model",
"for",
"and",
"of",
"pooling",
"methods"
],
[
"algorithms",
"and",
"we",
"in",
"structured",
"the",
"clustering",
"to",
"of",
"on"
],
[
"we",
"of",
"bound",
"delta",
"for",
"the",
"lower",
"is",
"algorithm",
"in"
],
[
"we",
"algorithm",
"the",
"that",
"and",
"for",
"to",
"of",
"in",
"is"
]
] | 9.002439 | average_word_embeddings_glove.6B.300d | 0.366667 | -0.016607 | 0.19747 | 0.570943 |
42 | ArXiv ML Papers | BERTopic | 50 | [
[
"devices",
"to",
"and",
"with",
"on",
"in",
"the",
"of",
"for",
"accuracy"
],
[
"neural",
"networks",
"network",
"to",
"the",
"and",
"of",
"with",
"in",
"is"
],
[
"of",
"to",
"adversarial",
"and",
"attacks",
"the",
"we",
"attack",
"model",
"against"
],
[
"to",
"language",
"the",
"we",
"of",
"in",
"and",
"word",
"that",
"visual"
],
[
"of",
"to",
"the",
"in",
"and",
"we",
"that",
"representations",
"as",
"data"
],
[
"in",
"learning",
"the",
"this",
"that",
"tasks",
"and",
"of",
"we",
"to"
],
[
"and",
"the",
"research",
"in",
"of",
"we",
"user",
"users",
"to",
"learning"
],
[
"to",
"node",
"the",
"of",
"graph",
"graphs",
"we",
"and",
"in",
"on"
],
[
"image",
"in",
"images",
"of",
"and",
"the",
"for",
"with",
"to",
"we"
],
[
"the",
"for",
"forecast",
"day",
"forecasting",
"forecasts",
"demand",
"and",
"of",
"weather"
],
[
"and",
"of",
"data",
"to",
"the",
"in",
"traffic",
"is",
"network",
"deep"
],
[
"the",
"to",
"in",
"is",
"model",
"for",
"and",
"of",
"pooling",
"methods"
],
[
"algorithms",
"and",
"we",
"in",
"structured",
"the",
"clustering",
"to",
"of",
"on"
],
[
"we",
"of",
"bound",
"delta",
"for",
"the",
"lower",
"is",
"algorithm",
"in"
],
[
"we",
"algorithm",
"the",
"that",
"and",
"for",
"to",
"of",
"in",
"is"
]
] | 8.378359 | average_word_embeddings_glove.6B.300d | 0.366667 | -0.016607 | 0.19747 | 0.559362 |
42 | ArXiv ML Papers | NMF | 10 | [
[
"the",
"and",
"of",
"proposed",
"to",
"on",
"by",
"based",
"method",
"is"
],
[
"to",
"the",
"be",
"this",
"neural",
"model",
"from",
"can",
"networks",
"an"
],
[
"to",
"an",
"of",
"this",
"the",
"number",
"these",
"are",
"our",
"with"
],
[
"models",
"both",
"are",
"with",
"based",
"on",
"and",
"methods",
"as",
"performance"
],
[
"show",
"as",
"the",
"we",
"on",
"with",
"our",
"propose",
"that",
"can"
],
[
"are",
"the",
"has",
"in",
"this",
"we",
"or",
"which",
"as",
"been"
],
[
"for",
"with",
"an",
"we",
"are",
"be",
"as",
"algorithm",
"can",
"used"
],
[
"training",
"the",
"for",
"is",
"data",
"from",
"as",
"on",
"model",
"models"
],
[
"machine",
"to",
"tasks",
"on",
"of",
"algorithms",
"and",
"learning",
"deep",
"reinforcement"
],
[
"algorithm",
"and",
"this",
"is",
"it",
"that",
"are",
"be",
"by",
"which"
]
] | 1.617862 | average_word_embeddings_glove.6B.300d | 0.49 | -0.008151 | 0.195331 | 0.466071 |
42 | ArXiv ML Papers | NMF | 20 | [
[
"using",
"information",
"between",
"proposed",
"to",
"the",
"results",
"by",
"of",
"two"
],
[
"from",
"the",
"it",
"due",
"an",
"to",
"their",
"framework",
"this",
"approach"
],
[
"these",
"set",
"the",
"this",
"number",
"an",
"of",
"new",
"state",
"art"
],
[
"ml",
"research",
"models",
"and",
"both",
"the",
"their",
"including",
"two",
"are"
],
[
"problem",
"which",
"our",
"show",
"we",
"and",
"the",
"propose",
"in",
"algorithm"
],
[
"in",
"or",
"are",
"work",
"paper",
"the",
"this",
"been",
"has",
"many"
],
[
"time",
"are",
"the",
"each",
"for",
"test",
"also",
"used",
"using",
"analysis"
],
[
"from",
"to",
"we",
"data",
"training",
"real",
"is",
"classification",
"of",
"time"
],
[
"deep",
"of",
"learning",
"machine",
"algorithms",
"reinforcement",
"tasks",
"policy",
"supervised",
"has"
],
[
"it",
"problem",
"which",
"where",
"algorithm",
"is",
"and",
"this",
"an",
"at"
],
[
"based",
"an",
"on",
"performance",
"by",
"datasets",
"this",
"tasks",
"classification",
"task"
],
[
"the",
"representations",
"are",
"this",
"show",
"not",
"more",
"than",
"also",
"that"
],
[
"training",
"deep",
"convolutional",
"to",
"is",
"network",
"networks",
"neural",
"we",
"layer"
],
[
"model",
"which",
"trained",
"we",
"performance",
"models",
"inference",
"in",
"our",
"based"
],
[
"to",
"have",
"an",
"are",
"as",
"such",
"or",
"well",
"these",
"using"
],
[
"with",
"dataset",
"high",
"accuracy",
"an",
"by",
"low",
"performance",
"results",
"while"
],
[
"and",
"state",
"methods",
"gradient",
"based",
"method",
"optimization",
"proposed",
"propose",
"our"
],
[
"we",
"training",
"models",
"robustness",
"attacks",
"adversarial",
"attack",
"to",
"against",
"the"
],
[
"by",
"be",
"this",
"can",
"from",
"it",
"or",
"which",
"used",
"how"
],
[
"the",
"graphs",
"graph",
"gnns",
"representation",
"nodes",
"node",
"and",
"our",
"we"
]
] | 2.621725 | average_word_embeddings_glove.6B.300d | 0.565 | 0.013258 | 0.170953 | 0.559186 |
42 | ArXiv ML Papers | NMF | 30 | [
[
"of",
"the",
"to",
"proposed",
"between",
"problem",
"results",
"information",
"two",
"space"
],
[
"end",
"human",
"it",
"have",
"search",
"due",
"their",
"to",
"use",
"framework"
],
[
"and",
"terms",
"these",
"the",
"of",
"number",
"set",
"analysis",
"art",
"large"
],
[
"accuracy",
"the",
"two",
"and",
"research",
"both",
"between",
"including",
"ml",
"their"
],
[
"present",
"the",
"we",
"propose",
"show",
"use",
"which",
"in",
"demonstrate",
"where"
],
[
"been",
"in",
"or",
"applications",
"has",
"work",
"many",
"recent",
"have",
"high"
],
[
"using",
"the",
"prediction",
"time",
"test",
"used",
"each",
"regression",
"for",
"also"
],
[
"data",
"and",
"classification",
"real",
"training",
"time",
"to",
"synthetic",
"large",
"supervised"
],
[
"supervised",
"of",
"deep",
"reinforcement",
"learning",
"machine",
"tasks",
"has",
"metric",
"agent"
],
[
"it",
"which",
"where",
"one",
"is",
"not",
"at",
"than",
"when",
"information"
],
[
"on",
"performance",
"tasks",
"datasets",
"task",
"classification",
"art",
"trained",
"work",
"or"
],
[
"that",
"show",
"than",
"not",
"representations",
"the",
"results",
"also",
"more",
"find"
],
[
"training",
"convolutional",
"neural",
"networks",
"network",
"and",
"deep",
"to",
"layer",
"performance"
],
[
"the",
"model",
"performance",
"and",
"which",
"in",
"trained",
"proposed",
"inference",
"parameters"
],
[
"as",
"well",
"have",
"such",
"to",
"these",
"or",
"been",
"not",
"using"
],
[
"dataset",
"high",
"compared",
"performance",
"accuracy",
"low",
"results",
"with",
"while",
"achieve"
],
[
"using",
"approach",
"an",
"which",
"input",
"features",
"classification",
"image",
"high",
"its"
],
[
"attacks",
"to",
"adversarial",
"training",
"the",
"we",
"robustness",
"attack",
"against",
"examples"
],
[
"paper",
"this",
"problem",
"approach",
"the",
"distribution",
"process",
"new",
"work",
"been"
],
[
"generative",
"fidelity",
"models",
"these",
"of",
"trained",
"language",
"performance",
"inference",
"have"
],
[
"the",
"detection",
"based",
"on",
"system",
"propose",
"proposed",
"have",
"feature",
"recognition"
],
[
"be",
"can",
"used",
"which",
"how",
"or",
"it",
"quantum",
"they",
"image"
],
[
"where",
"different",
"are",
"algorithms",
"these",
"they",
"their",
"more",
"when",
"the"
],
[
"the",
"graphs",
"and",
"node",
"nodes",
"graph",
"representation",
"gnns",
"we",
"which"
],
[
"deep",
"feature",
"proposed",
"prediction",
"method",
"methods",
"existing",
"datasets",
"classification",
"large"
],
[
"time",
"by",
"images",
"approach",
"the",
"feature",
"series",
"high",
"between",
"into"
],
[
"gradient",
"optimization",
"stochastic",
"and",
"policy",
"function",
"convergence",
"problems",
"convex",
"optimal"
],
[
"framework",
"results",
"our",
"approach",
"method",
"we",
"state",
"at",
"art",
"to"
],
[
"algorithms",
"algorithm",
"to",
"problem",
"matrix",
"clustering",
"proposed",
"linear",
"sample",
"rank"
],
[
"from",
"task",
"using",
"tasks",
"language",
"representations",
"different",
"knowledge",
"information",
"object"
]
] | 3.656396 | average_word_embeddings_glove.6B.300d | 0.563333 | 0.022051 | 0.15894 | 0.631439 |
42 | ArXiv ML Papers | NMF | 40 | [
[
"of",
"proposed",
"between",
"results",
"the",
"information",
"test",
"error",
"two",
"strategy"
],
[
"to",
"due",
"human",
"their",
"it",
"use",
"have",
"framework",
"used",
"approach"
],
[
"these",
"number",
"of",
"set",
"terms",
"new",
"large",
"analysis",
"art",
"study"
],
[
"both",
"including",
"and",
"research",
"two",
"accuracy",
"ml",
"between",
"their",
"study"
],
[
"propose",
"we",
"show",
"demonstrate",
"new",
"study",
"use",
"present",
"provide",
"introduce"
],
[
"been",
"work",
"or",
"has",
"in",
"many",
"applications",
"recent",
"the",
"have"
],
[
"analysis",
"test",
"also",
"the",
"using",
"for",
"used",
"regression",
"one",
"large"
],
[
"large",
"data",
"classification",
"real",
"synthetic",
"the",
"high",
"driven",
"and",
"techniques"
],
[
"framework",
"policy",
"learn",
"learning",
"machine",
"metric",
"reinforcement",
"supervised",
"agent",
"of"
],
[
"than",
"the",
"at",
"information",
"one",
"is",
"when",
"where",
"it",
"its"
],
[
"all",
"datasets",
"rely",
"previous",
"work",
"on",
"or",
"focus",
"real",
"dataset"
],
[
"that",
"show",
"also",
"more",
"than",
"not",
"results",
"representations",
"find",
"prior"
],
[
"neural",
"architectures",
"convolutional",
"have",
"physics",
"and",
"recurrent",
"over",
"cnns",
"networks"
],
[
"trained",
"model",
"and",
"inference",
"the",
"to",
"uncertainty",
"predictions",
"parameters",
"in"
],
[
"have",
"these",
"well",
"the",
"been",
"to",
"such",
"as",
"or",
"not"
],
[
"high",
"accuracy",
"results",
"the",
"dataset",
"with",
"compared",
"all",
"low",
"while"
],
[
"between",
"the",
"by",
"approach",
"into",
"filter",
"high",
"while",
"class",
"function"
],
[
"robustness",
"of",
"to",
"adversarial",
"the",
"against",
"attack",
"attacks",
"detection",
"robust"
],
[
"this",
"approach",
"paper",
"new",
"work",
"been",
"process",
"system",
"challenge",
"presents"
],
[
"information",
"two",
"which",
"not",
"propose",
"feature",
"has",
"only",
"the",
"of"
],
[
"problems",
"optimal",
"problem",
"samples",
"matrix",
"to",
"and",
"the",
"rank",
"where"
],
[
"can",
"how",
"it",
"used",
"be",
"or",
"to",
"quantum",
"they",
"systems"
],
[
"algorithms",
"these",
"more",
"are",
"when",
"where",
"they",
"some",
"their",
"different"
],
[
"graph",
"the",
"and",
"graphs",
"node",
"representation",
"nodes",
"gnns",
"we",
"clustering"
],
[
"proposed",
"propose",
"prediction",
"and",
"method",
"feature",
"optimization",
"noise",
"large",
"using"
],
[
"inference",
"fidelity",
"trained",
"models",
"generative",
"these",
"have",
"machine",
"prediction",
"language"
],
[
"function",
"convex",
"optimization",
"stochastic",
"and",
"gradient",
"policy",
"convergence",
"problems",
"descent"
],
[
"approach",
"framework",
"results",
"our",
"state",
"to",
"we",
"first",
"art",
"best"
],
[
"to",
"proposed",
"clustering",
"the",
"label",
"linear",
"algorithms",
"algorithm",
"matrix",
"sample"
],
[
"different",
"from",
"using",
"object",
"features",
"set",
"language",
"between",
"information",
"representations"
],
[
"based",
"detection",
"the",
"proposed",
"system",
"propose",
"information",
"systems",
"user",
"approaches"
],
[
"time",
"series",
"real",
"in",
"of",
"at",
"forecasting",
"temporal",
"world",
"state"
],
[
"datasets",
"methods",
"have",
"problems",
"existing",
"both",
"the",
"been",
"other",
"many"
],
[
"high",
"accuracy",
"state",
"art",
"performance",
"of",
"classification",
"better",
"using",
"achieve"
],
[
"generative",
"image",
"segmentation",
"gan",
"images",
"text",
"we",
"object",
"resolution",
"to"
],
[
"network",
"neural",
"to",
"layer",
"prediction",
"architecture",
"structure",
"nodes",
"convolutional",
"show"
],
[
"language",
"the",
"tasks",
"task",
"domain",
"knowledge",
"to",
"multi",
"different",
"source"
],
[
"input",
"an",
"approach",
"features",
"its",
"using",
"system",
"it",
"high",
"technique"
],
[
"in",
"loss",
"training",
"we",
"trained",
"during",
"pre",
"to",
"set",
"federated"
],
[
"has",
"deep",
"learning",
"feature",
"features",
"have",
"been",
"speech",
"accuracy",
"results"
]
] | 5.210731 | average_word_embeddings_glove.6B.300d | 0.545 | 0.015451 | 0.157201 | 0.652538 |
42 | ArXiv ML Papers | NMF | 50 | [
[
"results",
"of",
"proposed",
"the",
"between",
"test",
"error",
"was",
"strategy",
"conditions"
],
[
"due",
"their",
"use",
"framework",
"to",
"able",
"compared",
"often",
"human",
"end"
],
[
"number",
"these",
"set",
"new",
"terms",
"study",
"parameters",
"large",
"analysis",
"of"
],
[
"and",
"both",
"two",
"including",
"accuracy",
"ml",
"research",
"study",
"between",
"respectively"
],
[
"propose",
"we",
"show",
"use",
"present",
"new",
"study",
"demonstrate",
"provide",
"introduce"
],
[
"particular",
"in",
"dimensional",
"terms",
"or",
"context",
"many",
"work",
"addition",
"space"
],
[
"large",
"one",
"test",
"used",
"day",
"also",
"analysis",
"fast",
"for",
"regression"
],
[
"labeled",
"synthetic",
"real",
"driven",
"large",
"data",
"samples",
"process",
"framework",
"techniques"
],
[
"machine",
"metric",
"active",
"learn",
"learning",
"framework",
"of",
"supervised",
"reinforcement",
"to"
],
[
"than",
"show",
"not",
"find",
"more",
"results",
"that",
"representations",
"prior",
"also"
],
[
"focus",
"rely",
"datasets",
"all",
"on",
"work",
"or",
"trained",
"real",
"previous"
],
[
"where",
"more",
"these",
"when",
"they",
"are",
"their",
"some",
"new",
"different"
],
[
"convolutional",
"networks",
"neural",
"and",
"architectures",
"over",
"cnns",
"more",
"physics",
"weights"
],
[
"the",
"model",
"trained",
"predictions",
"parameters",
"inference",
"uncertainty",
"while",
"global",
"free"
],
[
"or",
"as",
"such",
"known",
"through",
"to",
"using",
"well",
"processing",
"continuous"
],
[
"dataset",
"with",
"while",
"high",
"accuracy",
"respect",
"compared",
"results",
"all",
"multiple"
],
[
"propose",
"the",
"has",
"prediction",
"noise",
"proposed",
"method",
"large",
"using",
"optimization"
],
[
"large",
"federated",
"during",
"pre",
"train",
"training",
"we",
"trained",
"set",
"loss"
],
[
"be",
"can",
"used",
"how",
"they",
"quantum",
"or",
"systems",
"the",
"such"
],
[
"knowledge",
"source",
"adaptation",
"domains",
"domain",
"target",
"the",
"transfer",
"datasets",
"proposed"
],
[
"presents",
"work",
"new",
"this",
"paper",
"process",
"challenge",
"approach",
"system",
"strategies"
],
[
"have",
"been",
"the",
"these",
"machine",
"has",
"or",
"such",
"research",
"their"
],
[
"filter",
"between",
"by",
"while",
"the",
"into",
"approach",
"images",
"function",
"dimensional"
],
[
"the",
"graphs",
"graph",
"node",
"and",
"nodes",
"we",
"gnns",
"representation",
"clustering"
],
[
"of",
"state",
"the",
"we",
"propose",
"datasets",
"art",
"results",
"over",
"approaches"
],
[
"generative",
"fidelity",
"models",
"trained",
"inference",
"prediction",
"these",
"language",
"high",
"at"
],
[
"system",
"input",
"its",
"high",
"framework",
"3d",
"an",
"using",
"technique",
"approach"
],
[
"results",
"framework",
"approach",
"first",
"best",
"our",
"at",
"work",
"experiments",
"only"
],
[
"algorithm",
"the",
"to",
"proposed",
"regret",
"clustering",
"sample",
"exchange",
"selection",
"at"
],
[
"different",
"using",
"set",
"between",
"object",
"representations",
"from",
"knowledge",
"the",
"language"
],
[
"on",
"proposed",
"based",
"the",
"propose",
"approaches",
"user",
"systems",
"flow",
"better"
],
[
"forecasting",
"series",
"real",
"each",
"time",
"world",
"at",
"temporal",
"of",
"hierarchical"
],
[
"methods",
"both",
"datasets",
"existing",
"other",
"linear",
"many",
"be",
"hierarchical",
"several"
],
[
"than",
"the",
"is",
"one",
"where",
"when",
"its",
"most",
"at",
"there"
],
[
"generative",
"gan",
"segmentation",
"text",
"images",
"image",
"resolution",
"trained",
"we",
"visual"
],
[
"network",
"neural",
"layer",
"structure",
"architecture",
"nodes",
"convolutional",
"show",
"prediction",
"to"
],
[
"attacks",
"adversarial",
"the",
"against",
"robustness",
"to",
"attack",
"of",
"and",
"robust"
],
[
"classifiers",
"label",
"the",
"accuracy",
"class",
"high",
"classification",
"classifier",
"classes",
"using"
],
[
"reinforcement",
"the",
"policy",
"rl",
"agent",
"reward",
"agents",
"policies",
"learning",
"optimal"
],
[
"learning",
"deep",
"accuracy",
"in",
"results",
"speech",
"paper",
"ensemble",
"recognition",
"dnn"
],
[
"convex",
"function",
"gradient",
"optimization",
"stochastic",
"convergence",
"problems",
"descent",
"order",
"non"
],
[
"achieve",
"their",
"performance",
"better",
"design",
"proposed",
"computing",
"different",
"cost",
"to"
],
[
"search",
"the",
"algorithms",
"variance",
"two",
"other",
"more",
"problems",
"to",
"than"
],
[
"of",
"selection",
"and",
"features",
"the",
"feature",
"to",
"attention",
"speech",
"results"
],
[
"different",
"multi",
"language",
"task",
"tasks",
"knowledge",
"each",
"prediction",
"meta",
"solving"
],
[
"propose",
"two",
"of",
"distribution",
"which",
"not",
"only",
"has",
"to",
"novel"
],
[
"its",
"but",
"has",
"be",
"and",
"not",
"also",
"at",
"it",
"well"
],
[
"the",
"low",
"problems",
"optimal",
"matrix",
"problem",
"rank",
"samples",
"distribution",
"linear"
],
[
"between",
"semantic",
"information",
"user",
"each",
"two",
"channel",
"multi",
"local",
"the"
],
[
"dataset",
"object",
"detection",
"detect",
"attacks",
"system",
"using",
"in",
"rate",
"the"
]
] | 6.391296 | average_word_embeddings_glove.6B.300d | 0.57 | 0.003507 | 0.147042 | 0.675528 |
42 | ArXiv ML Papers | LDA | 10 | [
[
"and",
"object",
"of",
"the",
"in",
"we",
"for",
"objects",
"on",
"shape"
],
[
"of",
"the",
"data",
"and",
"in",
"for",
"to",
"series",
"is",
"time"
],
[
"and",
"of",
"the",
"to",
"we",
"models",
"in",
"that",
"adversarial",
"training"
],
[
"and",
"data",
"to",
"for",
"of",
"model",
"in",
"the",
"is",
"prediction"
],
[
"is",
"to",
"for",
"the",
"in",
"we",
"of",
"and",
"that",
"with"
],
[
"to",
"and",
"detection",
"of",
"the",
"based",
"is",
"in",
"user",
"this"
],
[
"of",
"the",
"to",
"and",
"in",
"we",
"that",
"on",
"is",
"for"
],
[
"ad",
"establishing",
"too",
"devise",
"exhaustive",
"82",
"definition",
"note",
"go",
"want"
],
[
"to",
"we",
"of",
"is",
"the",
"in",
"and",
"for",
"that",
"this"
],
[
"of",
"to",
"the",
"and",
"is",
"we",
"learning",
"in",
"that",
"this"
]
] | 11.326483 | average_word_embeddings_glove.6B.300d | 0.37 | -0.069928 | 0.242925 | 0.555165 |
42 | ArXiv ML Papers | LDA | 20 | [
[
"object",
"objects",
"shape",
"on",
"compact",
"in",
"geometric",
"geometry",
"3d",
"and"
],
[
"of",
"signals",
"noise",
"quantum",
"frequency",
"the",
"to",
"an",
"in",
"machine"
],
[
"the",
"and",
"to",
"we",
"of",
"models",
"in",
"image",
"that",
"generative"
],
[
"for",
"to",
"traffic",
"data",
"and",
"the",
"we",
"of",
"devices",
"flow"
],
[
"to",
"of",
"and",
"the",
"in",
"we",
"for",
"that",
"is",
"with"
],
[
"attacks",
"black",
"to",
"box",
"and",
"the",
"attack",
"concepts",
"demand",
"of"
],
[
"of",
"the",
"and",
"we",
"to",
"in",
"networks",
"graph",
"neural",
"network"
],
[
"describes",
"composed",
"appealing",
"establishing",
"nas",
"correctness",
"occur",
"storing",
"integrating",
"ell_2"
],
[
"the",
"to",
"of",
"we",
"in",
"and",
"that",
"are",
"for",
"this"
],
[
"video",
"and",
"to",
"in",
"the",
"of",
"label",
"is",
"we",
"labels"
],
[
"and",
"for",
"on",
"in",
"of",
"to",
"with",
"is",
"the",
"we"
],
[
"data",
"to",
"and",
"in",
"the",
"of",
"for",
"is",
"that",
"we"
],
[
"for",
"gradient",
"and",
"optimization",
"convergence",
"stochastic",
"convex",
"method",
"we",
"descent"
],
[
"data",
"in",
"the",
"of",
"and",
"with",
"to",
"we",
"stream",
"identity"
],
[
"and",
"to",
"the",
"we",
"of",
"in",
"is",
"on",
"learning",
"this"
],
[
"shape",
"design",
"sentence",
"sentences",
"optical",
"assisted",
"space",
"translation",
"interesting",
"task"
],
[
"of",
"the",
"we",
"adversarial",
"that",
"in",
"and",
"training",
"to",
"on"
],
[
"on",
"the",
"we",
"learning",
"of",
"task",
"and",
"to",
"that",
"in"
],
[
"planning",
"private",
"bias",
"the",
"that",
"and",
"ensemble",
"vae",
"of",
"variance"
],
[
"and",
"of",
"this",
"in",
"to",
"the",
"that",
"we",
"for",
"is"
]
] | 15.574985 | average_word_embeddings_glove.6B.300d | 0.415 | -0.06419 | 0.203814 | 0.61154 |
42 | ArXiv ML Papers | LDA | 30 | [
[
"objects",
"shape",
"the",
"3d",
"in",
"geometric",
"and",
"sequences",
"geometry",
"of"
],
[
"and",
"the",
"of",
"text",
"documents",
"classification",
"boosting",
"for",
"day",
"multi"
],
[
"and",
"that",
"to",
"visual",
"of",
"models",
"in",
"the",
"we",
"image"
],
[
"and",
"for",
"inference",
"the",
"we",
"to",
"of",
"data",
"in",
"posterior"
],
[
"of",
"to",
"the",
"and",
"we",
"in",
"that",
"for",
"is",
"this"
],
[
"attacks",
"the",
"box",
"attack",
"adversarial",
"to",
"black",
"detection",
"and",
"of"
],
[
"the",
"of",
"and",
"we",
"to",
"networks",
"graph",
"network",
"in",
"neural"
],
[
"personalized",
"perspective",
"performed",
"performing",
"performs",
"period",
"periodic",
"periods",
"person",
"personal"
],
[
"in",
"to",
"and",
"of",
"we",
"the",
"that",
"is",
"for",
"on"
],
[
"to",
"the",
"of",
"label",
"in",
"and",
"metric",
"video",
"is",
"labels"
],
[
"the",
"of",
"and",
"to",
"on",
"for",
"with",
"in",
"we",
"is"
],
[
"that",
"for",
"in",
"data",
"the",
"to",
"is",
"and",
"of",
"this"
],
[
"optimization",
"gradient",
"convergence",
"stochastic",
"convex",
"for",
"and",
"descent",
"method",
"we"
],
[
"kernel",
"in",
"the",
"of",
"and",
"kernels",
"subspace",
"distance",
"we",
"with"
],
[
"is",
"we",
"of",
"in",
"to",
"and",
"the",
"on",
"that",
"learning"
],
[
"action",
"actions",
"continuous",
"reinforcement",
"discrete",
"spaces",
"domains",
"control",
"policy",
"complex"
],
[
"we",
"that",
"on",
"and",
"in",
"for",
"of",
"the",
"to",
"this"
],
[
"and",
"learning",
"the",
"of",
"to",
"we",
"tasks",
"task",
"agent",
"that"
],
[
"variational",
"and",
"the",
"bias",
"latent",
"that",
"of",
"al",
"models",
"et"
],
[
"and",
"this",
"the",
"we",
"that",
"in",
"our",
"of",
"to",
"for"
],
[
"accuracies",
"measurement",
"and",
"the",
"of",
"additive",
"that",
"perturbations",
"we",
"dnns"
],
[
"we",
"of",
"to",
"and",
"in",
"the",
"generative",
"adversarial",
"gan",
"generation"
],
[
"and",
"in",
"of",
"to",
"privacy",
"the",
"data",
"learning",
"is",
"quantum"
],
[
"the",
"image",
"and",
"to",
"of",
"for",
"segmentation",
"in",
"with",
"on"
],
[
"online",
"distributed",
"to",
"of",
"the",
"communication",
"agents",
"planning",
"learning",
"and"
],
[
"personalized",
"perspective",
"performed",
"performing",
"performs",
"period",
"periodic",
"periods",
"person",
"personal"
],
[
"in",
"social",
"ai",
"the",
"of",
"to",
"user",
"users",
"and",
"media"
],
[
"the",
"model",
"is",
"and",
"in",
"to",
"proposed",
"channel",
"of",
"loss"
],
[
"scaling",
"depth",
"compromising",
"strategies",
"estimation",
"types",
"width",
"examples",
"domain",
"various"
],
[
"super",
"images",
"conditional",
"realistic",
"resolution",
"image",
"pixel",
"conditioning",
"details",
"model"
]
] | 16.76239 | average_word_embeddings_glove.6B.300d | 0.426667 | -0.049465 | 0.188912 | 0.657253 |
42 | ArXiv ML Papers | LDA | 40 | [
[
"and",
"on",
"geometric",
"left",
"the",
"frac",
"in",
"right",
"varepsilon",
"this"
],
[
"an",
"the",
"text",
"in",
"techniques",
"of",
"and",
"documents",
"implementations",
"library"
],
[
"in",
"translation",
"and",
"we",
"the",
"to",
"that",
"of",
"with",
"machine"
],
[
"to",
"the",
"in",
"and",
"for",
"models",
"of",
"we",
"free",
"reports"
],
[
"of",
"we",
"the",
"and",
"to",
"in",
"for",
"that",
"is",
"with"
],
[
"detection",
"attacks",
"black",
"the",
"box",
"of",
"adversarial",
"to",
"attack",
"and"
],
[
"of",
"graph",
"and",
"we",
"the",
"to",
"in",
"that",
"on",
"networks"
],
[
"personalized",
"perspective",
"performed",
"performing",
"performs",
"period",
"periodic",
"periods",
"person",
"personal"
],
[
"the",
"of",
"and",
"in",
"to",
"that",
"on",
"we",
"are",
"this"
],
[
"to",
"and",
"of",
"robot",
"the",
"style",
"in",
"we",
"that",
"al"
],
[
"the",
"in",
"for",
"to",
"time",
"and",
"of",
"on",
"models",
"prediction"
],
[
"to",
"the",
"for",
"and",
"of",
"in",
"that",
"data",
"is",
"learning"
],
[
"to",
"analysis",
"and",
"unstructured",
"operation",
"big",
"acceleration",
"data",
"classification",
"fusion"
],
[
"functions",
"in",
"convex",
"and",
"the",
"of",
"imaging",
"condition",
"methods",
"we"
],
[
"we",
"the",
"in",
"is",
"and",
"learning",
"to",
"of",
"on",
"this"
],
[
"actions",
"spaces",
"extends",
"domains",
"half",
"composed",
"robot",
"platform",
"action",
"combining"
],
[
"that",
"of",
"we",
"to",
"the",
"in",
"training",
"by",
"adversarial",
"robustness"
],
[
"the",
"and",
"of",
"agents",
"that",
"agent",
"we",
"to",
"on",
"task"
],
[
"the",
"variance",
"gradient",
"and",
"batch",
"stochastic",
"we",
"descent",
"that",
"boosting"
],
[
"in",
"to",
"that",
"the",
"user",
"of",
"we",
"and",
"for",
"users"
],
[
"sparse",
"latent",
"inference",
"the",
"end",
"structures",
"structure",
"structural",
"to",
"dynamic"
],
[
"to",
"and",
"we",
"the",
"of",
"in",
"human",
"design",
"research",
"flow"
],
[
"we",
"in",
"and",
"for",
"of",
"the",
"with",
"to",
"models",
"on"
],
[
"resolution",
"of",
"and",
"in",
"to",
"for",
"the",
"sequence",
"on",
"event"
],
[
"data",
"object",
"of",
"objects",
"detection",
"in",
"the",
"processes",
"new",
"stream"
],
[
"meta",
"correlated",
"games",
"sum",
"we",
"for",
"agents",
"and",
"solving",
"solvers"
],
[
"models",
"generative",
"conditional",
"autoregressive",
"samples",
"likelihood",
"of",
"we",
"procedure",
"that"
],
[
"is",
"demand",
"to",
"and",
"network",
"the",
"energy",
"of",
"for",
"in"
],
[
"predictive",
"selection",
"performance",
"experimental",
"for",
"detection",
"outlier",
"feature",
"bayesian",
"the"
],
[
"personalized",
"perspective",
"performed",
"performing",
"performs",
"period",
"periodic",
"periods",
"person",
"personal"
],
[
"we",
"and",
"is",
"to",
"on",
"learning",
"in",
"of",
"the",
"data"
],
[
"the",
"noise",
"visual",
"of",
"reasoning",
"and",
"in",
"codes",
"to",
"that"
],
[
"dropout",
"rnn",
"hidden",
"units",
"recurrent",
"rnns",
"language",
"sequential",
"generalization",
"forces"
],
[
"bias",
"loss",
"the",
"for",
"face",
"and",
"in",
"cross",
"of",
"recognition"
],
[
"concepts",
"robust",
"error",
"depth",
"neural",
"we",
"explanations",
"the",
"networks",
"width"
],
[
"in",
"the",
"to",
"training",
"is",
"on",
"of",
"we",
"and",
"with"
],
[
"the",
"in",
"and",
"to",
"of",
"we",
"for",
"is",
"that",
"with"
],
[
"of",
"and",
"the",
"in",
"is",
"we",
"to",
"for",
"that",
"on"
],
[
"the",
"clustering",
"and",
"in",
"we",
"cluster",
"to",
"of",
"for",
"are"
],
[
"of",
"to",
"gan",
"the",
"feedback",
"that",
"in",
"agent",
"and",
"interactive"
]
] | 17.04971 | average_word_embeddings_glove.6B.300d | 0.405 | -0.061252 | 0.182069 | 0.630669 |
42 | ArXiv ML Papers | LDA | 50 | [
[
"and",
"robust",
"in",
"of",
"we",
"error",
"the",
"predictor",
"on",
"to"
],
[
"signal",
"the",
"signals",
"frequency",
"spectral",
"separation",
"which",
"for",
"of",
"and"
],
[
"rank",
"tensor",
"translation",
"and",
"to",
"the",
"we",
"local",
"low",
"in"
],
[
"energy",
"and",
"to",
"for",
"free",
"we",
"algorithm",
"the",
"in",
"of"
],
[
"with",
"of",
"in",
"the",
"that",
"to",
"we",
"and",
"is",
"for"
],
[
"flow",
"the",
"traffic",
"recognition",
"system",
"detection",
"based",
"flows",
"vehicle",
"is"
],
[
"in",
"graph",
"the",
"that",
"to",
"and",
"of",
"we",
"for",
"on"
],
[
"personalized",
"perspective",
"performed",
"performing",
"performs",
"period",
"periodic",
"periods",
"person",
"personal"
],
[
"of",
"and",
"the",
"in",
"to",
"we",
"that",
"language",
"for",
"is"
],
[
"to",
"the",
"and",
"of",
"be",
"can",
"social",
"in",
"we",
"that"
],
[
"and",
"the",
"for",
"of",
"to",
"in",
"with",
"model",
"an",
"on"
],
[
"of",
"the",
"in",
"to",
"is",
"data",
"for",
"and",
"this",
"we"
],
[
"we",
"of",
"the",
"convergence",
"strongly",
"convex",
"our",
"for",
"complexity",
"condition"
],
[
"forward",
"feed",
"of",
"learning",
"in",
"rules",
"we",
"active",
"to",
"the"
],
[
"of",
"the",
"is",
"we",
"in",
"and",
"to",
"that",
"this",
"on"
],
[
"personalized",
"perspective",
"performed",
"performing",
"performs",
"period",
"periodic",
"periods",
"person",
"personal"
],
[
"to",
"are",
"the",
"ground",
"for",
"of",
"in",
"that",
"on",
"we"
],
[
"the",
"to",
"we",
"of",
"calibration",
"meta",
"and",
"that",
"probabilistic",
"distributions"
],
[
"forgetting",
"concept",
"learners",
"learning",
"and",
"the",
"in",
"that",
"for",
"of"
],
[
"we",
"in",
"of",
"time",
"for",
"that",
"to",
"the",
"and",
"our"
],
[
"multiple",
"community",
"perturbations",
"dnns",
"are",
"to",
"that",
"the",
"communities",
"and"
],
[
"in",
"of",
"the",
"and",
"we",
"that",
"are",
"these",
"to",
"with"
],
[
"and",
"to",
"communication",
"data",
"for",
"distributed",
"the",
"of",
"in",
"federated"
],
[
"and",
"to",
"in",
"causal",
"sequence",
"gan",
"gans",
"for",
"the",
"of"
],
[
"in",
"and",
"online",
"data",
"variable",
"we",
"to",
"the",
"of",
"learning"
],
[
"sum",
"joint",
"concept",
"games",
"correlated",
"meta",
"studied",
"outside",
"response",
"general"
],
[
"auc",
"device",
"devices",
"operators",
"healthcare",
"resource",
"low",
"to",
"and",
"the"
],
[
"of",
"and",
"quantum",
"to",
"in",
"is",
"the",
"topic",
"we",
"topics"
],
[
"variance",
"oracle",
"reduction",
"variant",
"accelerated",
"acceleration",
"popular",
"convex",
"known",
"effective"
],
[
"personalized",
"perspective",
"performed",
"performing",
"performs",
"period",
"periodic",
"periods",
"person",
"personal"
],
[
"the",
"to",
"of",
"and",
"in",
"model",
"we",
"this",
"learning",
"is"
],
[
"noise",
"the",
"of",
"visual",
"reasoning",
"and",
"to",
"software",
"in",
"codes"
],
[
"random",
"the",
"left",
"varepsilon",
"forest",
"frac",
"right",
"gpu",
"programming",
"code"
],
[
"cross",
"of",
"loss",
"the",
"is",
"in",
"bias",
"for",
"face",
"and"
],
[
"realistic",
"conditional",
"resolution",
"images",
"image",
"high",
"pixel",
"super",
"samples",
"pixels"
],
[
"to",
"in",
"the",
"of",
"and",
"we",
"on",
"data",
"is",
"for"
],
[
"to",
"and",
"is",
"model",
"we",
"that",
"for",
"the",
"in",
"of"
],
[
"data",
"in",
"and",
"that",
"the",
"is",
"of",
"to",
"we",
"for"
],
[
"and",
"the",
"from",
"in",
"ai",
"of",
"to",
"safety",
"nn",
"were"
],
[
"agents",
"reward",
"to",
"control",
"the",
"reinforcement",
"of",
"agent",
"in",
"learning"
],
[
"problems",
"method",
"gradient",
"methods",
"stochastic",
"of",
"descent",
"and",
"optimization",
"we"
],
[
"variational",
"in",
"of",
"inference",
"the",
"and",
"we",
"to",
"al",
"is"
],
[
"label",
"the",
"channel",
"clustering",
"is",
"with",
"multi",
"of",
"labels",
"semantic"
],
[
"the",
"in",
"performance",
"to",
"of",
"and",
"on",
"is",
"network",
"with"
],
[
"of",
"in",
"bayesian",
"is",
"we",
"the",
"for",
"distribution",
"posterior",
"and"
],
[
"language",
"models",
"the",
"representations",
"in",
"of",
"to",
"and",
"we",
"on"
],
[
"in",
"the",
"on",
"to",
"of",
"we",
"learning",
"policy",
"and",
"reinforcement"
],
[
"the",
"latent",
"generation",
"diffusion",
"to",
"of",
"text",
"in",
"for",
"and"
],
[
"adversarial",
"the",
"to",
"attacks",
"training",
"that",
"attack",
"robustness",
"in",
"we"
],
[
"we",
"to",
"in",
"and",
"the",
"of",
"that",
"on",
"for",
"is"
]
] | 18.716448 | average_word_embeddings_glove.6B.300d | 0.376 | -0.060056 | 0.190723 | 0.641459 |
42 | ArXiv ML Papers | Top2Vec | 10 | [
[
"using",
"example",
"methods",
"use",
"specific",
"required",
"particular",
"instance",
"systems",
"method"
],
[
"possible",
"use",
"example",
"methods",
"particular",
"specific",
"similar",
"using",
"instance",
"need"
],
[
"example",
"particular",
"instance",
"use",
"specific",
"certain",
"different",
"using",
"similar",
"context"
],
[
"different",
"particular",
"methods",
"use",
"example",
"certain",
"instance",
"need",
"specific",
"using"
],
[
"particular",
"instance",
"use",
"example",
"specific",
"provide",
"need",
"using",
"different",
"addition"
],
[
"particular",
"function",
"example",
"specific",
"instance",
"methods",
"using",
"method",
"parameters",
"types"
],
[
"example",
"using",
"particular",
"instance",
"specific",
"use",
"different",
"methods",
"data",
"types"
],
[
"instance",
"example",
"using",
"particular",
"data",
"use",
"specific",
"provide",
"different",
"similar"
],
[
"example",
"particular",
"methods",
"instance",
"method",
"specific",
"using",
"different",
"function",
"certain"
],
[
"methods",
"example",
"method",
"function",
"analysis",
"using",
"parameters",
"algorithms",
"particular",
"processes"
]
] | 8.997187 | average_word_embeddings_glove.6B.300d | 0.25 | -0.098371 | 0.253819 | 0.7812 |
42 | ArXiv ML Papers | Top2Vec | 20 | [
[
"example",
"using",
"use",
"systems",
"methods",
"data",
"required",
"specific",
"particular",
"provide"
],
[
"example",
"instance",
"specific",
"using",
"methods",
"systems",
"use",
"particular",
"types",
"method"
],
[
"possible",
"use",
"example",
"methods",
"particular",
"specific",
"similar",
"using",
"instance",
"need"
],
[
"example",
"particular",
"instance",
"use",
"specific",
"certain",
"different",
"using",
"similar",
"context"
],
[
"methods",
"particular",
"specific",
"example",
"instance",
"using",
"different",
"types",
"function",
"certain"
],
[
"different",
"particular",
"methods",
"use",
"example",
"certain",
"instance",
"need",
"specific",
"using"
],
[
"particular",
"instance",
"use",
"example",
"specific",
"provide",
"need",
"using",
"different",
"addition"
],
[
"particular",
"function",
"example",
"specific",
"instance",
"methods",
"using",
"method",
"parameters",
"types"
],
[
"example",
"using",
"particular",
"instance",
"specific",
"use",
"different",
"methods",
"data",
"types"
],
[
"instance",
"particular",
"example",
"result",
"data",
"using",
"use",
"methods",
"different",
"need"
],
[
"use",
"using",
"example",
"particular",
"provide",
"specific",
"instance",
"addition",
"similar",
"different"
],
[
"methods",
"example",
"using",
"method",
"particular",
"specific",
"instance",
"analysis",
"use",
"different"
],
[
"particular",
"methods",
"method",
"using",
"example",
"specific",
"analysis",
"instance",
"different",
"types"
],
[
"specific",
"particular",
"given",
"certain",
"example",
"function",
"method",
"instance",
"means",
"result"
],
[
"methods",
"example",
"method",
"function",
"analysis",
"using",
"parameters",
"algorithms",
"particular",
"processes"
]
] | 8.604245 | average_word_embeddings_glove.6B.300d | 0.186667 | -0.100363 | 0.247966 | 0.774992 |
42 | ArXiv ML Papers | Top2Vec | 30 | [
[
"example",
"using",
"use",
"systems",
"methods",
"data",
"required",
"specific",
"particular",
"provide"
],
[
"example",
"instance",
"specific",
"using",
"methods",
"systems",
"use",
"particular",
"types",
"method"
],
[
"possible",
"use",
"example",
"methods",
"particular",
"specific",
"similar",
"using",
"instance",
"need"
],
[
"example",
"particular",
"instance",
"use",
"specific",
"certain",
"different",
"using",
"similar",
"context"
],
[
"methods",
"particular",
"specific",
"example",
"instance",
"using",
"different",
"types",
"function",
"certain"
],
[
"different",
"particular",
"methods",
"use",
"example",
"certain",
"instance",
"need",
"specific",
"using"
],
[
"particular",
"instance",
"use",
"example",
"specific",
"provide",
"need",
"using",
"different",
"addition"
],
[
"particular",
"function",
"example",
"specific",
"instance",
"methods",
"using",
"method",
"parameters",
"types"
],
[
"example",
"using",
"particular",
"instance",
"specific",
"use",
"different",
"methods",
"data",
"types"
],
[
"instance",
"particular",
"example",
"result",
"data",
"using",
"use",
"methods",
"different",
"need"
],
[
"use",
"using",
"example",
"particular",
"provide",
"specific",
"instance",
"addition",
"similar",
"different"
],
[
"methods",
"example",
"using",
"method",
"particular",
"specific",
"instance",
"analysis",
"use",
"different"
],
[
"particular",
"methods",
"method",
"using",
"example",
"specific",
"analysis",
"instance",
"different",
"types"
],
[
"specific",
"particular",
"given",
"certain",
"example",
"function",
"method",
"instance",
"means",
"result"
],
[
"methods",
"example",
"method",
"function",
"analysis",
"using",
"parameters",
"algorithms",
"particular",
"processes"
]
] | 8.604781 | average_word_embeddings_glove.6B.300d | 0.186667 | -0.100363 | 0.247966 | 0.780283 |
42 | ArXiv ML Papers | Top2Vec | 40 | [
[
"example",
"using",
"use",
"systems",
"methods",
"data",
"required",
"specific",
"particular",
"provide"
],
[
"example",
"instance",
"specific",
"using",
"methods",
"systems",
"use",
"particular",
"types",
"method"
],
[
"possible",
"use",
"example",
"methods",
"particular",
"specific",
"similar",
"using",
"instance",
"need"
],
[
"example",
"particular",
"instance",
"use",
"specific",
"certain",
"different",
"using",
"similar",
"context"
],
[
"methods",
"particular",
"specific",
"example",
"instance",
"using",
"different",
"types",
"function",
"certain"
],
[
"different",
"particular",
"methods",
"use",
"example",
"certain",
"instance",
"need",
"specific",
"using"
],
[
"particular",
"instance",
"use",
"example",
"specific",
"provide",
"need",
"using",
"different",
"addition"
],
[
"particular",
"function",
"example",
"specific",
"instance",
"methods",
"using",
"method",
"parameters",
"types"
],
[
"example",
"using",
"particular",
"instance",
"specific",
"use",
"different",
"methods",
"data",
"types"
],
[
"instance",
"particular",
"example",
"result",
"data",
"using",
"use",
"methods",
"different",
"need"
],
[
"use",
"using",
"example",
"particular",
"provide",
"specific",
"instance",
"addition",
"similar",
"different"
],
[
"methods",
"example",
"using",
"method",
"particular",
"specific",
"instance",
"analysis",
"use",
"different"
],
[
"particular",
"methods",
"method",
"using",
"example",
"specific",
"analysis",
"instance",
"different",
"types"
],
[
"specific",
"particular",
"given",
"certain",
"example",
"function",
"method",
"instance",
"means",
"result"
],
[
"methods",
"example",
"method",
"function",
"analysis",
"using",
"parameters",
"algorithms",
"particular",
"processes"
]
] | 8.980426 | average_word_embeddings_glove.6B.300d | 0.186667 | -0.100363 | 0.247966 | 0.77516 |
42 | ArXiv ML Papers | Top2Vec | 50 | [
[
"example",
"using",
"use",
"systems",
"methods",
"data",
"required",
"specific",
"particular",
"provide"
],
[
"example",
"instance",
"specific",
"using",
"methods",
"systems",
"use",
"particular",
"types",
"method"
],
[
"possible",
"use",
"example",
"methods",
"particular",
"specific",
"similar",
"using",
"instance",
"need"
],
[
"example",
"particular",
"instance",
"use",
"specific",
"certain",
"different",
"using",
"similar",
"context"
],
[
"methods",
"particular",
"specific",
"example",
"instance",
"using",
"different",
"types",
"function",
"certain"
],
[
"different",
"particular",
"methods",
"use",
"example",
"certain",
"instance",
"need",
"specific",
"using"
],
[
"particular",
"instance",
"use",
"example",
"specific",
"provide",
"need",
"using",
"different",
"addition"
],
[
"particular",
"function",
"example",
"specific",
"instance",
"methods",
"using",
"method",
"parameters",
"types"
],
[
"example",
"using",
"particular",
"instance",
"specific",
"use",
"different",
"methods",
"data",
"types"
],
[
"instance",
"particular",
"example",
"result",
"data",
"using",
"use",
"methods",
"different",
"need"
],
[
"use",
"using",
"example",
"particular",
"provide",
"specific",
"instance",
"addition",
"similar",
"different"
],
[
"methods",
"example",
"using",
"method",
"particular",
"specific",
"instance",
"analysis",
"use",
"different"
],
[
"particular",
"methods",
"method",
"using",
"example",
"specific",
"analysis",
"instance",
"different",
"types"
],
[
"specific",
"particular",
"given",
"certain",
"example",
"function",
"method",
"instance",
"means",
"result"
],
[
"methods",
"example",
"method",
"function",
"analysis",
"using",
"parameters",
"algorithms",
"particular",
"processes"
]
] | 8.776014 | average_word_embeddings_glove.6B.300d | 0.186667 | -0.100363 | 0.247966 | 0.772195 |
42 | ArXiv ML Papers | S³ | 10 | [
[
"similarities",
"relationship",
"phenomena",
"expression",
"character",
"distinct",
"influence",
"understanding",
"literature",
"expressive"
],
[
"images",
"sensors",
"audio",
"camera",
"digital",
"sensor",
"video",
"devices",
"equipped",
"acoustic"
],
[
"topology",
"vertices",
"graph",
"graphs",
"orthogonal",
"topological",
"symmetric",
"vertex",
"tensor",
"nodes"
],
[
"statistical",
"model",
"forecast",
"models",
"forecasting",
"projections",
"forecasts",
"growth",
"estimates",
"prediction"
],
[
"molecular",
"technique",
"method",
"algorithm",
"physics",
"techniques",
"neural",
"equations",
"quantum",
"geometry"
],
[
"engineering",
"research",
"languages",
"biology",
"science",
"library",
"physics",
"foundation",
"literature",
"published"
],
[
"time",
"points",
"number",
"10",
"minutes",
"half",
"second",
"point",
"average",
"rate"
],
[
"model",
"style",
"version",
"mode",
"graphical",
"simplicity",
"models",
"sound",
"image",
"compact"
],
[
"71",
"min",
"rl",
"91",
"vs",
"averaged",
"83",
"82",
"88",
"94"
],
[
"energy",
"infrastructure",
"alleviate",
"forces",
"accelerate",
"stability",
"network",
"pipelines",
"relations",
"operations"
]
] | 1.009519 | average_word_embeddings_glove.6B.300d | 0.96 | -0.262821 | 0.270151 | 0.884171 |
42 | ArXiv ML Papers | S³ | 20 | [
[
"settings",
"predefined",
"pairwise",
"combinations",
"vary",
"sizes",
"randomly",
"depending",
"varying",
"different"
],
[
"forecasts",
"uncertainties",
"catastrophic",
"forecast",
"predicting",
"forecasting",
"weather",
"traffic",
"predictions",
"periods"
],
[
"attack",
"agents",
"attacks",
"deployed",
"prevent",
"forces",
"attacker",
"armed",
"enforce",
"bandit"
],
[
"calibration",
"controller",
"evaluation",
"parameter",
"data",
"parameters",
"classifier",
"module",
"gps",
"configuration"
],
[
"sum",
"right",
"equation",
"integer",
"function",
"sentence",
"polynomial",
"word",
"cause",
"zero"
],
[
"network",
"neural",
"networks",
"neurons",
"channel",
"channels",
"connections",
"backpropagation",
"wireless",
"cells"
],
[
"problems",
"difficulties",
"loss",
"poor",
"performance",
"losses",
"training",
"suffer",
"severe",
"lack"
],
[
"branch",
"classification",
"classes",
"priori",
"training",
"groups",
"centers",
"areas",
"communities",
"rigorous"
],
[
"relations",
"implementation",
"policy",
"framework",
"stability",
"proposed",
"compliance",
"implementing",
"objectives",
"comprehensive"
],
[
"goal",
"target",
"consensus",
"algorithm",
"conclusion",
"correct",
"prediction",
"algorithms",
"scenario",
"precise"
],
[
"diseases",
"cancer",
"clinical",
"protein",
"brain",
"disease",
"cells",
"patients",
"genetic",
"diagnosis"
],
[
"vertex",
"graph",
"node",
"graphs",
"vertices",
"nodes",
"convex",
"nonconvex",
"topological",
"edges"
],
[
"consists",
"joint",
"code",
"named",
"model",
"unit",
"special",
"assigned",
"consisting",
"functions"
],
[
"data",
"statistical",
"estimate",
"compression",
"textual",
"layer",
"paper",
"text",
"texts",
"derived"
],
[
"technology",
"imaging",
"efficiency",
"energy",
"maximize",
"enhancing",
"efficient",
"quality",
"technologies",
"products"
],
[
"gradient",
"signals",
"subspace",
"signal",
"relaxation",
"https",
"filtering",
"frequencies",
"denoising",
"frequency"
],
[
"negative",
"biased",
"conducted",
"observations",
"positive",
"evidence",
"results",
"studies",
"relationship",
"bias"
],
[
"web",
"twitter",
"online",
"internet",
"users",
"server",
"user",
"website",
"databases",
"software"
],
[
"stochastic",
"iterative",
"bayesian",
"dirichlet",
"optimization",
"nonlinear",
"multivariate",
"algorithm",
"variational",
"minimization"
],
[
"variables",
"behaviors",
"mechanisms",
"processes",
"factors",
"specific",
"minimize",
"patterns",
"reduce",
"function"
]
] | 0.988541 | average_word_embeddings_glove.6B.300d | 0.975 | -0.269014 | 0.254041 | 0.877098 |
42 | ArXiv ML Papers | S³ | 30 | [
[
"2019",
"2017",
"autoregressive",
"28",
"19",
"2018",
"preliminary",
"14",
"final",
"16"
],
[
"ensemble",
"elements",
"noisy",
"initialization",
"constraints",
"dimension",
"architectural",
"element",
"complexity",
"compositional"
],
[
"false",
"questions",
"documents",
"explanations",
"explanation",
"answers",
"answer",
"valid",
"details",
"proof"
],
[
"points",
"score",
"scoring",
"remarkable",
"highest",
"impressive",
"gained",
"goals",
"region",
"best"
],
[
"literature",
"studies",
"science",
"research",
"physics",
"biology",
"scientific",
"topics",
"survey",
"studied"
],
[
"online",
"random",
"videos",
"video",
"randomly",
"samples",
"sample",
"music",
"combinations",
"interesting"
],
[
"classic",
"designs",
"design",
"style",
"adaptation",
"techniques",
"novel",
"topics",
"feature",
"series"
],
[
"backpropagation",
"network",
"convolutional",
"neural",
"networks",
"channels",
"neurons",
"connections",
"bayesian",
"channel"
],
[
"measured",
"smooth",
"process",
"consensus",
"gradient",
"positive",
"responses",
"results",
"solution",
"candidates"
],
[
"consumption",
"online",
"price",
"weighted",
"share",
"weaker",
"market",
"outperform",
"rates",
"trend"
],
[
"bayesian",
"estimator",
"regression",
"multivariate",
"variational",
"bayes",
"iterative",
"stochastic",
"inference",
"probabilistic"
],
[
"conditioning",
"standard",
"safety",
"noise",
"discrimination",
"overhead",
"practice",
"strict",
"problems",
"privacy"
],
[
"predictions",
"forecasts",
"forecasting",
"forecast",
"weather",
"traffic",
"catastrophic",
"flows",
"maps",
"prediction"
],
[
"molecular",
"detector",
"particle",
"quantum",
"physics",
"energy",
"magnetic",
"machines",
"chemical",
"sensors"
],
[
"width",
"parameter",
"posterior",
"length",
"density",
"approximately",
"approximate",
"depth",
"variance",
"maximum"
],
[
"performances",
"performance",
"production",
"produced",
"factorization",
"ensemble",
"performs",
"performed",
"dirichlet",
"audio"
],
[
"evaluations",
"positive",
"conditional",
"performances",
"exhibits",
"efficacy",
"material",
"gains",
"observation",
"uniformly"
],
[
"temporal",
"sources",
"nets",
"agent",
"modes",
"inefficient",
"encoding",
"averaging",
"deep",
"storage"
],
[
"gains",
"tremendous",
"confidence",
"bit",
"deeper",
"uncertainty",
"deep",
"enormous",
"sentiment",
"slow"
],
[
"functionality",
"user",
"kernel",
"software",
"interface",
"hardware",
"graphical",
"platforms",
"server",
"runtime"
],
[
"built",
"models",
"size",
"model",
"version",
"fit",
"modeling",
"engine",
"car",
"design"
],
[
"word",
"languages",
"english",
"speech",
"language",
"words",
"translation",
"text",
"texts",
"dictionary"
],
[
"threat",
"malicious",
"targeting",
"attacks",
"malware",
"attacker",
"armed",
"attack",
"deployed",
"forces"
],
[
"occur",
"events",
"variables",
"perceptual",
"behaviors",
"cognitive",
"temporal",
"phenomena",
"discrimination",
"stochastic"
],
[
"population",
"optimal",
"randomly",
"sampling",
"genetic",
"locations",
"location",
"inference",
"size",
"variability"
],
[
"errors",
"languages",
"error",
"classification",
"intermediate",
"class",
"classes",
"requirements",
"variables",
"predictors"
],
[
"mathematical",
"quantum",
"computational",
"physics",
"computer",
"cognitive",
"computation",
"algorithms",
"fourier",
"asymptotic"
],
[
"approximations",
"computationally",
"diverse",
"tasks",
"efficiently",
"fewer",
"perform",
"pairwise",
"execute",
"gpus"
],
[
"substantially",
"widespread",
"increased",
"increasing",
"structural",
"compression",
"reduction",
"costs",
"considerable",
"substantial"
],
[
"transfer",
"thousands",
"block",
"process",
"prevent",
"property",
"construction",
"beta",
"free",
"closed"
]
] | 1.507303 | average_word_embeddings_glove.6B.300d | 0.926667 | -0.318092 | 0.233842 | 0.892498 |
42 | ArXiv ML Papers | S³ | 40 | [
[
"generator",
"variables",
"paradigm",
"adversarial",
"agent",
"power",
"latent",
"node",
"generation",
"dirichlet"
],
[
"vertices",
"edges",
"nodes",
"node",
"graphs",
"graph",
"vertex",
"benchmark",
"embedding",
"weighted"
],
[
"differentiable",
"convex",
"integer",
"vector",
"vectors",
"subset",
"polynomial",
"euclidean",
"linearly",
"space"
],
[
"cognitive",
"functions",
"behavior",
"activity",
"behaviors",
"dynamics",
"brain",
"behaviour",
"function",
"dynamical"
],
[
"deeper",
"concrete",
"differences",
"points",
"depth",
"wide",
"yield",
"deep",
"broad",
"posterior"
],
[
"distribution",
"universal",
"music",
"criterion",
"shelf",
"sample",
"generative",
"criteria",
"quality",
"satisfying"
],
[
"malicious",
"classification",
"detect",
"detecting",
"detection",
"semantic",
"malware",
"weighted",
"categories",
"denoising"
],
[
"predicting",
"forecasts",
"trends",
"forecast",
"scenarios",
"predictions",
"forecasting",
"predict",
"weather",
"global"
],
[
"sentence",
"penalty",
"attention",
"dimensionality",
"evidence",
"sentences",
"summary",
"quantization",
"exhaustive",
"scenes"
],
[
"accelerated",
"rapid",
"descent",
"convergence",
"gradient",
"gradients",
"rate",
"normalization",
"converges",
"process"
],
[
"separation",
"strict",
"procedure",
"medical",
"limitations",
"notions",
"imposes",
"practice",
"care",
"adversarial"
],
[
"solver",
"problem",
"solvers",
"solve",
"optimization",
"solved",
"problems",
"solving",
"intractable",
"solves"
],
[
"python",
"derivatives",
"compositional",
"latent",
"leveraged",
"ensembles",
"underlying",
"initialization",
"synthetic",
"frameworks"
],
[
"drastically",
"adjust",
"optimize",
"varying",
"mitigate",
"reduce",
"various",
"degrees",
"ratio",
"analyzed"
],
[
"discrimination",
"biased",
"gender",
"fairness",
"biases",
"bias",
"diversity",
"discriminate",
"ranking",
"perception"
],
[
"forecasts",
"weaker",
"net",
"gains",
"market",
"benchmark",
"forecast",
"credit",
"sentiment",
"share"
],
[
"spectral",
"textual",
"positive",
"discrimination",
"class",
"classification",
"suite",
"contains",
"iii",
"ml"
],
[
"pipeline",
"reconstruction",
"posterior",
"2018",
"feasibility",
"joint",
"resolution",
"inter",
"preliminary",
"logistic"
],
[
"randomly",
"learning",
"dimensionality",
"clusters",
"clustering",
"sizes",
"subspace",
"biases",
"preferences",
"cluster"
],
[
"factorization",
"anomalies",
"resonance",
"spectral",
"entries",
"popularity",
"similarities",
"recent",
"com",
"performances"
],
[
"voice",
"sampled",
"speech",
"audio",
"sound",
"acoustic",
"signals",
"signal",
"message",
"speaker"
],
[
"memory",
"textual",
"attention",
"sequential",
"capacity",
"heavy",
"computation",
"scoring",
"inputs",
"output"
],
[
"region",
"autonomous",
"established",
"remains",
"preserving",
"superiority",
"preserve",
"stability",
"central",
"gap"
],
[
"games",
"game",
"matches",
"players",
"goals",
"scoring",
"super",
"match",
"play",
"series"
],
[
"intermediate",
"testing",
"tests",
"latency",
"protocols",
"tested",
"classes",
"benchmark",
"requirements",
"predictors"
],
[
"particle",
"physics",
"energy",
"detector",
"generate",
"accelerators",
"generator",
"light",
"magnetic",
"chemical"
],
[
"classification",
"modal",
"generic",
"attractive",
"reasonable",
"desirable",
"frequencies",
"variant",
"conditions",
"ability"
],
[
"heuristic",
"markov",
"analytics",
"variational",
"estimator",
"inference",
"statistical",
"bayesian",
"bayes",
"posterior"
],
[
"screening",
"layers",
"safety",
"acoustic",
"privacy",
"aggregating",
"magnetic",
"random",
"detector",
"rigorous"
],
[
"ensemble",
"brings",
"student",
"network",
"collaborative",
"teacher",
"excellent",
"relationship",
"interaction",
"degree"
],
[
"perform",
"task",
"tasks",
"answering",
"robotic",
"performed",
"operations",
"performs",
"favorably",
"robots"
],
[
"unstructured",
"extracted",
"stored",
"data",
"datasets",
"storing",
"compressed",
"encoded",
"databases",
"information"
],
[
"architectural",
"superiority",
"structure",
"depth",
"mining",
"invariance",
"construction",
"complexity",
"coarse",
"algorithmic"
],
[
"resolution",
"proposal",
"proposing",
"proposes",
"novel",
"proposed",
"measure",
"adaptation",
"issue",
"proposals"
],
[
"accuracy",
"correctly",
"prediction",
"calibration",
"device",
"errors",
"module",
"sensor",
"error",
"validation"
],
[
"tensor",
"tensors",
"vectors",
"matrices",
"manifold",
"euclidean",
"vector",
"differentiable",
"orthogonal",
"quadratic"
],
[
"testing",
"enhancing",
"capability",
"capabilities",
"efficacy",
"imaging",
"transmission",
"capacity",
"tests",
"boosting"
],
[
"frequencies",
"spectrum",
"digital",
"imaging",
"representation",
"orthogonal",
"wireless",
"gps",
"coherent",
"resolution"
],
[
"proportional",
"systems",
"consists",
"weighted",
"consisting",
"classification",
"standard",
"based",
"classifier",
"results"
],
[
"ranging",
"distribution",
"margin",
"largest",
"stationary",
"total",
"generative",
"million",
"dual",
"semi"
]
] | 1.467583 | average_word_embeddings_glove.6B.300d | 0.8975 | -0.327229 | 0.204065 | 0.889762 |
42 | ArXiv ML Papers | S³ | 50 | [
[
"distribution",
"margin",
"dual",
"boundary",
"stationary",
"proportional",
"posterior",
"normalizing",
"magnetic",
"soft"
],
[
"downstream",
"identify",
"latent",
"integrate",
"communities",
"developers",
"outperform",
"detect",
"regions",
"countries"
],
[
"learnable",
"topics",
"benchmark",
"ensembles",
"bounds",
"outlier",
"valid",
"intermediate",
"predictors",
"criteria"
],
[
"python",
"leverage",
"achieve",
"attain",
"initialization",
"outperform",
"latent",
"vanilla",
"outperforming",
"exploit"
],
[
"mining",
"architectural",
"coarse",
"numerical",
"superiority",
"considerable",
"activities",
"construction",
"depth",
"huge"
],
[
"framework",
"solution",
"basis",
"samples",
"modalities",
"scenarios",
"layer",
"yield",
"feasibility",
"elements"
],
[
"market",
"gains",
"weaker",
"benchmark",
"price",
"exchange",
"derivatives",
"federated",
"share",
"leveraged"
],
[
"languages",
"linguistic",
"word",
"translation",
"dictionary",
"population",
"english",
"words",
"language",
"meaning"
],
[
"recognition",
"gpus",
"express",
"gps",
"normalized",
"relations",
"expressed",
"respect",
"temporal",
"distance"
],
[
"process",
"processes",
"gaussian",
"inspection",
"stochastic",
"automated",
"monitoring",
"markov",
"automatic",
"continuous"
],
[
"universal",
"architectural",
"criteria",
"color",
"patterns",
"association",
"collect",
"items",
"measurement",
"discover"
],
[
"facial",
"mri",
"discrimination",
"posterior",
"classes",
"transferred",
"class",
"involving",
"preference",
"procedure"
],
[
"attributes",
"techniques",
"strengths",
"characteristic",
"distinguishing",
"characteristics",
"quality",
"products",
"simplicity",
"uses"
],
[
"addressing",
"acoustic",
"speech",
"speaker",
"signals",
"signal",
"voice",
"audio",
"noise",
"noisy"
],
[
"unstructured",
"data",
"stored",
"databases",
"compressed",
"evidence",
"storage",
"database",
"information",
"lightweight"
],
[
"classifier",
"conditional",
"modal",
"accurately",
"similarity",
"reasonable",
"pairwise",
"detect",
"person",
"facial"
],
[
"patch",
"survey",
"vanilla",
"deep",
"logistic",
"regression",
"tree",
"convex",
"recovery",
"root"
],
[
"rules",
"synthetic",
"modifications",
"bounds",
"modified",
"modification",
"generic",
"limits",
"changes",
"category"
],
[
"generating",
"energy",
"pruning",
"power",
"randomly",
"binary",
"consumption",
"generator",
"transformers",
"solely"
],
[
"hardware",
"scalability",
"atari",
"platforms",
"implementations",
"scalable",
"server",
"architectures",
"gpus",
"computing"
],
[
"enhance",
"utilization",
"efficient",
"efficiency",
"rl",
"enhancing",
"rich",
"enhancement",
"attractive",
"augment"
],
[
"shape",
"direction",
"paradigm",
"structure",
"agnostic",
"directions",
"coherent",
"broadly",
"unified",
"hierarchical"
],
[
"code",
"imposes",
"rules",
"strict",
"safety",
"codes",
"absolute",
"privacy",
"labeling",
"rl"
],
[
"gender",
"denoising",
"respect",
"perturbations",
"diversity",
"individual",
"challenges",
"discriminate",
"fairness",
"equal"
],
[
"taxonomy",
"classification",
"protocol",
"changing",
"et",
"defines",
"entropy",
"temperature",
"evolving",
"evolution"
],
[
"overhead",
"classical",
"view",
"music",
"maximization",
"dictionary",
"camera",
"settings",
"art",
"modal"
],
[
"hypothesis",
"perception",
"biases",
"negative",
"bias",
"cognitive",
"poisoning",
"distortion",
"perceptual",
"reasoning"
],
[
"reliable",
"costly",
"convolutional",
"mcmc",
"labels",
"pairwise",
"distances",
"accurate",
"expensive",
"subspaces"
],
[
"verification",
"training",
"evaluation",
"investigation",
"retrieval",
"inspection",
"insight",
"intelligence",
"analysis",
"thorough"
],
[
"online",
"internet",
"mobile",
"google",
"wireless",
"phenomena",
"interactive",
"dirichlet",
"videos",
"streaming"
],
[
"coherent",
"errors",
"magnitude",
"malicious",
"visual",
"results",
"aggregating",
"similarity",
"correction",
"sequences"
],
[
"overview",
"correction",
"autoregressive",
"results",
"consistency",
"sketch",
"approximation",
"epsilon",
"statistical",
"abstraction"
],
[
"applicability",
"navigation",
"distances",
"ml",
"kl",
"logistic",
"distribution",
"resolution",
"sizes",
"available"
],
[
"gaussian",
"beta",
"variance",
"random",
"sampling",
"bernoulli",
"error",
"distributions",
"sample",
"estimator"
],
[
"match",
"game",
"games",
"players",
"play",
"matches",
"scoring",
"goals",
"super",
"score"
],
[
"graph",
"graphs",
"nodes",
"embedding",
"vertex",
"vertices",
"benchmark",
"edges",
"node",
"embeddings"
],
[
"interventions",
"statistically",
"hypothesis",
"strategy",
"outcomes",
"programming",
"successful",
"feasible",
"vector",
"effects"
],
[
"classic",
"convergence",
"logistic",
"superiority",
"regions",
"region",
"origin",
"central",
"captured",
"inequality"
],
[
"favorably",
"superior",
"margin",
"ranking",
"highest",
"weighted",
"biased",
"compared",
"posterior",
"computationally"
],
[
"networks",
"neural",
"programming",
"convolutional",
"channels",
"network",
"channel",
"backpropagation",
"cells",
"cnn"
],
[
"regret",
"issue",
"agent",
"intractable",
"discussed",
"documents",
"revisit",
"details",
"confirm",
"question"
],
[
"detection",
"malicious",
"detect",
"detecting",
"detector",
"attack",
"threat",
"malware",
"targeting",
"target"
],
[
"mild",
"severe",
"weather",
"magnitude",
"grows",
"event",
"category",
"occurs",
"hybrid",
"brings"
],
[
"tensors",
"tensor",
"vector",
"algebra",
"matrices",
"vectors",
"hessian",
"gradient",
"symmetric",
"raw"
],
[
"obstacles",
"solving",
"task",
"tasks",
"goals",
"challenges",
"answering",
"points",
"difficulty",
"capturing"
],
[
"prove",
"generator",
"superiority",
"confidence",
"proof",
"reliable",
"empirically",
"belief",
"proving",
"usefulness"
],
[
"quantization",
"perceptual",
"trees",
"pruning",
"inputs",
"rl",
"selection",
"invariance",
"shed",
"label"
],
[
"achieve",
"frequency",
"rank",
"achieved",
"paper",
"pattern",
"papers",
"uniform",
"degrees",
"rate"
],
[
"score",
"scoring",
"points",
"goals",
"textual",
"scores",
"output",
"outputs",
"generator",
"variable"
],
[
"rapid",
"ascent",
"accelerated",
"descent",
"gradient",
"gradients",
"rates",
"rate",
"convergence",
"stochastic"
]
] | 2.024288 | average_word_embeddings_glove.6B.300d | 0.884 | -0.340219 | 0.194242 | 0.893565 |
42 | ArXiv ML Papers | CombinedTM | 10 | [
[
"adversarial",
"against",
"that",
"model",
"models",
"attack",
"examples",
"training",
"attacks",
"robust"
],
[
"distributions",
"estimator",
"obtain",
"variational",
"posterior",
"gaussian",
"numerical",
"establish",
"observations",
"nearest"
],
[
"research",
"and",
"machine",
"from",
"human",
"to",
"this",
"these",
"in",
"are"
],
[
"previous",
"net",
"cnn",
"hardware",
"segmentation",
"challenge",
"recurrent",
"architecture",
"achieved",
"baseline"
],
[
"policy",
"algorithm",
"optimization",
"agent",
"reward",
"problem",
"rl",
"stochastic",
"gradient",
"reinforcement"
],
[
"and",
"is",
"model",
"the",
"proposed",
"based",
"data",
"on",
"to",
"learning"
],
[
"algorithm",
"matrix",
"for",
"in",
"problem",
"is",
"that",
"the",
"of",
"by"
],
[
"to",
"the",
"network",
"networks",
"neural",
"data",
"of",
"we",
"and",
"as"
],
[
"on",
"embedding",
"graph",
"language",
"learn",
"representations",
"node",
"graphs",
"tasks",
"from"
],
[
"symbolic",
"concept",
"94",
"important",
"future",
"challenges",
"implementation",
"traditional",
"biological",
"intelligence"
]
] | 552.939632 | average_word_embeddings_glove.6B.300d | 0.84 | -0.043813 | 0.139993 | 0.692827 |
42 | ArXiv ML Papers | CombinedTM | 20 | [
[
"word",
"language",
"transformer",
"text",
"languages",
"visual",
"representations",
"pre",
"cross",
"speech"
],
[
"supervised",
"that",
"learning",
"datasets",
"self",
"user",
"on",
"contrastive",
"data",
"unlabeled"
],
[
"actions",
"reinforcement",
"control",
"state",
"rl",
"classical",
"policy",
"action",
"robot",
"imitation"
],
[
"lower",
"bounds",
"estimator",
"delta",
"sample",
"algorithm",
"bound",
"bandits",
"bandit",
"problem"
],
[
"the",
"method",
"is",
"data",
"proposed",
"clustering",
"methods",
"kernel",
"variables",
"classification"
],
[
"as",
"of",
"neural",
"relu",
"parameters",
"with",
"physics",
"network",
"networks",
"we"
],
[
"bayesian",
"likelihood",
"posterior",
"variational",
"monte",
"gaussian",
"chain",
"inference",
"carlo",
"mixture"
],
[
"model",
"architecture",
"training",
"is",
"to",
"on",
"the",
"segmentation",
"teacher",
"and"
],
[
"deep",
"with",
"traffic",
"and",
"based",
"learning",
"performance",
"accuracy",
"classification",
"detection"
],
[
"gradient",
"stochastic",
"descent",
"non",
"convergence",
"convex",
"optimization",
"for",
"problems",
"methods"
],
[
"in",
"is",
"of",
"the",
"that",
"algorithm",
"for",
"matrix",
"we",
"queries"
],
[
"adversarial",
"attacks",
"training",
"against",
"attack",
"model",
"robustness",
"models",
"box",
"defense"
],
[
"been",
"challenges",
"techniques",
"current",
"recommendation",
"comprehensive",
"made",
"clinical",
"years",
"related"
],
[
"memory",
"neural",
"networks",
"precision",
"depth",
"bit",
"scale",
"quantization",
"performance",
"network"
],
[
"to",
"the",
"of",
"in",
"and",
"by",
"day",
"iot",
"health",
"sensor"
],
[
"is",
"to",
"this",
"agent",
"that",
"learning",
"agents",
"in",
"reinforcement",
"environment"
],
[
"and",
"data",
"to",
"be",
"can",
"we",
"in",
"from",
"that",
"models"
],
[
"lstm",
"imbalanced",
"imbalance",
"per",
"overall",
"segmentation",
"among",
"curve",
"showed",
"challenging"
],
[
"movement",
"94",
"sufficiently",
"alternative",
"invariance",
"kinds",
"oracle",
"strict",
"detected",
"rarely"
],
[
"node",
"graph",
"nodes",
"graphs",
"structure",
"feature",
"embedding",
"representation",
"gnns",
"channel"
]
] | 587.474442 | average_word_embeddings_glove.6B.300d | 0.795 | -0.041496 | 0.129791 | 0.724267 |
42 | ArXiv ML Papers | CombinedTM | 30 | [
[
"actions",
"rl",
"action",
"policy",
"control",
"reinforcement",
"agent",
"imitation",
"environment",
"policies"
],
[
"time",
"our",
"series",
"from",
"that",
"we",
"hierarchical",
"on",
"datasets",
"not"
],
[
"dnn",
"connected",
"gender",
"segmentation",
"net",
"map",
"layer",
"operation",
"idea",
"maps"
],
[
"have",
"challenges",
"clinical",
"deep",
"pooling",
"research",
"these",
"has",
"and",
"been"
],
[
"for",
"detection",
"data",
"is",
"traffic",
"system",
"prediction",
"speed",
"based",
"market"
],
[
"bound",
"case",
"estimator",
"log",
"delta",
"unknown",
"sample",
"frac",
"lower",
"bandit"
],
[
"proposed",
"devices",
"transmission",
"server",
"to",
"communication",
"the",
"federated",
"is",
"channel"
],
[
"with",
"architecture",
"neural",
"and",
"on",
"efficient",
"performance",
"nas",
"cnns",
"accuracy"
],
[
"neural",
"that",
"we",
"is",
"for",
"network",
"in",
"of",
"relu",
"matrix"
],
[
"estimators",
"optimization",
"problems",
"nearest",
"gaussian",
"distributed",
"provide",
"minimax",
"tensor",
"neighbor"
],
[
"gan",
"generative",
"neurons",
"networks",
"images",
"to",
"of",
"image",
"the",
"generator"
],
[
"topological",
"autoencoder",
"topic",
"vae",
"3d",
"geometric",
"image",
"autoencoders",
"generating",
"molecular"
],
[
"distillation",
"student",
"the",
"model",
"group",
"teacher",
"on",
"signals",
"to",
"iot"
],
[
"stochastic",
"gradient",
"for",
"convex",
"complexity",
"methods",
"optimization",
"convergence",
"descent",
"epsilon"
],
[
"is",
"that",
"regret",
"sqrt",
"prior",
"the",
"we",
"agent",
"communication",
"problem"
],
[
"neural",
"networks",
"adversarial",
"activation",
"that",
"training",
"perturbations",
"examples",
"robust",
"robustness"
],
[
"model",
"learning",
"we",
"attacks",
"adversarial",
"as",
"attack",
"models",
"privacy",
"to"
],
[
"test",
"bias",
"for",
"and",
"face",
"day",
"human",
"dataset",
"research",
"groups"
],
[
"of",
"patient",
"side",
"patients",
"the",
"effects",
"and",
"an",
"cancer",
"ct"
],
[
"data",
"regression",
"clustering",
"kernel",
"algorithms",
"structured",
"concept",
"stream",
"points",
"algorithm"
],
[
"architectures",
"recurrent",
"memory",
"rnn",
"hardware",
"art",
"state",
"parallel",
"character",
"significant"
],
[
"this",
"in",
"can",
"to",
"that",
"robot",
"from",
"diverse",
"search",
"learning"
],
[
"extensively",
"population",
"rarely",
"ask",
"center",
"tune",
"answer",
"observing",
"kinds",
"description"
],
[
"information",
"language",
"words",
"nlp",
"visual",
"word",
"text",
"sentence",
"context",
"bleu"
],
[
"graph",
"graphs",
"node",
"nodes",
"embedding",
"feature",
"gnns",
"link",
"classification",
"capture"
],
[
"tasks",
"pre",
"training",
"vision",
"transformer",
"trained",
"domain",
"supervised",
"speech",
"on"
],
[
"the",
"sensor",
"in",
"treatment",
"queries",
"of",
"is",
"by",
"that",
"for"
],
[
"data",
"labeled",
"domain",
"learning",
"target",
"feature",
"unlabeled",
"the",
"source",
"and"
],
[
"days",
"technology",
"collected",
"presented",
"project",
"physics",
"already",
"traditional",
"explainable",
"focuses"
],
[
"uncertainty",
"likelihood",
"variational",
"posterior",
"distribution",
"bayesian",
"estimation",
"inference",
"monte",
"approximate"
]
] | 617.524268 | average_word_embeddings_glove.6B.300d | 0.813333 | -0.047543 | 0.125235 | 0.746035 |
42 | ArXiv ML Papers | CombinedTM | 40 | [
[
"service",
"modes",
"each",
"are",
"which",
"one",
"reference",
"from",
"multiple",
"different"
],
[
"rank",
"algorithms",
"data",
"structured",
"metric",
"algorithm",
"matrix",
"stream",
"learning",
"that"
],
[
"to",
"server",
"is",
"computation",
"communication",
"proposed",
"federated",
"devices",
"central",
"optimization"
],
[
"gnns",
"graph",
"structure",
"link",
"attribute",
"nodes",
"graphs",
"node",
"embeddings",
"structural"
],
[
"the",
"to",
"is",
"meta",
"of",
"an",
"this",
"learning",
"in",
"learner"
],
[
"dynamics",
"media",
"paradigm",
"machine",
"fact",
"control",
"quantum",
"constraints",
"be",
"how"
],
[
"research",
"in",
"social",
"design",
"deep",
"and",
"memory",
"health",
"challenges",
"have"
],
[
"images",
"of",
"to",
"segmentation",
"gan",
"our",
"we",
"as",
"image",
"neurons"
],
[
"we",
"and",
"to",
"the",
"is",
"that",
"of",
"quantum",
"19",
"communities"
],
[
"real",
"metric",
"actions",
"robot",
"graph",
"unstructured",
"planning",
"imitation",
"action",
"long"
],
[
"real",
"that",
"systems",
"learning",
"to",
"recommender",
"item",
"robot",
"interactive",
"recommendation"
],
[
"training",
"examples",
"robustness",
"perturbations",
"robust",
"adversarial",
"attacks",
"against",
"box",
"defense"
],
[
"of",
"test",
"the",
"hypothesis",
"distribution",
"divergence",
"testing",
"is",
"calibration",
"channels"
],
[
"can",
"ml",
"the",
"and",
"in",
"of",
"day",
"for",
"94",
"patients"
],
[
"pre",
"language",
"tasks",
"from",
"languages",
"bert",
"downstream",
"task",
"scratch",
"transformer"
],
[
"gaussian",
"monte",
"variational",
"posterior",
"approach",
"approximate",
"carlo",
"control",
"objective",
"mcmc"
],
[
"in",
"of",
"as",
"graph",
"methods",
"laplacian",
"link",
"been",
"dimensional",
"relationships"
],
[
"such",
"or",
"assessment",
"material",
"as",
"tests",
"bias",
"important",
"nlp",
"importance"
],
[
"detected",
"2018",
"would",
"base",
"identified",
"abstract",
"xgboost",
"al",
"contain",
"those"
],
[
"with",
"resolution",
"at",
"performance",
"on",
"cnns",
"bit",
"neural",
"scales",
"channel"
],
[
"introduce",
"vae",
"resolution",
"autoencoder",
"super",
"music",
"gender",
"experience",
"speaker",
"acoustic"
],
[
"probability",
"minimizing",
"obtain",
"pac",
"distributed",
"family",
"means",
"variants",
"estimators",
"differential"
],
[
"rl",
"action",
"state",
"agent",
"agents",
"reinforcement",
"decentralized",
"learning",
"trajectory",
"games"
],
[
"convolutional",
"net",
"network",
"convolution",
"proposes",
"neural",
"hardware",
"segmentation",
"networks",
"convolutions"
],
[
"linear",
"frac",
"big",
"delta",
"log",
"bandit",
"bound",
"arm",
"bandits",
"unknown"
],
[
"matrix",
"relu",
"greedy",
"that",
"random",
"subspaces",
"error",
"for",
"width",
"bounds"
],
[
"neural",
"training",
"network",
"gnns",
"generalization",
"noisy",
"pruning",
"networks",
"loss",
"its"
],
[
"entropy",
"slow",
"guarantees",
"compositional",
"dropout",
"flow",
"uncertainty",
"kernels",
"parameter",
"estimation"
],
[
"to",
"object",
"from",
"and",
"was",
"images",
"using",
"models",
"pose",
"of"
],
[
"labeled",
"learning",
"data",
"domain",
"datasets",
"supervised",
"deep",
"unlabeled",
"self",
"training"
],
[
"attacks",
"model",
"box",
"attack",
"models",
"against",
"privacy",
"adversarial",
"malware",
"black"
],
[
"boosting",
"reports",
"lstm",
"2d",
"patch",
"application",
"score",
"competition",
"created",
"simulation"
],
[
"mri",
"comprehensive",
"imaging",
"clinical",
"computer",
"deep",
"years",
"usage",
"physics",
"current"
],
[
"speech",
"coding",
"attention",
"the",
"feature",
"source",
"model",
"audio",
"information",
"target"
],
[
"convergence",
"gradient",
"stochastic",
"convex",
"methods",
"descent",
"optimization",
"problems",
"point",
"proximal"
],
[
"that",
"our",
"we",
"regret",
"for",
"bounds",
"algorithm",
"bound",
"bandit",
"losses"
],
[
"models",
"series",
"distributions",
"hierarchical",
"causal",
"fidelity",
"model",
"forecasting",
"probabilistic",
"forecast"
],
[
"representations",
"language",
"text",
"words",
"word",
"image",
"semantic",
"translation",
"visual",
"sentence"
],
[
"method",
"time",
"feature",
"clustering",
"series",
"regression",
"data",
"is",
"patient",
"sensor"
],
[
"vehicle",
"traffic",
"based",
"driving",
"anomaly",
"sensor",
"detection",
"and",
"cancer",
"on"
]
] | 603.515563 | average_word_embeddings_glove.6B.300d | 0.7825 | -0.066818 | 0.127927 | 0.766605 |
42 | ArXiv ML Papers | CombinedTM | 50 | [
[
"solving",
"wireless",
"include",
"learnt",
"procedures",
"deal",
"outside",
"what",
"kinds",
"learns"
],
[
"attribute",
"feature",
"structure",
"dynamic",
"by",
"sparse",
"cnn",
"outlier",
"embedding",
"proposal"
],
[
"data",
"time",
"imputation",
"patient",
"missing",
"values",
"method",
"series",
"variables",
"strategy"
],
[
"domain",
"the",
"brain",
"classes",
"target",
"resolution",
"feature",
"reports",
"cell",
"segmentation"
],
[
"supervised",
"training",
"classification",
"cifar",
"unlabeled",
"unsupervised",
"datasets",
"labeled",
"self",
"semi"
],
[
"biases",
"fairness",
"quantum",
"tasks",
"machine",
"systems",
"learning",
"can",
"human",
"robot"
],
[
"embeddings",
"representations",
"language",
"visual",
"word",
"ability",
"reasoning",
"nlp",
"words",
"text"
],
[
"to",
"networks",
"for",
"neurons",
"neural",
"from",
"deep",
"modes",
"as",
"approach"
],
[
"is",
"for",
"data",
"matrix",
"projections",
"structured",
"matrices",
"regression",
"in",
"computing"
],
[
"interaction",
"in",
"humans",
"to",
"of",
"individual",
"behavior",
"dynamics",
"body",
"entities"
],
[
"varepsilon",
"sample",
"sensitivity",
"log",
"estimator",
"shape",
"interval",
"frac",
"component",
"least"
],
[
"variational",
"carlo",
"bayesian",
"variance",
"topological",
"monte",
"mini",
"inference",
"gradient",
"uses"
],
[
"to",
"processes",
"the",
"an",
"by",
"of",
"eye",
"and",
"in",
"regions"
],
[
"the",
"channel",
"gan",
"hypothesis",
"generator",
"coding",
"generative",
"source",
"transmission",
"latent"
],
[
"on",
"covid",
"for",
"performance",
"and",
"with",
"svm",
"based",
"accuracy",
"eeg"
],
[
"extracted",
"lstm",
"detailed",
"filter",
"prediction",
"forecasting",
"svm",
"rnn",
"tensor",
"frequency"
],
[
"decentralized",
"model",
"federated",
"server",
"centralized",
"learning",
"devices",
"framework",
"distributed",
"privacy"
],
[
"dl",
"we",
"to",
"synthetic",
"real",
"data",
"of",
"augmentation",
"images",
"simulated"
],
[
"rl",
"agents",
"agent",
"the",
"reinforcement",
"is",
"communication",
"with",
"an",
"learning"
],
[
"for",
"bound",
"algorithm",
"expert",
"error",
"our",
"interventions",
"correct",
"bounds",
"squared"
],
[
"descent",
"gradient",
"optimization",
"pooling",
"stochastic",
"problems",
"proximal",
"convex",
"minimax",
"ascent"
],
[
"linear",
"algorithm",
"rank",
"noise",
"sequential",
"kernel",
"acquisition",
"efficiently",
"regression",
"constrained"
],
[
"causal",
"minimum",
"label",
"variables",
"free",
"the",
"problem",
"of",
"by",
"recovery"
],
[
"the",
"30",
"94",
"day",
"of",
"calibration",
"test",
"for",
"method",
"is"
],
[
"techniques",
"kernels",
"or",
"are",
"compositional",
"such",
"related",
"large",
"more",
"open"
],
[
"algorithm",
"the",
"method",
"pairwise",
"of",
"proposed",
"data",
"metric",
"treatment",
"to"
],
[
"graphs",
"graph",
"representation",
"capture",
"contrastive",
"representations",
"from",
"existing",
"our",
"conditioning"
],
[
"review",
"explain",
"players",
"users",
"collection",
"sensing",
"modelling",
"largely",
"recognize",
"boosting"
],
[
"shift",
"bias",
"normalization",
"degree",
"gender",
"vae",
"joint",
"landscape",
"pair",
"matching"
],
[
"have",
"been",
"clinical",
"imaging",
"comprehensive",
"image",
"gans",
"deep",
"explanation",
"explanations"
],
[
"filters",
"networks",
"side",
"graph",
"effects",
"gnns",
"network",
"convolutional",
"node",
"cnns"
],
[
"attention",
"memory",
"compared",
"architectures",
"resolution",
"recurrent",
"character",
"decoding",
"competitive",
"driving"
],
[
"comparable",
"patch",
"reveal",
"completely",
"oracle",
"errors",
"just",
"select",
"fashion",
"operators"
],
[
"geometric",
"modified",
"example",
"capacity",
"storage",
"geometry",
"values",
"compress",
"concept",
"3d"
],
[
"neural",
"activation",
"network",
"networks",
"relu",
"physics",
"informed",
"with",
"of",
"our"
],
[
"in",
"influence",
"quantum",
"models",
"that",
"can",
"to",
"for",
"we",
"be"
],
[
"which",
"for",
"algorithm",
"policies",
"state",
"convergence",
"markov",
"policy",
"greedy",
"epsilon"
],
[
"em",
"we",
"that",
"algorithms",
"of",
"queries",
"as",
"topic",
"is",
"in"
],
[
"content",
"fact",
"lead",
"software",
"information",
"document",
"working",
"sentences",
"style",
"textual"
],
[
"bandits",
"combinatorial",
"arm",
"armed",
"exploitation",
"exploration",
"regret",
"prove",
"bandit",
"setting"
],
[
"inference",
"distributions",
"hierarchical",
"series",
"distribution",
"forecast",
"probabilistic",
"relations",
"posterior",
"likelihood"
],
[
"trajectory",
"actions",
"imitation",
"rl",
"policy",
"planning",
"action",
"objects",
"reinforcement",
"control"
],
[
"fidelity",
"but",
"defense",
"training",
"attacks",
"adversarial",
"federated",
"models",
"on",
"trade"
],
[
"box",
"black",
"against",
"adversarial",
"inputs",
"attacks",
"examples",
"unstructured",
"malicious",
"behaviors"
],
[
"detect",
"vehicle",
"location",
"lines",
"sensor",
"and",
"driving",
"behavior",
"ai",
"detection"
],
[
"auxiliary",
"adversarial",
"gradient",
"student",
"group",
"distillation",
"training",
"teacher",
"to",
"targets"
],
[
"and",
"researchers",
"research",
"design",
"scientific",
"challenges",
"machine",
"data",
"recommendation",
"in"
],
[
"transformer",
"model",
"performance",
"downstream",
"speech",
"end",
"scratch",
"tasks",
"bert",
"language"
],
[
"and",
"malware",
"iot",
"model",
"attack",
"to",
"the",
"detection",
"models",
"rate"
],
[
"clustering",
"acoustic",
"separation",
"audio",
"paper",
"speech",
"correlated",
"scientific",
"core",
"heterogeneous"
]
] | 577.830082 | average_word_embeddings_glove.6B.300d | 0.792 | -0.099059 | 0.122829 | 0.781891 |
42 | ArXiv ML Papers | ZeroShotTM | 10 | [
[
"feature",
"network",
"node",
"graph",
"networks",
"graphs",
"nodes",
"embedding",
"gnns",
"learn"
],
[
"for",
"classification",
"segmentation",
"deep",
"data",
"images",
"image",
"convolutional",
"using",
"models"
],
[
"agents",
"reinforcement",
"agent",
"policy",
"we",
"that",
"algorithm",
"rl",
"decision",
"action"
],
[
"of",
"the",
"is",
"data",
"distribution",
"for",
"that",
"noise",
"in",
"matrix"
],
[
"efficiency",
"learning",
"cost",
"performance",
"accuracy",
"models",
"model",
"to",
"memory",
"training"
],
[
"in",
"the",
"detection",
"are",
"and",
"attacks",
"to",
"of",
"model",
"models"
],
[
"tasks",
"task",
"language",
"visual",
"word",
"representations",
"human",
"semantic",
"text",
"context"
],
[
"inference",
"output",
"parameter",
"measurements",
"tensor",
"weather",
"uncertainty",
"parameters",
"dnn",
"component"
],
[
"assessment",
"forecasting",
"years",
"techniques",
"faces",
"detected",
"growing",
"material",
"2015",
"root"
],
[
"optimization",
"gradient",
"linear",
"stochastic",
"convex",
"algorithm",
"order",
"convergence",
"method",
"problems"
]
] | 545.146259 | average_word_embeddings_glove.6B.300d | 0.89 | -0.030824 | 0.157314 | 0.733868 |
42 | ArXiv ML Papers | ZeroShotTM | 20 | [
[
"their",
"can",
"to",
"are",
"models",
"that",
"we",
"our",
"these",
"face"
],
[
"from",
"of",
"based",
"are",
"study",
"and",
"data",
"as",
"in",
"the"
],
[
"learning",
"tasks",
"training",
"trained",
"model",
"task",
"models",
"data",
"performance",
"transfer"
],
[
"control",
"policies",
"reinforcement",
"rl",
"policy",
"exploration",
"agent",
"environment",
"actions",
"action"
],
[
"malicious",
"lstm",
"days",
"pooling",
"usage",
"short",
"movement",
"volume",
"out",
"focuses"
],
[
"attacks",
"model",
"models",
"perturbations",
"robust",
"adversarial",
"box",
"training",
"attack",
"against"
],
[
"in",
"be",
"kernel",
"data",
"of",
"classification",
"methods",
"the",
"algorithms",
"method"
],
[
"neural",
"memory",
"network",
"to",
"energy",
"service",
"deep",
"networks",
"with",
"power"
],
[
"of",
"patients",
"the",
"detection",
"an",
"detect",
"and",
"cancer",
"from",
"based"
],
[
"efficiency",
"image",
"segmentation",
"on",
"convolutional",
"model",
"performance",
"with",
"cnn",
"accuracy"
],
[
"word",
"image",
"speech",
"transformer",
"text",
"audio",
"sequence",
"language",
"recognition",
"visual"
],
[
"in",
"the",
"for",
"of",
"is",
"this",
"distribution",
"test",
"noise",
"to"
],
[
"where",
"agent",
"regret",
"communication",
"bandit",
"online",
"algorithm",
"agents",
"we",
"problem"
],
[
"learning",
"tasks",
"robot",
"visual",
"self",
"goal",
"representations",
"learn",
"objects",
"human"
],
[
"the",
"matrix",
"problem",
"is",
"algorithm",
"rank",
"for",
"of",
"that",
"number"
],
[
"structure",
"features",
"node",
"domain",
"graphs",
"representation",
"graph",
"nodes",
"interactions",
"feature"
],
[
"gradient",
"optimization",
"linear",
"problems",
"stochastic",
"order",
"convex",
"convergence",
"solving",
"methods"
],
[
"social",
"researchers",
"research",
"challenges",
"media",
"information",
"techniques",
"software",
"clinical",
"engineering"
],
[
"distribution",
"shift",
"probability",
"family",
"98",
"approximate",
"factor",
"vertex",
"parameter",
"chain"
],
[
"classical",
"graphs",
"graph",
"dimensional",
"tensor",
"sparse",
"statistical",
"distributions",
"classifiers",
"space"
]
] | 546.179139 | average_word_embeddings_glove.6B.300d | 0.8 | -0.019383 | 0.153126 | 0.732215 |
42 | ArXiv ML Papers | ZeroShotTM | 30 | [
[
"metric",
"data",
"causal",
"of",
"as",
"embedding",
"methods",
"learning",
"on",
"in"
],
[
"our",
"models",
"model",
"detection",
"detect",
"to",
"privacy",
"based",
"attacks",
"face"
],
[
"respectively",
"from",
"detection",
"object",
"and",
"was",
"of",
"images",
"segmentation",
"the"
],
[
"adversarial",
"dnns",
"model",
"accuracy",
"models",
"robust",
"training",
"fidelity",
"scaling",
"overfitting"
],
[
"representations",
"word",
"language",
"visual",
"text",
"semantic",
"style",
"robot",
"words",
"reasoning"
],
[
"gnn",
"we",
"teacher",
"training",
"that",
"gnns",
"networks",
"knowledge",
"student",
"adversarial"
],
[
"the",
"of",
"in",
"and",
"detection",
"cancer",
"to",
"for",
"driving",
"as"
],
[
"methods",
"is",
"from",
"data",
"method",
"the",
"based",
"time",
"for",
"model"
],
[
"supervised",
"speech",
"cnn",
"audio",
"signals",
"quality",
"visual",
"feature",
"signal",
"channel"
],
[
"reinforcement",
"action",
"optimal",
"value",
"policies",
"goal",
"policy",
"trajectory",
"algorithms",
"both"
],
[
"treatment",
"is",
"in",
"distribution",
"of",
"for",
"time",
"day",
"test",
"the"
],
[
"regression",
"kernel",
"distributions",
"gaussian",
"privacy",
"tensor",
"private",
"synthetic",
"empirical",
"approximations"
],
[
"data",
"method",
"proposed",
"methods",
"domain",
"classification",
"feature",
"is",
"learning",
"datasets"
],
[
"and",
"this",
"ml",
"should",
"ai",
"be",
"recommendation",
"these",
"systems",
"community"
],
[
"relu",
"loss",
"pruning",
"network",
"networks",
"neural",
"filters",
"activation",
"architectures",
"functions"
],
[
"to",
"model",
"adversarial",
"models",
"attacks",
"attack",
"black",
"are",
"we",
"box"
],
[
"variational",
"requires",
"markov",
"carlo",
"validation",
"estimates",
"inference",
"free",
"continuous",
"graphical"
],
[
"agent",
"actions",
"reinforcement",
"learning",
"environment",
"robot",
"rl",
"reward",
"exploration",
"tasks"
],
[
"agent",
"sqrt",
"we",
"agents",
"is",
"communication",
"that",
"where",
"regret",
"bound"
],
[
"performance",
"high",
"energy",
"with",
"efficiency",
"devices",
"accuracy",
"time",
"on",
"memory"
],
[
"is",
"of",
"generative",
"that",
"gan",
"to",
"we",
"scene",
"model",
"the"
],
[
"node",
"graph",
"graphs",
"nodes",
"representation",
"embedding",
"contrastive",
"features",
"capture",
"embeddings"
],
[
"clinical",
"machine",
"engineering",
"deep",
"techniques",
"research",
"disease",
"software",
"data",
"computer"
],
[
"matrix",
"that",
"is",
"of",
"number",
"in",
"the",
"we",
"rank",
"input"
],
[
"net",
"device",
"lstm",
"channels",
"numbers",
"filter",
"modified",
"wireless",
"active",
"performed"
],
[
"algorithms",
"proposed",
"the",
"learning",
"is",
"algorithm",
"in",
"problem",
"of",
"private"
],
[
"term",
"concepts",
"objects",
"classifiers",
"mechanism",
"normalization",
"sound",
"event",
"account",
"events"
],
[
"zero",
"translation",
"pre",
"language",
"languages",
"encoder",
"text",
"task",
"decoder",
"transformer"
],
[
"relations",
"server",
"head",
"millions",
"competition",
"super",
"participants",
"frameworks",
"answer",
"major"
],
[
"order",
"epsilon",
"stochastic",
"problems",
"convergence",
"first",
"convex",
"optimization",
"gradient",
"non"
]
] | 545.88194 | average_word_embeddings_glove.6B.300d | 0.743333 | -0.049198 | 0.136654 | 0.740591 |
42 | ArXiv ML Papers | ZeroShotTM | 40 | [
[
"label",
"distance",
"methods",
"metric",
"classification",
"data",
"proposed",
"method",
"feature",
"art"
],
[
"graphs",
"tensor",
"spectrum",
"anomaly",
"decomposition",
"measurement",
"group",
"lstm",
"anomalies",
"full"
],
[
"network",
"nodes",
"node",
"structure",
"graph",
"attribute",
"embedding",
"networks",
"representation",
"interactions"
],
[
"exploration",
"control",
"dynamics",
"policies",
"policy",
"reinforcement",
"rl",
"space",
"learn",
"learning"
],
[
"deep",
"training",
"neural",
"networks",
"model",
"accuracy",
"traffic",
"models",
"network",
"prediction"
],
[
"networks",
"filters",
"convolutional",
"pruning",
"accelerators",
"neural",
"hardware",
"network",
"bit",
"layers"
],
[
"stochastic",
"convex",
"gradient",
"constrained",
"optimization",
"policy",
"convergence",
"algorithms",
"descent",
"problems"
],
[
"based",
"data",
"as",
"and",
"in",
"accuracy",
"model",
"prediction",
"from",
"for"
],
[
"training",
"neural",
"with",
"networks",
"network",
"deep",
"performance",
"architectures",
"accuracy",
"memory"
],
[
"users",
"events",
"text",
"computer",
"sound",
"profiles",
"bias",
"usage",
"social",
"profile"
],
[
"rl",
"online",
"regret",
"agent",
"communication",
"agents",
"is",
"learning",
"an",
"distributed"
],
[
"representation",
"tasks",
"supervised",
"domain",
"contrastive",
"graph",
"learning",
"training",
"framework",
"representations"
],
[
"agent",
"we",
"policy",
"decision",
"games",
"rl",
"action",
"agents",
"fair",
"multi"
],
[
"representation",
"of",
"the",
"image",
"which",
"by",
"to",
"an",
"generative",
"background"
],
[
"box",
"robustness",
"attacks",
"adversarial",
"training",
"examples",
"perturbations",
"fidelity",
"against",
"black"
],
[
"words",
"representations",
"visual",
"semantic",
"languages",
"translation",
"language",
"word",
"attention",
"text"
],
[
"and",
"image",
"classification",
"by",
"images",
"segmentation",
"object",
"for",
"detection",
"of"
],
[
"private",
"euclidean",
"distributed",
"majority",
"minimize",
"varepsilon",
"joint",
"privacy",
"locally",
"neighbor"
],
[
"attacks",
"model",
"attack",
"we",
"are",
"have",
"iot",
"models",
"based",
"in"
],
[
"of",
"high",
"classification",
"methods",
"in",
"stream",
"data",
"as",
"be",
"are"
],
[
"and",
"the",
"of",
"for",
"is",
"training",
"method",
"on",
"in",
"an"
],
[
"textit",
"score",
"negative",
"boosting",
"urban",
"demand",
"temporal",
"weather",
"depth",
"predictors"
],
[
"software",
"clinical",
"techniques",
"challenges",
"ml",
"industry",
"research",
"engineering",
"years",
"community"
],
[
"probabilistic",
"models",
"hierarchical",
"model",
"uncertainty",
"inference",
"bayesian",
"forecasts",
"forecasting",
"estimation"
],
[
"concepts",
"scientific",
"useful",
"approximately",
"correct",
"represent",
"part",
"called",
"pooling",
"very"
],
[
"estimator",
"probability",
"regression",
"linear",
"error",
"frac",
"bias",
"sensitivity",
"bounds",
"interval"
],
[
"prediction",
"classification",
"selection",
"proposed",
"method",
"methods",
"from",
"time",
"feature",
"new"
],
[
"clustering",
"graph",
"matrix",
"problem",
"algorithms",
"graphs",
"nodes",
"cluster",
"clusters",
"theoretical"
],
[
"human",
"are",
"robot",
"to",
"tasks",
"that",
"diverse",
"objects",
"world",
"or"
],
[
"algorithm",
"bandit",
"we",
"regret",
"an",
"that",
"our",
"is",
"class",
"prior"
],
[
"super",
"resolution",
"sequence",
"sentences",
"net",
"video",
"relations",
"explanation",
"vanilla",
"baseline"
],
[
"end",
"speech",
"recognition",
"performance",
"on",
"downstream",
"image",
"loss",
"transformer",
"resolution"
],
[
"stochastic",
"optimization",
"problems",
"solving",
"variational",
"gradient",
"posterior",
"proximal",
"descent",
"function"
],
[
"to",
"research",
"human",
"on",
"questions",
"influence",
"are",
"these",
"and",
"fairness"
],
[
"algorithm",
"algorithms",
"of",
"sampling",
"private",
"the",
"in",
"that",
"label",
"is"
],
[
"and",
"detection",
"vehicle",
"to",
"health",
"detect",
"driving",
"in",
"of",
"model"
],
[
"knowledge",
"user",
"items",
"domain",
"adaptation",
"domains",
"query",
"target",
"search",
"objects"
],
[
"can",
"data",
"to",
"of",
"from",
"are",
"images",
"and",
"deep",
"be"
],
[
"embedding",
"deep",
"to",
"metric",
"methods",
"learning",
"framework",
"source",
"approach",
"knowledge"
],
[
"the",
"of",
"that",
"sensor",
"in",
"matrix",
"is",
"noise",
"this",
"low"
]
] | 548.787165 | average_word_embeddings_glove.6B.300d | 0.6575 | -0.040872 | 0.141755 | 0.74298 |
42 | ArXiv ML Papers | ZeroShotTM | 50 | [
[
"attention",
"visual",
"distortion",
"proposed",
"segmentation",
"separation",
"speech",
"signals",
"feature",
"audio"
],
[
"networks",
"architectures",
"pruning",
"loss",
"neural",
"efficient",
"architecture",
"art",
"search",
"parameters"
],
[
"clusters",
"capacity",
"net",
"vector",
"tensor",
"simple",
"anomalies",
"operation",
"support",
"lasso"
],
[
"convolutional",
"bit",
"networks",
"neural",
"memory",
"performance",
"hardware",
"with",
"deep",
"art"
],
[
"the",
"data",
"method",
"and",
"sensor",
"using",
"for",
"time",
"of",
"anomaly"
],
[
"networks",
"network",
"neural",
"convolutional",
"temperature",
"hardware",
"fitting",
"convolution",
"frequency",
"dnn"
],
[
"model",
"dnn",
"samples",
"training",
"models",
"adversarial",
"dnns",
"attack",
"that",
"parameters"
],
[
"is",
"the",
"proposed",
"data",
"for",
"of",
"method",
"matrix",
"methods",
"metric"
],
[
"domain",
"knowledge",
"user",
"from",
"recommendation",
"in",
"users",
"adaptation",
"task",
"representations"
],
[
"algorithms",
"problem",
"of",
"query",
"queries",
"data",
"problems",
"inference",
"complexity",
"from"
],
[
"we",
"online",
"agents",
"an",
"agent",
"communication",
"regret",
"neighbors",
"strategy",
"sqrt"
],
[
"stochastic",
"convex",
"linear",
"gradient",
"optimization",
"convergence",
"dimension",
"order",
"descent",
"strongly"
],
[
"theoretical",
"matrix",
"of",
"in",
"the",
"problem",
"is",
"objects",
"clustering",
"rank"
],
[
"data",
"with",
"training",
"networks",
"compact",
"model",
"parameters",
"ensemble",
"neural",
"accuracy"
],
[
"queries",
"is",
"are",
"the",
"distribution",
"causal",
"to",
"we",
"in",
"of"
],
[
"patients",
"from",
"sensor",
"with",
"and",
"vehicle",
"cancer",
"ct",
"the",
"of"
],
[
"traffic",
"systems",
"to",
"data",
"user",
"in",
"based",
"prediction",
"of",
"and"
],
[
"few",
"consistency",
"video",
"image",
"gans",
"label",
"frames",
"frame",
"previous",
"generate"
],
[
"visual",
"environments",
"learning",
"tasks",
"current",
"deep",
"systems",
"perception",
"robot",
"able"
],
[
"variational",
"generative",
"bayesian",
"optimization",
"inference",
"posterior",
"computational",
"methods",
"estimation",
"flows"
],
[
"conditions",
"covid",
"19",
"the",
"testing",
"of",
"for",
"day",
"treatment",
"missing"
],
[
"learning",
"graph",
"embedding",
"contrastive",
"methods",
"graphs",
"art",
"manifold",
"node",
"framework"
],
[
"to",
"we",
"models",
"attack",
"attacks",
"model",
"samples",
"dataset",
"malware",
"detection"
],
[
"network",
"performance",
"cnn",
"with",
"training",
"speech",
"model",
"it",
"networks",
"parameters"
],
[
"euclidean",
"market",
"diffusion",
"spectrum",
"assumptions",
"statistical",
"private",
"financial",
"extent",
"mcmc"
],
[
"policy",
"optimal",
"learning",
"action",
"policies",
"reinforcement",
"value",
"algorithm",
"convergence",
"problem"
],
[
"methods",
"gradient",
"dynamics",
"constraints",
"constrained",
"method",
"observations",
"stochastic",
"optimization",
"order"
],
[
"language",
"bleu",
"text",
"style",
"speech",
"languages",
"words",
"translation",
"word",
"visual"
],
[
"robot",
"human",
"environment",
"to",
"environments",
"systems",
"how",
"learning",
"reinforcement",
"or"
],
[
"estimating",
"right",
"frac",
"left",
"varepsilon",
"interval",
"locally",
"estimates",
"requires",
"graphical"
],
[
"learning",
"symbolic",
"space",
"robot",
"reward",
"reinforcement",
"policy",
"objects",
"tasks",
"behaviors"
],
[
"ml",
"machine",
"research",
"researchers",
"human",
"and",
"questions",
"challenges",
"of",
"these"
],
[
"task",
"tasks",
"performance",
"supervised",
"training",
"programs",
"trained",
"words",
"end",
"on"
],
[
"source",
"the",
"which",
"interaction",
"feature",
"features",
"representation",
"attention",
"interactions",
"target"
],
[
"cost",
"engineering",
"imaging",
"power",
"applications",
"machine",
"monitoring",
"detection",
"and",
"accuracy"
],
[
"models",
"forecasting",
"forecasts",
"approach",
"data",
"datasets",
"such",
"as",
"design",
"forecast"
],
[
"information",
"media",
"social",
"software",
"industry",
"spread",
"platform",
"countries",
"patterns",
"users"
],
[
"demands",
"major",
"assessment",
"represent",
"good",
"created",
"eight",
"promising",
"massive",
"idea"
],
[
"attacks",
"world",
"user",
"models",
"model",
"learning",
"are",
"active",
"to",
"based"
],
[
"exploration",
"rl",
"agent",
"agents",
"reward",
"multi",
"communication",
"environment",
"safety",
"reinforcement"
],
[
"which",
"prove",
"linear",
"theoretical",
"bounds",
"optimal",
"we",
"policy",
"setting",
"further"
],
[
"is",
"pooling",
"network",
"the",
"in",
"matrix",
"of",
"dimensional",
"space",
"problem"
],
[
"model",
"based",
"prediction",
"gan",
"and",
"performance",
"to",
"dataset",
"images",
"proposed"
],
[
"target",
"representations",
"images",
"pre",
"supervised",
"task",
"domain",
"tasks",
"image",
"language"
],
[
"accuracy",
"eeg",
"based",
"the",
"from",
"proposed",
"an",
"traffic",
"and",
"detection"
],
[
"we",
"that",
"model",
"forecasting",
"models",
"forecasts",
"data",
"fidelity",
"forecast",
"as"
],
[
"gnns",
"node",
"nodes",
"graph",
"spectral",
"filters",
"structure",
"network",
"local",
"position"
],
[
"where",
"for",
"that",
"algorithm",
"is",
"distribution",
"bound",
"sample",
"class",
"prior"
],
[
"classification",
"between",
"of",
"classifier",
"representation",
"structure",
"feature",
"in",
"novel",
"detection"
],
[
"attacks",
"perturbations",
"defense",
"robustness",
"against",
"adversarial",
"box",
"examples",
"robust",
"attack"
]
] | 625.48978 | average_word_embeddings_glove.6B.300d | 0.608 | -0.037376 | 0.13713 | 0.739937 |
42 | BBC News | BERTopic | 10 | [
[
"and",
"he",
"to",
"for",
"in",
"of",
"the",
"but",
"we",
"his"
],
[
"and",
"of",
"her",
"in",
"to",
"the",
"she",
"world",
"was",
"indoor"
],
[
"and",
"open",
"the",
"to",
"his",
"was",
"in",
"of",
"but",
"for"
],
[
"for",
"that",
"of",
"in",
"and",
"to",
"said",
"on",
"the",
"is"
],
[
"to",
"of",
"for",
"in",
"the",
"and",
"film",
"was",
"it",
"best"
],
[
"it",
"for",
"are",
"to",
"in",
"and",
"is",
"of",
"the",
"that"
]
] | 5.115115 | average_word_embeddings_glove.6B.300d | 0.4 | -0.023793 | 0.287191 | 0.564488 |
42 | BBC News | BERTopic | 20 | [
[
"and",
"he",
"to",
"for",
"in",
"of",
"the",
"but",
"we",
"his"
],
[
"and",
"of",
"her",
"in",
"to",
"the",
"she",
"world",
"was",
"indoor"
],
[
"and",
"open",
"the",
"to",
"his",
"was",
"in",
"of",
"but",
"for"
],
[
"for",
"that",
"of",
"in",
"and",
"to",
"said",
"on",
"the",
"is"
],
[
"to",
"of",
"for",
"in",
"the",
"and",
"film",
"was",
"it",
"best"
],
[
"it",
"for",
"are",
"to",
"in",
"and",
"is",
"of",
"the",
"that"
]
] | 4.639692 | average_word_embeddings_glove.6B.300d | 0.4 | -0.023793 | 0.287191 | 0.571718 |
42 | BBC News | BERTopic | 30 | [
[
"and",
"he",
"to",
"for",
"in",
"of",
"the",
"but",
"we",
"his"
],
[
"and",
"of",
"her",
"in",
"to",
"the",
"she",
"world",
"was",
"indoor"
],
[
"and",
"open",
"the",
"to",
"his",
"was",
"in",
"of",
"but",
"for"
],
[
"for",
"that",
"of",
"in",
"and",
"to",
"said",
"on",
"the",
"is"
],
[
"to",
"of",
"for",
"in",
"the",
"and",
"film",
"was",
"it",
"best"
],
[
"it",
"for",
"are",
"to",
"in",
"and",
"is",
"of",
"the",
"that"
]
] | 4.414884 | average_word_embeddings_glove.6B.300d | 0.4 | -0.023793 | 0.287191 | 0.5738 |
42 | BBC News | BERTopic | 40 | [
[
"and",
"he",
"to",
"for",
"in",
"of",
"the",
"but",
"we",
"his"
],
[
"and",
"of",
"her",
"in",
"to",
"the",
"she",
"world",
"was",
"indoor"
],
[
"and",
"open",
"the",
"to",
"his",
"was",
"in",
"of",
"but",
"for"
],
[
"for",
"that",
"of",
"in",
"and",
"to",
"said",
"on",
"the",
"is"
],
[
"to",
"of",
"for",
"in",
"the",
"and",
"film",
"was",
"it",
"best"
],
[
"it",
"for",
"are",
"to",
"in",
"and",
"is",
"of",
"the",
"that"
]
] | 4.642083 | average_word_embeddings_glove.6B.300d | 0.4 | -0.023793 | 0.287191 | 0.5685 |
42 | BBC News | BERTopic | 50 | [
[
"and",
"he",
"to",
"for",
"in",
"of",
"the",
"but",
"we",
"his"
],
[
"and",
"of",
"her",
"in",
"to",
"the",
"she",
"world",
"was",
"indoor"
],
[
"and",
"open",
"the",
"to",
"his",
"was",
"in",
"of",
"but",
"for"
],
[
"for",
"that",
"of",
"in",
"and",
"to",
"said",
"on",
"the",
"is"
],
[
"to",
"of",
"for",
"in",
"the",
"and",
"film",
"was",
"it",
"best"
],
[
"it",
"for",
"are",
"to",
"in",
"and",
"is",
"of",
"the",
"that"
]
] | 4.597936 | average_word_embeddings_glove.6B.300d | 0.4 | -0.023793 | 0.287191 | 0.569225 |
42 | BBC News | NMF | 10 | [
[
"at",
"by",
"the",
"first",
"which",
"it",
"on",
"to",
"time",
"from"
],
[
"not",
"we",
"to",
"that",
"and",
"is",
"with",
"have",
"of",
"it"
],
[
"be",
"are",
"they",
"by",
"the",
"that",
"to",
"said",
"on",
"have"
],
[
"and",
"his",
"to",
"mr",
"was",
"he",
"said",
"for",
"would",
"party"
],
[
"in",
"was",
"to",
"and",
"with",
"at",
"his",
"but",
"for",
"the"
],
[
"its",
"that",
"was",
"said",
"were",
"of",
"the",
"by",
"in",
"has"
],
[
"with",
"and",
"for",
"of",
"more",
"are",
"is",
"as",
"they",
"on"
],
[
"that",
"is",
"it",
"to",
"music",
"what",
"in",
"you",
"not",
"if"
],
[
"minimum",
"will",
"for",
"at",
"in",
"be",
"is",
"increase",
"and",
"that"
],
[
"for",
"25",
"song",
"robbie",
"of",
"last",
"years",
"best",
"it",
"and"
]
] | 1.756336 | average_word_embeddings_glove.6B.300d | 0.49 | -0.001058 | 0.269654 | 0.572547 |
42 | BBC News | NMF | 20 | [
[
"which",
"the",
"has",
"as",
"to",
"of",
"by",
"at",
"this",
"first"
],
[
"and",
"that",
"we",
"to",
"of",
"is",
"not",
"with",
"have",
"in"
],
[
"the",
"of",
"to",
"by",
"up",
"out",
"and",
"with",
"take",
"over"
],
[
"him",
"silk",
"his",
"he",
"and",
"the",
"has",
"is",
"was",
"party"
],
[
"her",
"was",
"at",
"uk",
"last",
"the",
"year",
"in",
"world",
"their"
],
[
"than",
"was",
"by",
"were",
"an",
"of",
"one",
"new",
"or",
"as"
],
[
"in",
"into",
"and",
"world",
"as",
"work",
"with",
"also",
"other",
"both"
],
[
"is",
"in",
"to",
"it",
"you",
"that",
"but",
"as",
"has",
"not"
],
[
"film",
"year",
"also",
"for",
"at",
"who",
"is",
"best",
"up",
"won"
],
[
"robbie",
"and",
"years",
"for",
"best",
"song",
"25",
"in",
"last",
"music"
],
[
"are",
"they",
"have",
"their",
"on",
"people",
"you",
"is",
"the",
"can"
],
[
"was",
"his",
"but",
"on",
"the",
"with",
"after",
"had",
"it",
"from"
],
[
"be",
"will",
"on",
"would",
"new",
"have",
"in",
"is",
"could",
"at"
],
[
"and",
"she",
"black",
"what",
"to",
"music",
"that",
"her",
"the",
"in"
],
[
"would",
"it",
"said",
"be",
"was",
"and",
"he",
"for",
"had",
"were"
],
[
"minimum",
"for",
"and",
"of",
"to",
"in",
"will",
"that",
"increase",
"are"
],
[
"to",
"for",
"have",
"that",
"in",
"file",
"many",
"firms",
"this",
"software"
],
[
"he",
"blair",
"on",
"mr",
"labour",
"election",
"that",
"party",
"brown",
"was"
],
[
"are",
"mobile",
"more",
"with",
"on",
"of",
"as",
"people",
"tv",
"and"
],
[
"is",
"us",
"in",
"on",
"has",
"its",
"and",
"said",
"by",
"with"
]
] | 2.883286 | average_word_embeddings_glove.6B.300d | 0.45 | 0.001947 | 0.236665 | 0.584661 |
42 | BBC News | NMF | 30 | [
[
"the",
"of",
"to",
"time",
"first",
"only",
"most",
"by",
"made",
"into"
],
[
"that",
"and",
"with",
"in",
"to",
"is",
"of",
"we",
"not",
"be"
],
[
"to",
"work",
"this",
"up",
"make",
"out",
"take",
"of",
"also",
"000"
],
[
"the",
"he",
"his",
"has",
"at",
"him",
"with",
"who",
"was",
"been"
],
[
"world",
"the",
"in",
"year",
"their",
"european",
"first",
"last",
"her",
"uk"
],
[
"new",
"all",
"most",
"part",
"from",
"some",
"of",
"than",
"one",
"an"
],
[
"and",
"but",
"it",
"in",
"you",
"to",
"if",
"the",
"like",
"can"
],
[
"and",
"best",
"it",
"song",
"that",
"in",
"25",
"years",
"but",
"of"
],
[
"has",
"are",
"is",
"be",
"there",
"not",
"but",
"this",
"in",
"you"
],
[
"best",
"the",
"film",
"at",
"her",
"who",
"she",
"won",
"awards",
"award"
],
[
"are",
"their",
"they",
"or",
"were",
"people",
"have",
"them",
"by",
"with"
],
[
"in",
"but",
"that",
"not",
"have",
"we",
"liverpool",
"would",
"are",
"it"
],
[
"be",
"could",
"and",
"but",
"to",
"not",
"should",
"would",
"new",
"law"
],
[
"and",
"music",
"to",
"what",
"she",
"that",
"black",
"as",
"people",
"it"
],
[
"work",
"the",
"also",
"and",
"other",
"world",
"into",
"two",
"both",
"between"
],
[
"to",
"and",
"in",
"for",
"that",
"is",
"minimum",
"of",
"will",
"be"
],
[
"for",
"file",
"technology",
"in",
"that",
"not",
"software",
"and",
"to",
"this"
],
[
"election",
"mr",
"labour",
"blair",
"the",
"howard",
"brown",
"he",
"minister",
"party"
],
[
"silk",
"this",
"for",
"and",
"in",
"to",
"party",
"he",
"has",
"have"
],
[
"in",
"us",
"is",
"company",
"has",
"it",
"by",
"its",
"with",
"yukos"
],
[
"added",
"from",
"up",
"high",
"an",
"000",
"the",
"also",
"for",
"while"
],
[
"after",
"was",
"who",
"the",
"when",
"by",
"but",
"had",
"she",
"were"
],
[
"his",
"with",
"after",
"the",
"from",
"england",
"but",
"at",
"game",
"to"
],
[
"more",
"technology",
"tv",
"to",
"mobile",
"are",
"people",
"with",
"services",
"digital"
],
[
"the",
"which",
"gadget",
"it",
"apple",
"has",
"be",
"gadgets",
"mobile",
"at"
],
[
"will",
"new",
"is",
"have",
"at",
"be",
"by",
"which",
"this",
"next"
],
[
"show",
"on",
"in",
"day",
"some",
"the",
"friday",
"also",
"from",
"which"
],
[
"year",
"from",
"growth",
"at",
"economy",
"economic",
"in",
"us",
"by",
"sales"
],
[
"it",
"mr",
"were",
"said",
"he",
"by",
"the",
"had",
"uk",
"added"
],
[
"such",
"well",
"as",
"to",
"an",
"with",
"more",
"new",
"one",
"however"
]
] | 4.106758 | average_word_embeddings_glove.6B.300d | 0.446667 | 0.012502 | 0.23947 | 0.626326 |
42 | BBC News | NMF | 40 | [
[
"by",
"time",
"made",
"of",
"only",
"most",
"first",
"to",
"the",
"into"
],
[
"are",
"not",
"be",
"is",
"this",
"in",
"to",
"but",
"of",
"there"
],
[
"also",
"up",
"to",
"take",
"over",
"this",
"out",
"make",
"any",
"000"
],
[
"were",
"when",
"in",
"was",
"had",
"the",
"been",
"by",
"after",
"which"
],
[
"the",
"their",
"in",
"world",
"uk",
"european",
"new",
"first",
"final",
"years"
],
[
"new",
"of",
"one",
"all",
"than",
"part",
"and",
"some",
"most",
"many"
],
[
"world",
"into",
"the",
"work",
"and",
"years",
"from",
"also",
"other",
"both"
],
[
"and",
"to",
"best",
"that",
"song",
"of",
"in",
"years",
"it",
"25"
],
[
"new",
"by",
"this",
"year",
"which",
"will",
"be",
"service",
"is",
"have"
],
[
"award",
"won",
"best",
"who",
"director",
"the",
"film",
"awards",
"with",
"actress"
],
[
"of",
"will",
"and",
"in",
"increase",
"for",
"to",
"be",
"that",
"minimum"
],
[
"have",
"we",
"not",
"but",
"liverpool",
"would",
"if",
"had",
"think",
"that"
],
[
"such",
"time",
"to",
"as",
"an",
"new",
"their",
"one",
"well",
"however"
],
[
"music",
"in",
"black",
"the",
"and",
"what",
"that",
"stone",
"she",
"people"
],
[
"the",
"on",
"and",
"from",
"also",
"an",
"day",
"after",
"some",
"friday"
],
[
"government",
"to",
"tax",
"local",
"public",
"in",
"budget",
"would",
"council",
"000"
],
[
"and",
"that",
"to",
"this",
"about",
"file",
"make",
"not",
"on",
"for"
],
[
"mr",
"he",
"told",
"minister",
"prime",
"blair",
"brown",
"the",
"had",
"by"
],
[
"to",
"silk",
"party",
"in",
"he",
"this",
"and",
"for",
"of",
"mr"
],
[
"but",
"if",
"at",
"the",
"like",
"very",
"all",
"it",
"they",
"and"
],
[
"000",
"up",
"added",
"for",
"also",
"an",
"new",
"year",
"news",
"while"
],
[
"firm",
"with",
"its",
"in",
"by",
"company",
"which",
"yukos",
"it",
"oil"
],
[
"we",
"and",
"to",
"not",
"have",
"with",
"of",
"that",
"be",
"government"
],
[
"people",
"the",
"are",
"by",
"and",
"or",
"were",
"their",
"have",
"they"
],
[
"her",
"olympic",
"year",
"with",
"but",
"first",
"she",
"who",
"after",
"had"
],
[
"mini",
"it",
"software",
"apple",
"to",
"pc",
"with",
"mac",
"computer",
"and"
],
[
"his",
"who",
"the",
"him",
"he",
"with",
"but",
"not",
"told",
"would"
],
[
"in",
"year",
"growth",
"by",
"economy",
"from",
"sales",
"prices",
"market",
"than"
],
[
"he",
"the",
"said",
"it",
"were",
"being",
"had",
"uk",
"added",
"rights"
],
[
"would",
"could",
"be",
"not",
"to",
"should",
"have",
"if",
"but",
"new"
],
[
"are",
"technology",
"is",
"of",
"dvd",
"high",
"be",
"on",
"games",
"in"
],
[
"are",
"to",
"mobile",
"technology",
"data",
"3g",
"but",
"networks",
"on",
"of"
],
[
"have",
"number",
"show",
"the",
"at",
"best",
"and",
"his",
"end",
"theatre"
],
[
"howard",
"the",
"labour",
"party",
"election",
"tory",
"would",
"tories",
"tax",
"in"
],
[
"are",
"your",
"with",
"there",
"can",
"and",
"that",
"game",
"if",
"you"
],
[
"been",
"has",
"have",
"this",
"is",
"last",
"said",
"since",
"the",
"now"
],
[
"which",
"mobile",
"of",
"it",
"and",
"the",
"gadget",
"be",
"gadgets",
"in"
],
[
"china",
"with",
"in",
"dollar",
"budget",
"trade",
"us",
"bush",
"deficit",
"about"
],
[
"but",
"with",
"his",
"england",
"after",
"from",
"game",
"half",
"ireland",
"the"
],
[
"people",
"tv",
"with",
"broadband",
"more",
"digital",
"net",
"services",
"service",
"video"
]
] | 5.877973 | average_word_embeddings_glove.6B.300d | 0.4575 | 0.020847 | 0.227713 | 0.65189 |
42 | BBC News | NMF | 50 | [
[
"of",
"and",
"the",
"most",
"time",
"first",
"in",
"only",
"also",
"into"
],
[
"this",
"to",
"is",
"in",
"and",
"but",
"be",
"of",
"not",
"there"
],
[
"mr",
"added",
"of",
"said",
"were",
"and",
"had",
"uk",
"to",
"is"
],
[
"he",
"told",
"him",
"would",
"an",
"but",
"not",
"very",
"added",
"to"
],
[
"first",
"final",
"last",
"years",
"world",
"in",
"uk",
"european",
"new",
"set"
],
[
"and",
"than",
"one",
"of",
"most",
"the",
"to",
"their",
"all",
"also"
],
[
"into",
"two",
"other",
"also",
"and",
"work",
"between",
"both",
"years",
"over"
],
[
"it",
"song",
"of",
"and",
"best",
"25",
"to",
"in",
"years",
"that"
],
[
"make",
"digital",
"to",
"which",
"new",
"be",
"next",
"will",
"can",
"service"
],
[
"the",
"best",
"film",
"awards",
"who",
"award",
"won",
"director",
"actress",
"actor"
],
[
"of",
"to",
"for",
"will",
"minimum",
"increase",
"in",
"and",
"that",
"be"
],
[
"we",
"but",
"liverpool",
"think",
"would",
"had",
"club",
"steven",
"not",
"if"
],
[
"but",
"it",
"the",
"to",
"very",
"at",
"news",
"out",
"because",
"over"
],
[
"and",
"what",
"music",
"to",
"she",
"stone",
"black",
"from",
"or",
"the"
],
[
"that",
"out",
"this",
"not",
"about",
"the",
"to",
"there",
"by",
"file"
],
[
"economic",
"growth",
"year",
"economy",
"from",
"bank",
"rates",
"prices",
"sales",
"market"
],
[
"also",
"from",
"an",
"on",
"the",
"day",
"friday",
"some",
"wednesday",
"media"
],
[
"gadget",
"the",
"be",
"it",
"mobile",
"gadgets",
"which",
"first",
"list",
"or"
],
[
"to",
"party",
"silk",
"and",
"he",
"in",
"this",
"for",
"mr",
"on"
],
[
"their",
"were",
"people",
"they",
"and",
"who",
"them",
"when",
"on",
"those"
],
[
"year",
"while",
"week",
"up",
"new",
"also",
"for",
"an",
"000",
"added"
],
[
"shares",
"which",
"its",
"firm",
"also",
"would",
"london",
"company",
"market",
"deutsche"
],
[
"to",
"of",
"we",
"and",
"that",
"with",
"not",
"be",
"have",
"in"
],
[
"last",
"have",
"been",
"from",
"very",
"may",
"the",
"not",
"this",
"four"
],
[
"had",
"blair",
"mr",
"prime",
"brown",
"his",
"the",
"minister",
"told",
"chancellor"
],
[
"net",
"security",
"to",
"users",
"software",
"or",
"people",
"microsoft",
"attacks",
"site"
],
[
"was",
"had",
"were",
"after",
"been",
"when",
"in",
"but",
"which",
"one"
],
[
"to",
"out",
"this",
"take",
"up",
"make",
"any",
"set",
"which",
"however"
],
[
"the",
"from",
"were",
"by",
"which",
"been",
"000",
"an",
"up",
"number"
],
[
"well",
"as",
"such",
"an",
"time",
"part",
"many",
"however",
"one",
"industry"
],
[
"phones",
"people",
"music",
"more",
"mobile",
"with",
"to",
"phone",
"mobiles",
"the"
],
[
"she",
"who",
"her",
"year",
"the",
"first",
"but",
"after",
"olympic",
"world"
],
[
"number",
"show",
"at",
"the",
"theatre",
"royal",
"top",
"end",
"best",
"expected"
],
[
"party",
"the",
"labour",
"election",
"tory",
"howard",
"would",
"tax",
"tories",
"in"
],
[
"the",
"your",
"if",
"can",
"there",
"what",
"games",
"and",
"game",
"you"
],
[
"now",
"has",
"with",
"been",
"since",
"the",
"said",
"week",
"against",
"internet"
],
[
"it",
"and",
"in",
"of",
"be",
"with",
"have",
"on",
"but",
"children"
],
[
"to",
"the",
"and",
"mobile",
"of",
"3g",
"technology",
"networks",
"data",
"for"
],
[
"england",
"half",
"six",
"after",
"but",
"ireland",
"the",
"game",
"wales",
"their"
],
[
"broadband",
"video",
"people",
"uk",
"content",
"net",
"digital",
"tv",
"services",
"over"
],
[
"of",
"oil",
"in",
"russian",
"yukos",
"its",
"the",
"court",
"gazprom",
"that"
],
[
"be",
"is",
"games",
"the",
"dvd",
"high",
"and",
"technology",
"ray",
"in"
],
[
"mac",
"mini",
"and",
"the",
"it",
"apple",
"pc",
"for",
"computer",
"is"
],
[
"who",
"his",
"him",
"the",
"my",
"after",
"year",
"me",
"old",
"club"
],
[
"can",
"phone",
"companies",
"these",
"be",
"calls",
"people",
"about",
"not",
"is"
],
[
"not",
"law",
"should",
"government",
"be",
"would",
"could",
"lord",
"but",
"if"
],
[
"dollar",
"us",
"china",
"bush",
"trade",
"budget",
"eu",
"deficit",
"about",
"firms"
],
[
"an",
"out",
"the",
"from",
"with",
"over",
"one",
"hit",
"both",
"through"
],
[
"radio",
"in",
"is",
"hop",
"spanish",
"world",
"hip",
"says",
"but",
"rap"
],
[
"where",
"there",
"are",
"be",
"not",
"their",
"which",
"do",
"more",
"they"
]
] | 6.986562 | average_word_embeddings_glove.6B.300d | 0.482 | 0.026203 | 0.216141 | 0.682455 |
42 | BBC News | LDA | 10 | [
[
"and",
"his",
"to",
"the",
"was",
"mr",
"he",
"in",
"of",
"had"
],
[
"ms",
"her",
"the",
"she",
"to",
"in",
"was",
"said",
"air",
"airlines"
],
[
"for",
"to",
"in",
"of",
"that",
"the",
"said",
"and",
"is",
"it"
],
[
"was",
"to",
"the",
"on",
"of",
"in",
"for",
"and",
"is",
"it"
],
[
"to",
"liverpool",
"united",
"the",
"club",
"chelsea",
"smith",
"arsenal",
"he",
"manchester"
],
[
"the",
"in",
"of",
"to",
"said",
"on",
"that",
"be",
"and",
"was"
],
[
"to",
"the",
"and",
"of",
"that",
"it",
"are",
"is",
"in",
"be"
],
[
"it",
"that",
"of",
"be",
"and",
"the",
"in",
"is",
"to",
"for"
],
[
"to",
"the",
"in",
"of",
"and",
"on",
"for",
"said",
"by",
"uk"
],
[
"for",
"and",
"the",
"to",
"but",
"on",
"of",
"in",
"is",
"it"
]
] | 7.704844 | average_word_embeddings_glove.6B.300d | 0.33 | -0.026604 | 0.289131 | 0.575766 |
42 | BBC News | LDA | 20 | [
[
"he",
"in",
"to",
"the",
"was",
"his",
"and",
"it",
"of",
"but"
],
[
"of",
"she",
"was",
"said",
"to",
"her",
"the",
"ms",
"in",
"prince"
],
[
"is",
"it",
"in",
"for",
"the",
"that",
"said",
"of",
"to",
"and"
],
[
"to",
"the",
"and",
"of",
"in",
"for",
"was",
"it",
"on",
"is"
],
[
"and",
"club",
"united",
"we",
"to",
"at",
"the",
"league",
"he",
"has"
],
[
"to",
"in",
"be",
"of",
"was",
"the",
"said",
"and",
"it",
"for"
],
[
"the",
"and",
"that",
"to",
"of",
"said",
"it",
"in",
"is",
"users"
],
[
"that",
"the",
"to",
"and",
"in",
"of",
"is",
"he",
"it",
"on"
],
[
"said",
"and",
"of",
"to",
"in",
"the",
"for",
"by",
"on",
"is"
],
[
"the",
"to",
"in",
"is",
"and",
"for",
"of",
"on",
"at",
"have"
],
[
"the",
"men",
"on",
"cross",
"country",
"arrested",
"in",
"women",
"blue",
"from"
],
[
"the",
"to",
"of",
"and",
"office",
"in",
"linux",
"open",
"its",
"box"
],
[
"at",
"for",
"the",
"best",
"awards",
"theatre",
"actress",
"director",
"actor",
"musical"
],
[
"the",
"of",
"and",
"in",
"will",
"to",
"be",
"on",
"is",
"for"
],
[
"but",
"to",
"and",
"game",
"from",
"was",
"the",
"half",
"on",
"with"
],
[
"and",
"is",
"in",
"of",
"to",
"the",
"that",
"it",
"are",
"for"
],
[
"to",
"the",
"will",
"of",
"and",
"scottish",
"in",
"be",
"is",
"would"
],
[
"dvd",
"the",
"dvds",
"to",
"of",
"definition",
"high",
"be",
"and",
"studios"
],
[
"that",
"of",
"be",
"mr",
"to",
"said",
"in",
"and",
"the",
"was"
],
[
"the",
"of",
"in",
"to",
"and",
"dr",
"mr",
"jewish",
"smith",
"he"
]
] | 17.727674 | average_word_embeddings_glove.6B.300d | 0.32 | -0.017112 | 0.259832 | 0.607921 |
42 | BBC News | LDA | 30 | [
[
"he",
"the",
"to",
"was",
"his",
"had",
"and",
"in",
"of",
"for"
],
[
"the",
"was",
"she",
"in",
"ms",
"her",
"said",
"to",
"of",
"after"
],
[
"of",
"the",
"to",
"in",
"and",
"said",
"that",
"for",
"its",
"is"
],
[
"for",
"in",
"to",
"of",
"the",
"and",
"on",
"was",
"film",
"with"
],
[
"to",
"league",
"chelsea",
"united",
"the",
"club",
"arsenal",
"he",
"manchester",
"on"
],
[
"law",
"of",
"lord",
"in",
"the",
"to",
"said",
"and",
"rights",
"was"
],
[
"to",
"and",
"said",
"the",
"attacks",
"sites",
"of",
"site",
"that",
"in"
],
[
"the",
"and",
"to",
"of",
"he",
"said",
"in",
"on",
"that",
"mr"
],
[
"budget",
"japan",
"in",
"the",
"of",
"bush",
"to",
"and",
"was",
"prince"
],
[
"on",
"and",
"for",
"of",
"is",
"the",
"in",
"to",
"but",
"at"
],
[
"of",
"on",
"from",
"arrested",
"men",
"the",
"women",
"country",
"cross",
"police"
],
[
"and",
"the",
"to",
"ibm",
"software",
"linux",
"programs",
"use",
"of",
"call"
],
[
"the",
"for",
"awards",
"best",
"at",
"award",
"won",
"in",
"theatre",
"actress"
],
[
"at",
"number",
"the",
"to",
"and",
"top",
"chart",
"of",
"in",
"one"
],
[
"to",
"the",
"and",
"was",
"half",
"for",
"game",
"but",
"from",
"after"
],
[
"for",
"the",
"in",
"to",
"people",
"local",
"of",
"and",
"government",
"is"
],
[
"scottish",
"scotland",
"russian",
"executive",
"to",
"will",
"gazprom",
"the",
"and",
"in"
],
[
"the",
"dvds",
"dvd",
"studios",
"film",
"technology",
"to",
"formats",
"of",
"hollywood"
],
[
"of",
"users",
"the",
"to",
"that",
"security",
"and",
"software",
"mail",
"or"
],
[
"gas",
"contract",
"firm",
"oil",
"company",
"production",
"indian",
"contracts",
"the",
"energy"
],
[
"the",
"radio",
"and",
"china",
"in",
"of",
"to",
"on",
"will",
"is"
],
[
"within",
"companies",
"industries",
"television",
"such",
"indicated",
"research",
"led",
"film",
"career"
],
[
"to",
"race",
"world",
"in",
"olympic",
"of",
"the",
"at",
"and",
"indoor"
],
[
"mobile",
"to",
"the",
"phone",
"of",
"and",
"tv",
"people",
"service",
"are"
],
[
"the",
"and",
"to",
"irish",
"000",
"for",
"from",
"u2",
"of",
"drugs"
],
[
"of",
"to",
"the",
"and",
"he",
"has",
"been",
"for",
"in",
"was"
],
[
"it",
"in",
"is",
"for",
"and",
"to",
"that",
"of",
"the",
"as"
],
[
"to",
"the",
"of",
"be",
"for",
"in",
"said",
"and",
"that",
"is"
],
[
"wasps",
"bath",
"ask",
"thompson",
"search",
"leicester",
"computing",
"hardware",
"it",
"web"
],
[
"the",
"to",
"and",
"is",
"of",
"it",
"that",
"we",
"you",
"have"
]
] | 36.384504 | average_word_embeddings_glove.6B.300d | 0.446667 | -0.011615 | 0.215939 | 0.669564 |
42 | BBC News | LDA | 40 | [
[
"party",
"leader",
"he",
"the",
"and",
"mr",
"to",
"his",
"silk",
"as"
],
[
"kim",
"her",
"return",
"the",
"on",
"february",
"in",
"back",
"to",
"make"
],
[
"in",
"the",
"to",
"of",
"for",
"and",
"said",
"that",
"its",
"on"
],
[
"and",
"to",
"film",
"for",
"of",
"on",
"was",
"in",
"at",
"the"
],
[
"the",
"are",
"with",
"brother",
"and",
"big",
"on",
"to",
"life",
"show"
],
[
"to",
"the",
"peer",
"hunting",
"music",
"and",
"that",
"of",
"files",
"hunt"
],
[
"the",
"is",
"to",
"users",
"and",
"software",
"in",
"computer",
"of",
"programs"
],
[
"he",
"of",
"the",
"that",
"and",
"in",
"to",
"mr",
"on",
"is"
],
[
"the",
"broadband",
"japan",
"to",
"japanese",
"by",
"of",
"in",
"and",
"year"
],
[
"to",
"the",
"in",
"and",
"on",
"of",
"for",
"with",
"at",
"was"
],
[
"of",
"the",
"on",
"police",
"men",
"from",
"arrested",
"british",
"to",
"in"
],
[
"him",
"edward",
"for",
"at",
"to",
"prince",
"the",
"queen",
"harry",
"of"
],
[
"points",
"pointed",
"political",
"policy",
"policies",
"police",
"politicians",
"pointing",
"plus",
"pocket"
],
[
"and",
"to",
"in",
"the",
"super",
"of",
"group",
"said",
"its",
"at"
],
[
"irish",
"the",
"prize",
"is",
"for",
"and",
"000",
"book",
"of",
"novel"
],
[
"and",
"of",
"to",
"people",
"in",
"the",
"that",
"more",
"said",
"music"
],
[
"russia",
"the",
"its",
"state",
"yukos",
"oil",
"firm",
"company",
"gazprom",
"russian"
],
[
"studios",
"dvd",
"gas",
"dvds",
"energy",
"the",
"to",
"india",
"in",
"technology"
],
[
"of",
"the",
"to",
"sport",
"drugs",
"and",
"for",
"that",
"doping",
"collins"
],
[
"security",
"what",
"virus",
"sending",
"anti",
"mail",
"warnings",
"2004",
"itself",
"websites"
],
[
"in",
"of",
"to",
"the",
"on",
"radio",
"digital",
"air",
"and",
"will"
],
[
"points",
"pointed",
"political",
"policy",
"policies",
"police",
"politicians",
"pointing",
"plus",
"pocket"
],
[
"scottish",
"the",
"in",
"will",
"at",
"to",
"scotland",
"glasgow",
"and",
"list"
],
[
"to",
"mobile",
"the",
"and",
"phone",
"services",
"tv",
"service",
"of",
"will"
],
[
"french",
"album",
"singer",
"band",
"at",
"to",
"the",
"drug",
"music",
"and"
],
[
"in",
"the",
"england",
"to",
"robinson",
"for",
"after",
"has",
"six",
"of"
],
[
"and",
"the",
"in",
"of",
"for",
"it",
"is",
"to",
"that",
"said"
],
[
"to",
"and",
"in",
"is",
"of",
"that",
"the",
"be",
"it",
"for"
],
[
"code",
"or",
"microsoft",
"version",
"online",
"windows",
"system",
"security",
"released",
"pcs"
],
[
"and",
"to",
"it",
"is",
"that",
"the",
"of",
"you",
"not",
"in"
],
[
"the",
"her",
"in",
"she",
"to",
"was",
"has",
"year",
"said",
"but"
],
[
"to",
"search",
"site",
"the",
"users",
"sites",
"web",
"said",
"net",
"google"
],
[
"sporting",
"that",
"the",
"to",
"my",
"manager",
"director",
"and",
"we",
"frank"
],
[
"of",
"budget",
"the",
"to",
"and",
"bush",
"cards",
"be",
"spending",
"said"
],
[
"of",
"said",
"is",
"iraq",
"to",
"and",
"women",
"in",
"the",
"dr"
],
[
"the",
"of",
"in",
"he",
"for",
"to",
"and",
"his",
"it",
"is"
],
[
"in",
"it",
"bill",
"for",
"the",
"children",
"and",
"computing",
"to",
"of"
],
[
"in",
"game",
"it",
"was",
"from",
"the",
"and",
"of",
"to",
"we"
],
[
"the",
"to",
"of",
"in",
"and",
"said",
"he",
"was",
"on",
"that"
],
[
"the",
"of",
"and",
"in",
"as",
"world",
"to",
"said",
"with",
"has"
]
] | 57.039842 | average_word_embeddings_glove.6B.300d | 0.415 | -0.03507 | 0.201104 | 0.673093 |
42 | BBC News | LDA | 50 | [
[
"he",
"of",
"to",
"in",
"the",
"and",
"mr",
"his",
"for",
"said"
],
[
"took",
"number",
"the",
"box",
"last",
"weekend",
"movie",
"office",
"in",
"7m"
],
[
"of",
"said",
"the",
"and",
"to",
"in",
"for",
"that",
"its",
"on"
],
[
"the",
"for",
"on",
"of",
"in",
"to",
"and",
"film",
"was",
"best"
],
[
"manchester",
"ferguson",
"to",
"west",
"united",
"the",
"visit",
"everton",
"city",
"boss"
],
[
"guilty",
"peer",
"network",
"to",
"of",
"the",
"and",
"files",
"sharing",
"that"
],
[
"to",
"india",
"the",
"and",
"of",
"company",
"indian",
"000",
"in",
"it"
],
[
"and",
"it",
"that",
"to",
"of",
"is",
"you",
"the",
"are",
"have"
],
[
"to",
"south",
"countries",
"the",
"korea",
"japan",
"in",
"yen",
"and",
"of"
],
[
"to",
"in",
"the",
"and",
"of",
"on",
"for",
"with",
"was",
"but"
],
[
"points",
"pointed",
"political",
"policy",
"policies",
"police",
"politicians",
"pointing",
"plus",
"pocket"
],
[
"prince",
"of",
"in",
"harry",
"the",
"and",
"vehicles",
"to",
"said",
"it"
],
[
"points",
"pointed",
"political",
"policy",
"policies",
"police",
"politicians",
"pointing",
"plus",
"pocket"
],
[
"be",
"and",
"parents",
"group",
"in",
"super",
"to",
"children",
"the",
"of"
],
[
"irish",
"for",
"ireland",
"the",
"penalty",
"gara",
"and",
"try",
"driscoll",
"brian"
],
[
"the",
"for",
"in",
"to",
"of",
"and",
"people",
"more",
"is",
"said"
],
[
"and",
"the",
"he",
"head",
"prop",
"said",
"phil",
"injury",
"in",
"operation"
],
[
"harder",
"system",
"copy",
"piracy",
"copies",
"download",
"dvd",
"said",
"formats",
"new"
],
[
"and",
"to",
"the",
"ban",
"of",
"hunting",
"said",
"that",
"schools",
"hunt"
],
[
"of",
"johnny",
"wasps",
"leicester",
"what",
"the",
"van",
"strange",
"fire",
"murphy"
],
[
"to",
"china",
"the",
"of",
"and",
"in",
"on",
"is",
"chinese",
"be"
],
[
"points",
"pointed",
"political",
"policy",
"policies",
"police",
"politicians",
"pointing",
"plus",
"pocket"
],
[
"the",
"scottish",
"to",
"will",
"scotland",
"on",
"in",
"be",
"of",
"and"
],
[
"tv",
"and",
"the",
"services",
"phone",
"service",
"of",
"to",
"mobile",
"phones"
],
[
"singer",
"band",
"tour",
"album",
"french",
"to",
"the",
"at",
"and",
"u2"
],
[
"in",
"3g",
"the",
"and",
"for",
"to",
"of",
"drugs",
"has",
"will"
],
[
"the",
"and",
"is",
"of",
"to",
"it",
"in",
"for",
"that",
"said"
],
[
"that",
"for",
"of",
"is",
"in",
"to",
"be",
"it",
"and",
"the"
],
[
"centre",
"robinson",
"coach",
"squad",
"wales",
"england",
"andy",
"six",
"in",
"nations"
],
[
"and",
"of",
"music",
"the",
"it",
"to",
"is",
"that",
"in",
"song"
],
[
"in",
"she",
"was",
"said",
"the",
"to",
"her",
"year",
"it",
"but"
],
[
"search",
"site",
"the",
"users",
"to",
"net",
"sites",
"web",
"of",
"internet"
],
[
"points",
"pointed",
"political",
"policy",
"policies",
"police",
"politicians",
"pointing",
"plus",
"pocket"
],
[
"of",
"to",
"the",
"in",
"rates",
"and",
"brown",
"spending",
"will",
"economy"
],
[
"and",
"of",
"to",
"in",
"dvd",
"the",
"high",
"sony",
"be",
"is"
],
[
"the",
"his",
"for",
"he",
"to",
"in",
"of",
"it",
"and",
"is"
],
[
"computing",
"of",
"the",
"in",
"hardware",
"to",
"will",
"and",
"it",
"customers"
],
[
"the",
"was",
"of",
"and",
"in",
"it",
"to",
"game",
"we",
"but"
],
[
"and",
"of",
"he",
"the",
"to",
"in",
"was",
"mr",
"his",
"said"
],
[
"security",
"as",
"of",
"in",
"and",
"that",
"to",
"the",
"said",
"are"
],
[
"of",
"world",
"olympic",
"in",
"the",
"to",
"and",
"indoor",
"at",
"britain"
],
[
"points",
"pointed",
"political",
"policy",
"policies",
"police",
"politicians",
"pointing",
"plus",
"pocket"
],
[
"party",
"and",
"in",
"of",
"to",
"labour",
"the",
"election",
"said",
"he"
],
[
"shot",
"hunting",
"remote",
"session",
"shooting",
"texas",
"control",
"wild",
"camera",
"lets"
],
[
"points",
"pointed",
"political",
"policy",
"policies",
"police",
"politicians",
"pointing",
"plus",
"pocket"
],
[
"said",
"they",
"to",
"police",
"the",
"and",
"were",
"in",
"have",
"their"
],
[
"points",
"pointed",
"political",
"policy",
"policies",
"police",
"politicians",
"pointing",
"plus",
"pocket"
],
[
"the",
"be",
"on",
"of",
"and",
"mr",
"he",
"said",
"to",
"in"
],
[
"the",
"of",
"to",
"and",
"for",
"in",
"on",
"said",
"be",
"by"
],
[
"points",
"pointed",
"political",
"policy",
"policies",
"police",
"politicians",
"pointing",
"plus",
"pocket"
]
] | 68.63326 | average_word_embeddings_glove.6B.300d | 0.362 | -0.068876 | 0.200801 | 0.682867 |
42 | BBC News | Top2Vec | 10 | [
[
"time",
"coming",
"play",
"came",
"come",
"team",
"going",
"players",
"chance",
"win"
],
[
"year",
"winning",
"came",
"second",
"time",
"world",
"win",
"coming",
"final",
"took"
],
[
"win",
"tournament",
"play",
"round",
"second",
"coming",
"final",
"match",
"time",
"fourth"
],
[
"come",
"government",
"time",
"year",
"week",
"likely",
"month",
"expected",
"saying",
"make"
],
[
"time",
"best",
"come",
"came",
"year",
"years",
"way",
"movie",
"coming",
"later"
],
[
"way",
"come",
"example",
"time",
"make",
"need",
"use",
"internet",
"instance",
"making"
]
] | 4.874947 | average_word_embeddings_glove.6B.300d | 0.633333 | -0.058725 | 0.246285 | 0.785235 |
42 | BBC News | Top2Vec | 20 | [
[
"time",
"coming",
"play",
"came",
"come",
"team",
"going",
"players",
"chance",
"win"
],
[
"year",
"winning",
"came",
"second",
"time",
"world",
"win",
"coming",
"final",
"took"
],
[
"win",
"tournament",
"play",
"round",
"second",
"coming",
"final",
"match",
"time",
"fourth"
],
[
"come",
"government",
"time",
"year",
"week",
"likely",
"month",
"expected",
"saying",
"make"
],
[
"time",
"best",
"come",
"came",
"year",
"years",
"way",
"movie",
"coming",
"later"
],
[
"way",
"come",
"example",
"time",
"make",
"need",
"use",
"internet",
"instance",
"making"
]
] | 4.603987 | average_word_embeddings_glove.6B.300d | 0.633333 | -0.058725 | 0.246285 | 0.776698 |
42 | BBC News | Top2Vec | 30 | [
[
"time",
"coming",
"play",
"came",
"come",
"team",
"going",
"players",
"chance",
"win"
],
[
"year",
"winning",
"came",
"second",
"time",
"world",
"win",
"coming",
"final",
"took"
],
[
"win",
"tournament",
"play",
"round",
"second",
"coming",
"final",
"match",
"time",
"fourth"
],
[
"come",
"government",
"time",
"year",
"week",
"likely",
"month",
"expected",
"saying",
"make"
],
[
"time",
"best",
"come",
"came",
"year",
"years",
"way",
"movie",
"coming",
"later"
],
[
"way",
"come",
"example",
"time",
"make",
"need",
"use",
"internet",
"instance",
"making"
]
] | 4.519022 | average_word_embeddings_glove.6B.300d | 0.633333 | -0.058725 | 0.246285 | 0.784775 |
42 | BBC News | Top2Vec | 40 | [
[
"time",
"coming",
"play",
"came",
"come",
"team",
"going",
"players",
"chance",
"win"
],
[
"year",
"winning",
"came",
"second",
"time",
"world",
"win",
"coming",
"final",
"took"
],
[
"win",
"tournament",
"play",
"round",
"second",
"coming",
"final",
"match",
"time",
"fourth"
],
[
"come",
"government",
"time",
"year",
"week",
"likely",
"month",
"expected",
"saying",
"make"
],
[
"time",
"best",
"come",
"came",
"year",
"years",
"way",
"movie",
"coming",
"later"
],
[
"way",
"come",
"example",
"time",
"make",
"need",
"use",
"internet",
"instance",
"making"
]
] | 5.060041 | average_word_embeddings_glove.6B.300d | 0.633333 | -0.058725 | 0.246285 | 0.780766 |
42 | BBC News | Top2Vec | 50 | [
[
"time",
"coming",
"play",
"came",
"come",
"team",
"going",
"players",
"chance",
"win"
],
[
"year",
"winning",
"came",
"second",
"time",
"world",
"win",
"coming",
"final",
"took"
],
[
"win",
"tournament",
"play",
"round",
"second",
"coming",
"final",
"match",
"time",
"fourth"
],
[
"come",
"government",
"time",
"year",
"week",
"likely",
"month",
"expected",
"saying",
"make"
],
[
"time",
"best",
"come",
"came",
"year",
"years",
"way",
"movie",
"coming",
"later"
],
[
"way",
"come",
"example",
"time",
"make",
"need",
"use",
"internet",
"instance",
"making"
]
] | 4.640697 | average_word_embeddings_glove.6B.300d | 0.633333 | -0.058725 | 0.246285 | 0.780457 |
42 | BBC News | S³ | 10 | [
[
"athletics",
"sued",
"formally",
"consortium",
"broadcasting",
"doping",
"uefa",
"olympic",
"denies",
"denied"
],
[
"movie",
"world",
"country",
"american",
"russian",
"television",
"people",
"media",
"chinese",
"want"
],
[
"doubles",
"spaniard",
"roddick",
"slam",
"wimbledon",
"federer",
"seed",
"henman",
"singles",
"upset"
],
[
"risks",
"stability",
"fundamental",
"sustainable",
"implications",
"processes",
"innovation",
"efficiency",
"efficient",
"enhance"
],
[
"midfielder",
"mourinho",
"arsenal",
"striker",
"chelsea",
"coach",
"liverpool",
"everton",
"football",
"league"
],
[
"film",
"scorsese",
"actress",
"starring",
"actor",
"bafta",
"movie",
"films",
"comedy",
"oscar"
],
[
"adults",
"unemployment",
"children",
"population",
"poverty",
"disease",
"child",
"infected",
"living",
"age"
],
[
"wireless",
"pc",
"desktop",
"xbox",
"playstation",
"pcs",
"handsets",
"handheld",
"digital",
"ipod"
],
[
"viruses",
"drug",
"plant",
"doping",
"athletes",
"facilities",
"water",
"virus",
"drugs",
"infected"
],
[
"championship",
"championships",
"athletics",
"indoor",
"rugby",
"olympic",
"medal",
"olympics",
"cup",
"prix"
]
] | 1.044015 | average_word_embeddings_glove.6B.300d | 0.95 | -0.222157 | 0.26598 | 0.925145 |
42 | BBC News | S³ | 20 | [
[
"pension",
"tax",
"pensions",
"taxes",
"taxpayers",
"savings",
"budget",
"legislation",
"benefits",
"income"
],
[
"roddick",
"spaniard",
"federer",
"henman",
"seed",
"ranked",
"doubles",
"wimbledon",
"bt",
"tennis"
],
[
"barcelona",
"women",
"madrid",
"milan",
"italian",
"spanish",
"brazilian",
"people",
"woman",
"city"
],
[
"band",
"song",
"songs",
"singer",
"albums",
"album",
"chart",
"pop",
"rock",
"rap"
],
[
"nations",
"countries",
"european",
"eu",
"summit",
"france",
"ministers",
"imf",
"monetary",
"euro"
],
[
"olivier",
"french",
"dvd",
"nicolas",
"yen",
"prop",
"playing",
"plays",
"argentina",
"roddick"
],
[
"university",
"education",
"economist",
"health",
"studies",
"professor",
"correspondent",
"services",
"foundation",
"research"
],
[
"operate",
"seats",
"tickets",
"pay",
"airlines",
"fly",
"stake",
"owners",
"passengers",
"shareholders"
],
[
"guilty",
"judge",
"doping",
"prosecutors",
"convicted",
"lawsuit",
"charges",
"pleaded",
"filed",
"criminal"
],
[
"pence",
"shareholder",
"stock",
"deutsche",
"shares",
"vodafone",
"shareholders",
"takeover",
"investors",
"merger"
],
[
"strategic",
"russia",
"oil",
"yukos",
"producer",
"energy",
"gas",
"khodorkovsky",
"natural",
"gazprom"
],
[
"striker",
"midfielder",
"liverpool",
"chelsea",
"mourinho",
"arsenal",
"goalkeeper",
"everton",
"defender",
"barcelona"
],
[
"sequel",
"complex",
"shadow",
"chapter",
"tale",
"novel",
"spider",
"historic",
"historical",
"book"
],
[
"letwin",
"10bn",
"1bn",
"3bn",
"000m",
"9m",
"gara",
"2bn",
"dems",
"2m"
],
[
"blair",
"minister",
"democrat",
"prime",
"leader",
"election",
"secretary",
"mp",
"deputy",
"cabinet"
],
[
"film",
"starring",
"bafta",
"award",
"actress",
"actor",
"awards",
"oscar",
"films",
"directed"
],
[
"sport",
"households",
"premium",
"seats",
"luxury",
"tv",
"broadcasters",
"television",
"broadcaster",
"cent"
],
[
"wealth",
"genre",
"culture",
"values",
"racist",
"influence",
"opinions",
"belief",
"moral",
"passion"
],
[
"everybody",
"guys",
"really",
"bit",
"somebody",
"pretty",
"wrong",
"absolutely",
"feel",
"thing"
],
[
"penalty",
"penalties",
"terror",
"broadcast",
"viewers",
"violence",
"fraud",
"opponents",
"illegal",
"copyright"
]
] | 1.05877 | average_word_embeddings_glove.6B.300d | 0.98 | -0.208191 | 0.270039 | 0.929995 |
42 | BBC News | S³ | 30 | [
[
"traffic",
"mail",
"sites",
"landscape",
"royal",
"studies",
"cultural",
"correspondent",
"service",
"airline"
],
[
"pence",
"shareholders",
"shares",
"deutsche",
"frankfurt",
"stock",
"takeover",
"merger",
"exchange",
"index"
],
[
"son",
"friend",
"maker",
"convicted",
"father",
"ibm",
"daughter",
"wife",
"born",
"died"
],
[
"realise",
"3bn",
"gamers",
"9m",
"attitude",
"sensible",
"ambition",
"enjoy",
"genre",
"enjoying"
],
[
"schools",
"college",
"mac",
"students",
"attend",
"teachers",
"screening",
"education",
"association",
"commissioner"
],
[
"engines",
"brands",
"plant",
"cars",
"manufacturing",
"engine",
"factory",
"vehicles",
"sales",
"car"
],
[
"nominations",
"awards",
"rankings",
"categories",
"winners",
"chart",
"award",
"outstanding",
"000",
"prizes"
],
[
"diaries",
"rafael",
"jose",
"madrid",
"milan",
"barcelona",
"brazilian",
"maria",
"spanish",
"manifesto"
],
[
"poverty",
"disease",
"aids",
"developing",
"africa",
"poorest",
"drugs",
"virus",
"infected",
"drug"
],
[
"collection",
"loan",
"loans",
"banks",
"items",
"departments",
"department",
"securities",
"books",
"housing"
],
[
"taxpayers",
"savings",
"tax",
"pensions",
"taxes",
"pension",
"retirement",
"payments",
"budget",
"bills"
],
[
"spam",
"google",
"web",
"mail",
"malicious",
"viruses",
"virus",
"spammers",
"websites",
"software"
],
[
"jean",
"gara",
"nicolas",
"french",
"olivier",
"france",
"italian",
"spain",
"german",
"italy"
],
[
"metres",
"squad",
"athletes",
"bomb",
"relay",
"football",
"scrum",
"honours",
"olympics",
"rugby"
],
[
"findings",
"gmt",
"department",
"contained",
"assess",
"committee",
"environmental",
"carefully",
"complex",
"departments"
],
[
"mortgage",
"property",
"homes",
"buildings",
"bank",
"secure",
"land",
"chase",
"housing",
"units"
],
[
"keyboard",
"engineering",
"systems",
"laboratory",
"signals",
"device",
"capability",
"processes",
"computing",
"technology"
],
[
"society",
"research",
"art",
"artists",
"wealth",
"auction",
"institute",
"children",
"foundation",
"collection"
],
[
"song",
"album",
"songs",
"albums",
"band",
"singer",
"chart",
"rap",
"pop",
"charts"
],
[
"eu",
"trade",
"copyright",
"legislation",
"ban",
"restrictions",
"commerce",
"laws",
"regulations",
"promote"
],
[
"zealand",
"trading",
"index",
"000m",
"assembly",
"mixed",
"capt",
"pence",
"bell",
"commons"
],
[
"century",
"underground",
"revolution",
"lessons",
"abandoned",
"church",
"square",
"theme",
"gazprom",
"radical"
],
[
"rugby",
"tour",
"tournament",
"tennis",
"championships",
"wimbledon",
"sport",
"championship",
"club",
"football"
],
[
"slowing",
"economists",
"gdp",
"slowdown",
"growth",
"inflation",
"outlook",
"rate",
"recession",
"decline"
],
[
"editor",
"writer",
"sir",
"appointed",
"published",
"author",
"lord",
"novel",
"secretary",
"minister"
],
[
"minutes",
"france",
"seconds",
"really",
"minute",
"score",
"ball",
"right",
"french",
"got"
],
[
"spokeswoman",
"blunkett",
"spokesman",
"said",
"told",
"jeremy",
"jamie",
"clothes",
"kate",
"letwin"
],
[
"brussels",
"imports",
"engines",
"parliament",
"fuel",
"eu",
"britain",
"items",
"text",
"straw"
],
[
"playing",
"plays",
"fat",
"players",
"played",
"play",
"indian",
"yen",
"referee",
"board"
],
[
"forum",
"hosted",
"jim",
"breakfast",
"bob",
"sports",
"marketing",
"advertising",
"host",
"agriculture"
]
] | 1.185691 | average_word_embeddings_glove.6B.300d | 0.946667 | -0.274609 | 0.230971 | 0.923831 |
42 | BBC News | S³ | 40 | [
[
"doping",
"drugs",
"drug",
"tests",
"suspension",
"sprinter",
"tested",
"banned",
"positive",
"athletes"
],
[
"lewsey",
"andrew",
"davies",
"bafta",
"stephen",
"bell",
"edinburgh",
"australia",
"silk",
"wales"
],
[
"signing",
"contract",
"contracts",
"lewsey",
"scrum",
"relay",
"honours",
"woodward",
"football",
"replacements"
],
[
"score",
"goal",
"pass",
"penalty",
"kick",
"seconds",
"minutes",
"minute",
"scored",
"shot"
],
[
"billions",
"dollars",
"worth",
"million",
"billion",
"bought",
"buy",
"millions",
"invest",
"distribute"
],
[
"criminals",
"suspected",
"police",
"arrested",
"killing",
"killed",
"assault",
"convicted",
"detained",
"suspects"
],
[
"retail",
"operations",
"consultancy",
"planning",
"unit",
"management",
"chain",
"independent",
"gartner",
"assessment"
],
[
"university",
"director",
"spokesman",
"david",
"van",
"spokeswoman",
"said",
"robert",
"born",
"institute"
],
[
"church",
"1960s",
"1970s",
"records",
"contains",
"warner",
"abandoned",
"century",
"label",
"revolution"
],
[
"festive",
"consultancy",
"gartner",
"holiday",
"dvds",
"piracy",
"oil",
"christmas",
"games",
"retailers"
],
[
"storage",
"records",
"lord",
"stolen",
"ticket",
"tickets",
"collection",
"chambers",
"stored",
"keeper"
],
[
"engineering",
"indian",
"industries",
"sectors",
"sector",
"telecommunications",
"india",
"telecoms",
"stocks",
"gained"
],
[
"festival",
"theatre",
"opera",
"italy",
"concert",
"france",
"frankfurt",
"paris",
"germany",
"rome"
],
[
"scrum",
"corry",
"damien",
"olivier",
"flanker",
"replacements",
"capt",
"nicolas",
"france",
"jean"
],
[
"movies",
"film",
"films",
"starring",
"movie",
"sequel",
"hollywood",
"starred",
"blockbuster",
"directed"
],
[
"technical",
"direction",
"data",
"depth",
"keyboard",
"speeds",
"engine",
"pilot",
"recorder",
"miles"
],
[
"bank",
"sessions",
"sec",
"straw",
"session",
"agriculture",
"forum",
"meeting",
"liam",
"rubbish"
],
[
"ankle",
"knee",
"hamstring",
"surgery",
"injuries",
"injury",
"hip",
"foot",
"injured",
"suffered"
],
[
"voters",
"unions",
"convention",
"california",
"democratic",
"union",
"voting",
"votes",
"candidates",
"legislation"
],
[
"dem",
"khodorkovsky",
"nokia",
"exports",
"producing",
"contract",
"production",
"output",
"workforce",
"eu"
],
[
"frankfurt",
"deutsche",
"pence",
"shareholders",
"takeover",
"exchange",
"merger",
"euros",
"london",
"swap"
],
[
"area",
"water",
"gas",
"underground",
"shopping",
"shops",
"popular",
"land",
"areas",
"fuel"
],
[
"scotland",
"scrum",
"wales",
"rugby",
"zealand",
"ireland",
"australia",
"england",
"irish",
"squad"
],
[
"consoles",
"nintendo",
"console",
"playstation",
"xbox",
"psp",
"gamers",
"gaming",
"handheld",
"pc"
],
[
"streets",
"night",
"concert",
"rally",
"morning",
"afternoon",
"evening",
"festival",
"scheduled",
"saturday"
],
[
"antonio",
"manuel",
"president",
"wife",
"san",
"born",
"son",
"lived",
"friend",
"daughter"
],
[
"liberal",
"voters",
"election",
"candidate",
"conservatives",
"democratic",
"party",
"polls",
"votes",
"elections"
],
[
"total",
"votes",
"parliamentary",
"trophy",
"euros",
"000",
"trillion",
"list",
"lists",
"collected"
],
[
"lords",
"judge",
"constitutional",
"supreme",
"court",
"courts",
"appeals",
"laws",
"granted",
"law"
],
[
"health",
"screening",
"kong",
"college",
"hong",
"sex",
"blunkett",
"mac",
"sessions",
"schools"
],
[
"equipment",
"battery",
"manufacturing",
"eu",
"manufacturers",
"chip",
"producers",
"netherlands",
"machine",
"tommy"
],
[
"400m",
"800m",
"200m",
"500m",
"100m",
"1500m",
"hurdles",
"60m",
"medallist",
"sprint"
],
[
"director",
"watching",
"outdoor",
"arts",
"institute",
"cameras",
"sports",
"attended",
"university",
"forum"
],
[
"subscribers",
"handsets",
"wireless",
"mobile",
"3g",
"broadband",
"vodafone",
"phones",
"phone",
"handset"
],
[
"aid",
"poverty",
"nations",
"aids",
"africa",
"poorest",
"imf",
"tsunami",
"summit",
"countries"
],
[
"commerce",
"unveiled",
"remarks",
"unveil",
"summit",
"conference",
"meeting",
"forum",
"unveils",
"secretary"
],
[
"deficit",
"budget",
"funding",
"tax",
"taxes",
"budgets",
"deficits",
"spending",
"cuts",
"subsidies"
],
[
"software",
"malicious",
"virus",
"viruses",
"linux",
"windows",
"servers",
"microsoft",
"desktop",
"mac"
],
[
"newcastle",
"leeds",
"norwich",
"everton",
"bristol",
"manchester",
"sheffield",
"cardiff",
"leicester",
"liverpool"
],
[
"drug",
"research",
"cell",
"trials",
"drugs",
"analysis",
"laboratory",
"terrorism",
"technologies",
"studies"
]
] | 1.201045 | average_word_embeddings_glove.6B.300d | 0.92 | -0.234733 | 0.255469 | 0.93409 |
42 | BBC News | S³ | 50 | [
[
"theatre",
"opera",
"festival",
"concert",
"exhibition",
"performances",
"auction",
"conducted",
"art",
"venue"
],
[
"oscars",
"award",
"bafta",
"awards",
"nominations",
"oscar",
"prize",
"nomination",
"nominees",
"nominated"
],
[
"session",
"street",
"rally",
"church",
"trust",
"rubbish",
"bank",
"forum",
"councils",
"community"
],
[
"centre",
"departments",
"programmes",
"programme",
"centres",
"bafta",
"chart",
"labour",
"production",
"agriculture"
],
[
"oil",
"fuel",
"gas",
"prices",
"price",
"water",
"natural",
"supply",
"energy",
"reserves"
],
[
"praise",
"bosses",
"beat",
"score",
"deserved",
"scorsese",
"record",
"proud",
"fans",
"margin"
],
[
"board",
"arts",
"programs",
"science",
"environmental",
"committee",
"carefully",
"findings",
"program",
"index"
],
[
"engine",
"vehicles",
"car",
"engines",
"auto",
"cars",
"models",
"motorcycle",
"driving",
"model"
],
[
"bbc",
"broadcasting",
"tv",
"radio",
"broadcaster",
"channel",
"channels",
"broadcast",
"television",
"presenter"
],
[
"jobs",
"correspondent",
"hunt",
"workforce",
"chase",
"toshiba",
"trophy",
"freedom",
"apple",
"pace"
],
[
"laboratory",
"technology",
"engineering",
"electronics",
"research",
"technologies",
"toshiba",
"equipment",
"devices",
"energy"
],
[
"holiday",
"retailers",
"sales",
"stores",
"retail",
"store",
"shopping",
"christmas",
"clothing",
"shops"
],
[
"india",
"session",
"capt",
"sessions",
"asian",
"telecoms",
"indian",
"skipper",
"muslim",
"ministers"
],
[
"pass",
"goal",
"minute",
"scored",
"penalty",
"kick",
"seconds",
"minutes",
"goalkeeper",
"ball"
],
[
"seed",
"bt",
"roddick",
"henman",
"federer",
"spaniard",
"wimbledon",
"ranked",
"nicolas",
"sets"
],
[
"shot",
"jean",
"brazilian",
"right",
"ball",
"french",
"guy",
"veteran",
"manuel",
"injured"
],
[
"inflation",
"economists",
"rate",
"slowdown",
"slowing",
"rates",
"economist",
"forecast",
"unemployment",
"recession"
],
[
"console",
"consoles",
"psp",
"xbox",
"playstation",
"nintendo",
"handheld",
"game",
"gamers",
"gaming"
],
[
"hunting",
"gaming",
"underground",
"police",
"illegal",
"suspected",
"arrested",
"bands",
"hunt",
"betting"
],
[
"chicago",
"mexico",
"american",
"los",
"latin",
"america",
"states",
"california",
"brazil",
"angeles"
],
[
"cards",
"bank",
"security",
"currency",
"units",
"battery",
"card",
"roles",
"prison",
"regime"
],
[
"madrid",
"mourinho",
"jose",
"barcelona",
"milan",
"spanish",
"rafael",
"manuel",
"arsenal",
"chelsea"
],
[
"france",
"pictures",
"dvds",
"studios",
"dvd",
"french",
"format",
"images",
"disc",
"formats"
],
[
"party",
"democratic",
"election",
"seats",
"elections",
"votes",
"constituency",
"electoral",
"liberal",
"democrat"
],
[
"recognised",
"manufacturing",
"settled",
"confirm",
"relations",
"registered",
"asylum",
"trade",
"exports",
"interviewed"
],
[
"barcelona",
"lopez",
"milan",
"rafael",
"spokeswoman",
"madrid",
"antonio",
"maria",
"10bn",
"el"
],
[
"asset",
"analysis",
"analyst",
"strategist",
"stability",
"securities",
"flow",
"bank",
"trophy",
"alcohol"
],
[
"aged",
"adults",
"age",
"47",
"households",
"65",
"39",
"44",
"52",
"34"
],
[
"tsunami",
"poverty",
"relief",
"aid",
"disaster",
"aids",
"poorest",
"africa",
"estimated",
"victims"
],
[
"festival",
"violence",
"activists",
"muslim",
"racist",
"poll",
"jewish",
"motivated",
"mainstream",
"documentary"
],
[
"blunkett",
"oaten",
"pensioners",
"henman",
"gareth",
"australian",
"ferguson",
"corry",
"clive",
"nigel"
],
[
"exports",
"manufacturers",
"consumer",
"imports",
"export",
"minister",
"goods",
"household",
"loans",
"manufacturing"
],
[
"ireland",
"dublin",
"scottish",
"edinburgh",
"irish",
"glasgow",
"developers",
"theatre",
"windows",
"scotland"
],
[
"sports",
"watching",
"attended",
"outdoor",
"watched",
"attend",
"park",
"watch",
"exhibition",
"stadium"
],
[
"fat",
"yen",
"prop",
"french",
"players",
"plays",
"friendly",
"play",
"alcohol",
"player"
],
[
"school",
"male",
"german",
"students",
"russian",
"boys",
"alcohol",
"language",
"children",
"female"
],
[
"agency",
"approved",
"national",
"legislation",
"territory",
"proposal",
"approval",
"survey",
"ban",
"regulations"
],
[
"contract",
"signed",
"contracts",
"agreement",
"signing",
"deal",
"commitment",
"ambitious",
"2003",
"professional"
],
[
"award",
"prizes",
"awards",
"awarded",
"prize",
"ces",
"outstanding",
"medal",
"winners",
"achievement"
],
[
"queen",
"shot",
"british",
"gold",
"american",
"fa",
"shareholders",
"united",
"ball",
"red"
],
[
"court",
"legal",
"laws",
"courts",
"supreme",
"rights",
"appeals",
"constitutional",
"copyright",
"law"
],
[
"european",
"countries",
"eu",
"nations",
"trade",
"summit",
"ministers",
"brussels",
"imf",
"forum"
],
[
"hopefully",
"nominations",
"going",
"wait",
"nomination",
"happen",
"sure",
"lot",
"everybody",
"oscars"
],
[
"tennis",
"bodies",
"outdoor",
"lewsey",
"association",
"historical",
"body",
"motion",
"inquiry",
"experiences"
],
[
"mac",
"commissioner",
"index",
"screening",
"breakfast",
"intel",
"hong",
"kong",
"van",
"keane"
],
[
"told",
"said",
"chairman",
"spokesman",
"director",
"institute",
"robert",
"van",
"steve",
"spokeswoman"
],
[
"blair",
"prime",
"minister",
"scandal",
"allegations",
"blunkett",
"tony",
"downing",
"bush",
"talks"
],
[
"military",
"command",
"security",
"officers",
"police",
"senior",
"vladimir",
"officer",
"service",
"intelligence"
],
[
"china",
"japan",
"yuan",
"beijing",
"yen",
"kong",
"hong",
"chinese",
"japanese",
"korea"
],
[
"ofcom",
"executive",
"workforce",
"annual",
"gara",
"chairman",
"broadband",
"waste",
"statistics",
"review"
]
] | 1.397449 | average_word_embeddings_glove.6B.300d | 0.898 | -0.248271 | 0.244541 | 0.917434 |
42 | BBC News | CombinedTM | 10 | [
[
"of",
"the",
"to",
"and",
"is",
"that",
"it",
"in",
"for",
"be"
],
[
"economy",
"year",
"china",
"us",
"yukos",
"its",
"bank",
"oil",
"growth",
"prices"
],
[
"agency",
"low",
"once",
"executive",
"itself",
"sell",
"car",
"virus",
"mail",
"compensation"
],
[
"comedy",
"band",
"awards",
"film",
"album",
"director",
"nominated",
"prize",
"actress",
"award"
],
[
"was",
"his",
"half",
"after",
"ireland",
"second",
"final",
"in",
"wales",
"victory"
],
[
"steve",
"jump",
"soul",
"taking",
"named",
"argument",
"men",
"holder",
"else",
"indoor"
],
[
"chancellor",
"mr",
"said",
"would",
"he",
"brown",
"party",
"leader",
"election",
"prime"
],
[
"executive",
"spam",
"board",
"shares",
"poor",
"stock",
"major",
"profits",
"potential",
"mail"
],
[
"to",
"technology",
"and",
"mobile",
"of",
"are",
"digital",
"is",
"more",
"people"
],
[
"injury",
"arsenal",
"madrid",
"situation",
"barcelona",
"league",
"game",
"united",
"chelsea",
"face"
]
] | 480.499243 | average_word_embeddings_glove.6B.300d | 0.93 | -0.06176 | 0.163826 | 0.817069 |
42 | BBC News | CombinedTM | 20 | [
[
"bid",
"exchange",
"gazprom",
"oil",
"deal",
"deutsche",
"russian",
"london",
"company",
"talks"
],
[
"north",
"office",
"release",
"place",
"starring",
"box",
"films",
"7m",
"1m",
"weekend"
],
[
"in",
"her",
"was",
"title",
"world",
"indoor",
"record",
"for",
"olympic",
"he"
],
[
"and",
"it",
"is",
"definition",
"that",
"dvd",
"the",
"mini",
"you",
"mac"
],
[
"mr",
"silk",
"eu",
"was",
"his",
"the",
"lord",
"advice",
"he",
"said"
],
[
"police",
"court",
"would",
"said",
"plans",
"hunting",
"local",
"workers",
"id",
"murder"
],
[
"million",
"mobile",
"mobiles",
"3g",
"uk",
"broadband",
"will",
"tv",
"operators",
"phone"
],
[
"johnson",
"appointed",
"wonderful",
"matter",
"seats",
"age",
"longer",
"charged",
"poverty",
"heard"
],
[
"programs",
"software",
"spam",
"users",
"security",
"virus",
"windows",
"microsoft",
"viruses",
"websites"
],
[
"indian",
"shares",
"profits",
"products",
"low",
"euros",
"poor",
"latin",
"banks",
"profit"
],
[
"half",
"minutes",
"wales",
"robinson",
"nations",
"ireland",
"after",
"try",
"england",
"ball"
],
[
"interest",
"prices",
"dollar",
"rates",
"growth",
"rate",
"fall",
"2004",
"rise",
"december"
],
[
"of",
"the",
"to",
"and",
"that",
"song",
"robbie",
"25",
"in",
"it"
],
[
"gives",
"required",
"energy",
"winter",
"reason",
"employers",
"tsunami",
"remember",
"proposed",
"regulation"
],
[
"team",
"injury",
"cup",
"playing",
"play",
"round",
"against",
"match",
"start",
"league"
],
[
"mr",
"campaign",
"labour",
"brown",
"howard",
"he",
"prime",
"chancellor",
"election",
"blair"
],
[
"assault",
"committed",
"talking",
"women",
"west",
"tried",
"manager",
"comes",
"manage",
"white"
],
[
"to",
"hip",
"of",
"it",
"that",
"is",
"and",
"the",
"hop",
"you"
],
[
"film",
"rock",
"award",
"comedy",
"night",
"director",
"awards",
"star",
"actor",
"prize"
],
[
"tax",
"of",
"budget",
"for",
"yukos",
"said",
"to",
"by",
"aids",
"russian"
]
] | 363.609921 | average_word_embeddings_glove.6B.300d | 0.88 | -0.040945 | 0.160117 | 0.837652 |
42 | BBC News | CombinedTM | 30 | [
[
"the",
"and",
"of",
"is",
"hop",
"that",
"hip",
"radio",
"spanish",
"it"
],
[
"controlled",
"seed",
"attention",
"expert",
"feels",
"influence",
"ready",
"opinion",
"forces",
"seats"
],
[
"seconds",
"final",
"second",
"first",
"title",
"set",
"was",
"fourth",
"after",
"australian"
],
[
"aids",
"russian",
"bush",
"india",
"yukos",
"budget",
"tax",
"spending",
"by",
"oil"
],
[
"rise",
"rate",
"economy",
"economists",
"growth",
"prices",
"rates",
"interest",
"quarter",
"rises"
],
[
"microsoft",
"search",
"programs",
"software",
"security",
"virus",
"windows",
"viruses",
"spam",
"source"
],
[
"the",
"to",
"liverpool",
"his",
"he",
"in",
"that",
"of",
"and",
"stadium"
],
[
"after",
"minutes",
"hodgson",
"six",
"two",
"side",
"england",
"half",
"try",
"france"
],
[
"missing",
"greek",
"athletics",
"tests",
"test",
"doping",
"charges",
"drugs",
"she",
"olympics"
],
[
"its",
"us",
"china",
"oil",
"yukos",
"company",
"trade",
"bankruptcy",
"dollar",
"russian"
],
[
"awards",
"comedy",
"ceremony",
"band",
"album",
"award",
"rock",
"actor",
"categories",
"sky"
],
[
"holmes",
"gold",
"world",
"championships",
"collins",
"indoor",
"race",
"cross",
"champion",
"olympic"
],
[
"lord",
"police",
"said",
"would",
"home",
"plans",
"lords",
"secretary",
"clarke",
"law"
],
[
"team",
"sign",
"newcastle",
"club",
"arsenal",
"manager",
"situation",
"play",
"we",
"robinson"
],
[
"broadband",
"calls",
"fast",
"psp",
"bt",
"million",
"uk",
"speed",
"sony",
"service"
],
[
"centres",
"attack",
"attacks",
"are",
"sites",
"traffic",
"to",
"said",
"be",
"were"
],
[
"guys",
"lions",
"zealand",
"woodward",
"six",
"cup",
"wales",
"nations",
"rugby",
"tennis"
],
[
"definition",
"games",
"will",
"format",
"dvd",
"dvds",
"high",
"ray",
"generation",
"and"
],
[
"can",
"of",
"party",
"the",
"silk",
"to",
"and",
"premium",
"they",
"that"
],
[
"his",
"was",
"for",
"who",
"she",
"charles",
"her",
"lee",
"film",
"in"
],
[
"mac",
"file",
"that",
"mini",
"it",
"machine",
"said",
"and",
"to",
"sharing"
],
[
"starring",
"box",
"office",
"1m",
"festival",
"day",
"films",
"sequel",
"chart",
"release"
],
[
"rose",
"profits",
"value",
"january",
"2004",
"strong",
"figures",
"saw",
"demand",
"growth"
],
[
"shareholders",
"deal",
"talks",
"deutsche",
"bid",
"workers",
"union",
"unions",
"takeover",
"exchange"
],
[
"his",
"prime",
"iraq",
"he",
"was",
"advice",
"mr",
"blair",
"war",
"minister"
],
[
"premiership",
"scotland",
"simon",
"newcastle",
"face",
"saturday",
"capt",
"callaghan",
"united",
"league"
],
[
"banks",
"company",
"raise",
"airlines",
"fuel",
"iraq",
"program",
"shares",
"financial",
"revenue"
],
[
"areas",
"gas",
"energy",
"giant",
"elections",
"needs",
"pensions",
"payments",
"winter",
"gives"
],
[
"tv",
"music",
"phones",
"more",
"mobiles",
"gartner",
"people",
"mobile",
"services",
"digital"
],
[
"party",
"mr",
"brown",
"labour",
"election",
"he",
"howard",
"conservative",
"blair",
"tory"
]
] | 396.802948 | average_word_embeddings_glove.6B.300d | 0.876667 | -0.002967 | 0.157151 | 0.843641 |
42 | BBC News | CombinedTM | 40 | [
[
"police",
"asylum",
"they",
"hunting",
"be",
"children",
"cards",
"guilty",
"care",
"child"
],
[
"federer",
"after",
"goal",
"break",
"second",
"serve",
"ball",
"minutes",
"half",
"shot"
],
[
"recall",
"additional",
"mails",
"drug",
"safety",
"potential",
"vehicles",
"critical",
"social",
"programs"
],
[
"january",
"figures",
"december",
"sales",
"rose",
"retail",
"2004",
"stores",
"strong",
"growth"
],
[
"is",
"and",
"hop",
"hip",
"rap",
"to",
"spanish",
"it",
"that",
"radio"
],
[
"club",
"manager",
"sport",
"we",
"team",
"arsenal",
"ve",
"admitted",
"have",
"league"
],
[
"ipod",
"mini",
"mac",
"apple",
"search",
"jobs",
"design",
"desktop",
"storage",
"computer"
],
[
"his",
"mr",
"book",
"he",
"prime",
"stand",
"blair",
"allies",
"brown",
"had"
],
[
"he",
"that",
"it",
"the",
"and",
"liverpool",
"is",
"we",
"of",
"to"
],
[
"the",
"and",
"minimum",
"to",
"of",
"that",
"for",
"be",
"in",
"is"
],
[
"oil",
"its",
"yukos",
"auction",
"gazprom",
"bankruptcy",
"russia",
"russian",
"khodorkovsky",
"putin"
],
[
"the",
"suspects",
"to",
"rights",
"human",
"trial",
"of",
"said",
"lords",
"threat"
],
[
"injury",
"madrid",
"look",
"arsenal",
"mourinho",
"boss",
"wilkinson",
"manchester",
"broken",
"season"
],
[
"mobile",
"operators",
"more",
"report",
"mobiles",
"3g",
"handsets",
"digital",
"speeds",
"phones"
],
[
"traffic",
"sites",
"attacks",
"said",
"attack",
"machines",
"that",
"websites",
"site",
"data"
],
[
"star",
"her",
"school",
"woman",
"show",
"singer",
"band",
"released",
"book",
"concert"
],
[
"manufacturing",
"food",
"indian",
"india",
"factory",
"organisation",
"grow",
"plant",
"building",
"agriculture"
],
[
"marketing",
"commissioner",
"they",
"party",
"voting",
"to",
"and",
"the",
"numbers",
"telephone"
],
[
"bid",
"board",
"offer",
"deal",
"may",
"talks",
"takeover",
"exchange",
"group",
"compensation"
],
[
"gardener",
"in",
"race",
"her",
"title",
"60m",
"olympic",
"jump",
"was",
"final"
],
[
"edinburgh",
"scotland",
"squad",
"italy",
"capt",
"glasgow",
"murray",
"rugby",
"wasps",
"wales"
],
[
"female",
"male",
"awards",
"la",
"french",
"prize",
"feature",
"hosted",
"golden",
"u2"
],
[
"shares",
"profits",
"profit",
"stock",
"euros",
"sec",
"company",
"executive",
"exchange",
"insurance"
],
[
"and",
"dvds",
"to",
"content",
"high",
"will",
"dvd",
"on",
"definition",
"programmes"
],
[
"choice",
"union",
"prescott",
"sector",
"pension",
"correspondent",
"age",
"unions",
"strikes",
"workers"
],
[
"statement",
"advice",
"apology",
"lord",
"judge",
"said",
"inquiry",
"would",
"mr",
"attorney"
],
[
"charged",
"drugs",
"denied",
"conservative",
"elections",
"doping",
"stand",
"issues",
"mp",
"commons"
],
[
"didn",
"olympic",
"indoor",
"000m",
"seed",
"final",
"champion",
"world",
"open",
"gold"
],
[
"broadband",
"bt",
"uk",
"internet",
"service",
"via",
"security",
"found",
"spam",
"users"
],
[
"elections",
"eu",
"airlines",
"older",
"agency",
"development",
"poverty",
"pensions",
"fuel",
"payments"
],
[
"interest",
"us",
"rate",
"economists",
"slowdown",
"prices",
"rates",
"quarter",
"rises",
"growth"
],
[
"author",
"appeared",
"1999",
"killed",
"episode",
"norway",
"begin",
"eye",
"queen",
"sex"
],
[
"federal",
"bush",
"bank",
"budget",
"dollar",
"construction",
"economic",
"south",
"korea",
"economy"
],
[
"commit",
"targets",
"selected",
"gmt",
"blow",
"cross",
"contact",
"suffer",
"arrested",
"extended"
],
[
"actress",
"nominated",
"aviator",
"award",
"film",
"awards",
"best",
"night",
"for",
"category"
],
[
"spending",
"35bn",
"cuts",
"chancellor",
"would",
"tax",
"brown",
"stability",
"budget",
"taxes"
],
[
"linux",
"ibm",
"virus",
"mail",
"handheld",
"sony",
"psp",
"playstation",
"nintendo",
"stuff"
],
[
"their",
"ireland",
"six",
"penalty",
"wales",
"france",
"half",
"gara",
"england",
"nations"
],
[
"the",
"xbox",
"halo",
"consoles",
"game",
"nintendo",
"titles",
"games",
"gaming",
"of"
],
[
"album",
"office",
"1m",
"box",
"number",
"took",
"top",
"chart",
"5m",
"4m"
]
] | 383.932157 | average_word_embeddings_glove.6B.300d | 0.8825 | -0.043352 | 0.165744 | 0.860761 |
42 | BBC News | CombinedTM | 50 | [
[
"on",
"definition",
"dvd",
"ray",
"games",
"dvds",
"ces",
"high",
"images",
"toshiba"
],
[
"nintendo",
"games",
"sony",
"handheld",
"console",
"psp",
"gaming",
"playstation",
"screens",
"titles"
],
[
"cross",
"17",
"championships",
"race",
"gb",
"gold",
"000m",
"women",
"beginning",
"bronze"
],
[
"story",
"book",
"oscar",
"harry",
"actor",
"festival",
"musical",
"role",
"potter",
"queen"
],
[
"rights",
"human",
"hunt",
"hunting",
"ban",
"asylum",
"offence",
"not",
"groups",
"court"
],
[
"energy",
"gives",
"adults",
"1999",
"giant",
"pension",
"communications",
"significantly",
"fine",
"friend"
],
[
"qualifying",
"jump",
"of",
"to",
"title",
"the",
"mark",
"in",
"year",
"gardener"
],
[
"interest",
"rates",
"fed",
"prices",
"manufacturing",
"rises",
"economist",
"reserve",
"economists",
"bank"
],
[
"tax",
"cuts",
"budget",
"howard",
"taxes",
"election",
"spending",
"stability",
"35bn",
"mr"
],
[
"test",
"hospital",
"banned",
"conservative",
"manager",
"respect",
"decision",
"elections",
"mp",
"committee"
],
[
"payments",
"winter",
"killed",
"receive",
"special",
"swap",
"prison",
"fox",
"criticised",
"spend"
],
[
"virus",
"source",
"software",
"programs",
"linux",
"mails",
"mail",
"spam",
"ibm",
"windows"
],
[
"he",
"truth",
"his",
"comments",
"was",
"silk",
"robert",
"european",
"party",
"independence"
],
[
"studio",
"recorded",
"host",
"star",
"producer",
"concert",
"band",
"show",
"singer",
"stars"
],
[
"sky",
"awards",
"tipped",
"baby",
"album",
"aviator",
"award",
"favourite",
"categories",
"prize"
],
[
"term",
"stone",
"black",
"is",
"soul",
"music",
"that",
"of",
"cities",
"it"
],
[
"search",
"it",
"you",
"your",
"read",
"project",
"mini",
"school",
"power",
"google"
],
[
"open",
"federer",
"champion",
"my",
"australian",
"final",
"world",
"olympic",
"holmes",
"round"
],
[
"minutes",
"from",
"shot",
"celtic",
"ball",
"after",
"minute",
"post",
"goal",
"lead"
],
[
"rules",
"centres",
"councils",
"sites",
"traffic",
"airlines",
"communities",
"attacks",
"countries",
"data"
],
[
"calls",
"voice",
"users",
"service",
"bt",
"broadband",
"search",
"internet",
"sign",
"net"
],
[
"supply",
"social",
"chinese",
"china",
"reports",
"companies",
"broadband",
"oil",
"responsibility",
"corporate"
],
[
"law",
"said",
"prison",
"suspects",
"lords",
"lord",
"ruling",
"clarke",
"human",
"case"
],
[
"modern",
"the",
"gadgets",
"gadget",
"of",
"premium",
"magazine",
"list",
"100",
"be"
],
[
"would",
"police",
"id",
"plans",
"cards",
"said",
"howard",
"shadow",
"cabinet",
"davis"
],
[
"half",
"wasps",
"we",
"robinson",
"jonny",
"game",
"wilkinson",
"lewsey",
"got",
"try"
],
[
"government",
"workers",
"choice",
"pension",
"age",
"sector",
"prescott",
"union",
"unions",
"strikes"
],
[
"west",
"leaders",
"premiership",
"drawn",
"united",
"merger",
"everton",
"behaviour",
"newcastle",
"visit"
],
[
"growth",
"dollar",
"figures",
"trade",
"2004",
"economy",
"unemployment",
"exports",
"china",
"prices"
],
[
"will",
"be",
"of",
"account",
"budget",
"argentina",
"savings",
"fund",
"is",
"economic"
],
[
"commission",
"sec",
"exchange",
"company",
"financial",
"takeover",
"board",
"firm",
"deal",
"shareholders"
],
[
"broadband",
"uk",
"tv",
"text",
"digital",
"billion",
"total",
"records",
"mobile",
"26"
],
[
"club",
"chelsea",
"team",
"arsenal",
"he",
"me",
"smith",
"manager",
"scottish",
"told"
],
[
"france",
"england",
"thomas",
"cardiff",
"wales",
"half",
"robinson",
"ireland",
"nations",
"six"
],
[
"stop",
"they",
"the",
"to",
"labour",
"argument",
"their",
"party",
"voters",
"of"
],
[
"the",
"is",
"we",
"to",
"that",
"and",
"of",
"be",
"increase",
"minimum"
],
[
"unlikely",
"scotland",
"edinburgh",
"wales",
"rugby",
"italy",
"injury",
"squad",
"against",
"cup"
],
[
"woodward",
"liverpool",
"davis",
"lions",
"new",
"tim",
"cup",
"tennis",
"in",
"henman"
],
[
"who",
"for",
"was",
"best",
"lee",
"film",
"man",
"also",
"night",
"nominated"
],
[
"yukos",
"its",
"russian",
"bankruptcy",
"gazprom",
"court",
"russia",
"khodorkovsky",
"auction",
"oil"
],
[
"com",
"spain",
"language",
"radio",
"spanish",
"rap",
"says",
"hip",
"hop",
"and"
],
[
"office",
"number",
"box",
"took",
"chart",
"starring",
"weekend",
"5m",
"7m",
"comedy"
],
[
"said",
"networks",
"version",
"sharing",
"files",
"file",
"capacity",
"peer",
"that",
"systems"
],
[
"20",
"trade",
"exports",
"price",
"shares",
"strong",
"grew",
"raise",
"trading",
"recovery"
],
[
"arrested",
"send",
"elections",
"begin",
"immigration",
"denies",
"commit",
"commons",
"issues",
"banks"
],
[
"kennedy",
"labour",
"blair",
"politics",
"lib",
"election",
"said",
"mr",
"britain",
"fox"
],
[
"bid",
"its",
"deutsche",
"talks",
"offer",
"london",
"takeover",
"stock",
"exchange",
"deal"
],
[
"cabinet",
"minister",
"he",
"prime",
"allies",
"book",
"mr",
"brown",
"blair",
"poster"
],
[
"achieved",
"feels",
"banned",
"struggled",
"happened",
"blame",
"seemed",
"drugs",
"midfielder",
"test"
],
[
"saw",
"rose",
"profits",
"profit",
"sales",
"car",
"western",
"fell",
"vehicles",
"strong"
]
] | 419.089207 | average_word_embeddings_glove.6B.300d | 0.872 | -0.037199 | 0.151497 | 0.865992 |
42 | BBC News | ZeroShotTM | 10 | [
[
"police",
"human",
"lord",
"said",
"had",
"court",
"was",
"mr",
"would",
"rights"
],
[
"after",
"france",
"his",
"wales",
"ireland",
"two",
"six",
"nations",
"half",
"captain"
],
[
"box",
"series",
"movie",
"award",
"awards",
"star",
"director",
"actor",
"actress",
"category"
],
[
"mr",
"prime",
"blair",
"brown",
"chancellor",
"election",
"howard",
"labour",
"tax",
"tory"
],
[
"sure",
"open",
"liverpool",
"united",
"west",
"chelsea",
"manager",
"club",
"soon",
"old"
],
[
"it",
"is",
"and",
"the",
"of",
"to",
"that",
"in",
"for",
"you"
],
[
"mobile",
"people",
"services",
"networks",
"broadband",
"users",
"more",
"technology",
"are",
"service"
],
[
"programs",
"drugs",
"buy",
"anti",
"company",
"mail",
"mails",
"virus",
"send",
"windows"
],
[
"bank",
"growth",
"market",
"year",
"economic",
"sales",
"december",
"economy",
"rise",
"demand"
],
[
"he",
"of",
"minimum",
"in",
"the",
"that",
"is",
"and",
"to",
"mr"
]
] | 411.967711 | average_word_embeddings_glove.6B.300d | 0.91 | 0.011502 | 0.193145 | 0.788792 |
42 | BBC News | ZeroShotTM | 20 | [
[
"lord",
"lords",
"said",
"law",
"blunkett",
"he",
"secretary",
"mr",
"would",
"hunting"
],
[
"net",
"spam",
"users",
"service",
"security",
"broadband",
"search",
"site",
"sites",
"bt"
],
[
"olympic",
"christmas",
"championships",
"gold",
"holder",
"indoor",
"medal",
"race",
"jump",
"silver"
],
[
"office",
"single",
"chart",
"release",
"box",
"charts",
"sequel",
"series",
"track",
"ten"
],
[
"2004",
"euros",
"profit",
"profits",
"shares",
"strong",
"annual",
"rose",
"investment",
"figures"
],
[
"decision",
"elections",
"party",
"campaign",
"transport",
"unions",
"general",
"drugs",
"doping",
"spokesman"
],
[
"award",
"aviator",
"won",
"actor",
"best",
"awards",
"category",
"favourite",
"named",
"baby"
],
[
"woodward",
"six",
"lions",
"ireland",
"wales",
"nations",
"england",
"france",
"coach",
"captain"
],
[
"will",
"more",
"games",
"sony",
"nintendo",
"gaming",
"generation",
"mobile",
"mobiles",
"tv"
],
[
"match",
"federer",
"open",
"victory",
"seed",
"win",
"beat",
"australian",
"final",
"set"
],
[
"the",
"of",
"song",
"is",
"in",
"to",
"that",
"and",
"it",
"for"
],
[
"elvis",
"radio",
"include",
"stations",
"singer",
"single",
"ring",
"sold",
"artists",
"blue"
],
[
"after",
"title",
"mark",
"minutes",
"seconds",
"penalty",
"final",
"ireland",
"half",
"second"
],
[
"pensions",
"send",
"virus",
"mail",
"social",
"software",
"programs",
"itself",
"either",
"pension"
],
[
"rates",
"economic",
"us",
"economy",
"bush",
"oil",
"deficit",
"dollar",
"growth",
"president"
],
[
"would",
"are",
"be",
"tax",
"to",
"pay",
"said",
"government",
"councils",
"local"
],
[
"to",
"is",
"the",
"of",
"and",
"it",
"you",
"that",
"your",
"search"
],
[
"howard",
"mr",
"brown",
"party",
"prime",
"election",
"he",
"labour",
"blair",
"voters"
],
[
"united",
"manager",
"club",
"football",
"newcastle",
"arsenal",
"boss",
"liverpool",
"injury",
"chelsea"
],
[
"company",
"biggest",
"shareholders",
"figures",
"deutsche",
"its",
"december",
"sales",
"market",
"exchange"
]
] | 313.231972 | average_word_embeddings_glove.6B.300d | 0.91 | 0.005815 | 0.179557 | 0.864027 |
42 | BBC News | ZeroShotTM | 30 | [
[
"mr",
"silk",
"he",
"him",
"party",
"his",
"election",
"was",
"labour",
"voters"
],
[
"judge",
"law",
"was",
"lord",
"cards",
"id",
"said",
"trial",
"advice",
"would"
],
[
"indoor",
"race",
"world",
"olympic",
"championships",
"jump",
"cross",
"gold",
"60m",
"000m"
],
[
"comedy",
"sideways",
"favourite",
"actor",
"including",
"star",
"aviator",
"hill",
"bet",
"chart"
],
[
"seed",
"west",
"united",
"draw",
"everton",
"round",
"face",
"beat",
"saturday",
"open"
],
[
"exports",
"strong",
"figures",
"growth",
"2004",
"rose",
"overall",
"year",
"economic",
"previous"
],
[
"profits",
"banks",
"poor",
"low",
"sell",
"currency",
"indian",
"debt",
"shares",
"investors"
],
[
"spam",
"net",
"online",
"websites",
"site",
"sites",
"service",
"peer",
"internet",
"found"
],
[
"are",
"to",
"tax",
"tory",
"be",
"would",
"they",
"taxes",
"of",
"is"
],
[
"best",
"ceremony",
"film",
"award",
"director",
"awards",
"won",
"aviator",
"category",
"oscars"
],
[
"gazprom",
"russian",
"bankruptcy",
"oil",
"yukos",
"putin",
"us",
"russia",
"its",
"auction"
],
[
"prince",
"ban",
"death",
"arrested",
"drugs",
"television",
"charged",
"killed",
"fine",
"dutch"
],
[
"company",
"deal",
"bid",
"group",
"offer",
"deutsche",
"board",
"shareholders",
"talks",
"exchange"
],
[
"madrid",
"play",
"we",
"real",
"champions",
"arsenal",
"england",
"club",
"barcelona",
"robinson"
],
[
"commons",
"elections",
"unions",
"general",
"test",
"appointed",
"1997",
"debate",
"merger",
"expert"
],
[
"would",
"local",
"workers",
"plans",
"public",
"budget",
"cuts",
"pension",
"government",
"choice"
],
[
"generation",
"will",
"high",
"definition",
"tv",
"which",
"mobile",
"digital",
"be",
"capacity"
],
[
"hunting",
"law",
"said",
"rules",
"courts",
"cases",
"new",
"bill",
"children",
"rights"
],
[
"nations",
"lions",
"england",
"six",
"zealand",
"in",
"woodward",
"wales",
"ireland",
"coach"
],
[
"club",
"situation",
"football",
"team",
"manager",
"sport",
"rugby",
"my",
"players",
"assistant"
],
[
"and",
"apple",
"gadgets",
"of",
"the",
"it",
"is",
"magazine",
"gadget",
"mini"
],
[
"console",
"psp",
"microsoft",
"nintendo",
"handheld",
"sony",
"launch",
"windows",
"available",
"uk"
],
[
"the",
"to",
"of",
"and",
"is",
"that",
"minimum",
"in",
"it",
"be"
],
[
"minister",
"brown",
"mr",
"blair",
"he",
"prime",
"his",
"book",
"stand",
"had"
],
[
"half",
"from",
"lead",
"minutes",
"two",
"second",
"after",
"break",
"ball",
"penalty"
],
[
"security",
"mail",
"software",
"mails",
"programs",
"program",
"anti",
"itself",
"california",
"virus"
],
[
"is",
"to",
"network",
"that",
"and",
"systems",
"you",
"it",
"search",
"of"
],
[
"band",
"box",
"chart",
"christmas",
"number",
"single",
"released",
"release",
"office",
"album"
],
[
"his",
"it",
"best",
"and",
"was",
"song",
"the",
"of",
"that",
"he"
],
[
"dollar",
"rates",
"bank",
"us",
"manufacturing",
"hold",
"interest",
"rate",
"economy",
"output"
]
] | 291.856539 | average_word_embeddings_glove.6B.300d | 0.87 | 0.010478 | 0.176528 | 0.845005 |
42 | BBC News | ZeroShotTM | 40 | [
[
"in",
"their",
"england",
"zealand",
"nations",
"six",
"wales",
"callaghan",
"lions",
"cardiff"
],
[
"arsenal",
"chelsea",
"liverpool",
"league",
"ferguson",
"manchester",
"real",
"madrid",
"united",
"newcastle"
],
[
"sony",
"will",
"mobiles",
"uk",
"digital",
"tv",
"more",
"mobile",
"million",
"phones"
],
[
"england",
"six",
"nations",
"bath",
"club",
"players",
"wales",
"robinson",
"lions",
"squad"
],
[
"police",
"children",
"said",
"asylum",
"would",
"they",
"be",
"cases",
"immigration",
"child"
],
[
"final",
"set",
"serve",
"australian",
"roddick",
"open",
"henman",
"beat",
"victory",
"match"
],
[
"strong",
"profits",
"car",
"sales",
"grow",
"demand",
"europe",
"rose",
"rise",
"airline"
],
[
"russia",
"bankruptcy",
"yukos",
"gazprom",
"russian",
"oil",
"putin",
"gas",
"auction",
"khodorkovsky"
],
[
"of",
"is",
"are",
"to",
"search",
"that",
"said",
"and",
"attacks",
"can"
],
[
"us",
"dollar",
"budget",
"bush",
"economic",
"economy",
"deficit",
"euro",
"countries",
"eu"
],
[
"in",
"25",
"song",
"robbie",
"was",
"best",
"the",
"songs",
"of",
"but"
],
[
"dvd",
"studios",
"dvds",
"games",
"high",
"format",
"definition",
"is",
"more",
"and"
],
[
"drugs",
"doping",
"olympics",
"tests",
"decision",
"missing",
"greek",
"athens",
"charges",
"charged"
],
[
"and",
"the",
"is",
"to",
"for",
"will",
"in",
"minimum",
"be",
"that"
],
[
"race",
"holmes",
"indoor",
"olympic",
"gold",
"francis",
"world",
"championships",
"birmingham",
"cross"
],
[
"prize",
"film",
"bbc",
"director",
"story",
"show",
"sky",
"cast",
"oscar",
"awards"
],
[
"3g",
"speeds",
"mobile",
"and",
"with",
"network",
"networks",
"to",
"instead",
"operators"
],
[
"chambers",
"borders",
"edinburgh",
"unions",
"strategy",
"seats",
"friends",
"deliver",
"debate",
"pensions"
],
[
"government",
"said",
"military",
"was",
"advice",
"lord",
"judge",
"attorney",
"lords",
"inquiry"
],
[
"they",
"lords",
"said",
"not",
"rights",
"law",
"hunting",
"ruling",
"human",
"court"
],
[
"chart",
"single",
"band",
"album",
"number",
"box",
"christmas",
"took",
"released",
"top"
],
[
"we",
"wales",
"two",
"england",
"france",
"try",
"wing",
"injury",
"game",
"half"
],
[
"security",
"calls",
"programs",
"software",
"virus",
"windows",
"broadband",
"bt",
"microsoft",
"users"
],
[
"talks",
"offer",
"london",
"exchange",
"bid",
"deutsche",
"deal",
"shareholders",
"its",
"stock"
],
[
"labour",
"mr",
"election",
"bbc",
"tony",
"minister",
"blair",
"prime",
"he",
"brown"
],
[
"mail",
"pensions",
"companies",
"mails",
"research",
"pension",
"age",
"social",
"spam",
"science"
],
[
"stations",
"killed",
"death",
"show",
"died",
"shows",
"broadcast",
"star",
"victims",
"queen"
],
[
"budget",
"waste",
"taxes",
"tax",
"labour",
"cuts",
"spending",
"election",
"mr",
"chancellor"
],
[
"traffic",
"spam",
"file",
"sharing",
"site",
"data",
"sites",
"attacks",
"websites",
"spammers"
],
[
"film",
"aviator",
"nominated",
"best",
"award",
"actor",
"awards",
"actress",
"nominations",
"who"
],
[
"homes",
"government",
"local",
"council",
"would",
"public",
"councils",
"prescott",
"000",
"sector"
],
[
"of",
"and",
"the",
"to",
"silk",
"he",
"we",
"is",
"liverpool",
"new"
],
[
"is",
"and",
"gadget",
"of",
"it",
"the",
"gadgets",
"as",
"magazine",
"list"
],
[
"month",
"us",
"rate",
"december",
"bank",
"rates",
"economy",
"retailers",
"retail",
"interest"
],
[
"election",
"as",
"silk",
"was",
"he",
"mr",
"labour",
"his",
"party",
"of"
],
[
"company",
"iraq",
"recently",
"ministry",
"fuel",
"airlines",
"executive",
"insurance",
"drug",
"financial"
],
[
"to",
"and",
"of",
"the",
"is",
"are",
"that",
"it",
"be",
"in"
],
[
"shot",
"after",
"goal",
"ball",
"from",
"minutes",
"half",
"lead",
"scored",
"post"
],
[
"he",
"mr",
"brown",
"minister",
"apologise",
"prime",
"book",
"told",
"jewish",
"blair"
],
[
"january",
"quarter",
"2004",
"figures",
"strong",
"sales",
"rose",
"annual",
"growth",
"profits"
]
] | 373.056609 | average_word_embeddings_glove.6B.300d | 0.785 | 0.018343 | 0.173013 | 0.826066 |
42 | BBC News | ZeroShotTM | 50 | [
[
"sell",
"vehicles",
"company",
"deal",
"shares",
"profits",
"indian",
"buy",
"euros",
"share"
],
[
"said",
"re",
"would",
"have",
"of",
"not",
"countries",
"be",
"powers",
"human"
],
[
"actress",
"comedy",
"category",
"actor",
"best",
"awards",
"film",
"award",
"won",
"director"
],
[
"court",
"gazprom",
"yukos",
"bankruptcy",
"russia",
"russian",
"company",
"auction",
"its",
"law"
],
[
"to",
"of",
"the",
"new",
"government",
"that",
"police",
"he",
"on",
"by"
],
[
"title",
"olympic",
"final",
"britain",
"gardener",
"60m",
"seconds",
"relay",
"women",
"indoor"
],
[
"for",
"her",
"she",
"was",
"in",
"who",
"lee",
"best",
"sharp",
"at"
],
[
"we",
"in",
"and",
"the",
"of",
"to",
"minimum",
"will",
"increase",
"is"
],
[
"spending",
"bush",
"cuts",
"taxes",
"budget",
"tax",
"would",
"proposals",
"deficit",
"congress"
],
[
"playstation",
"handheld",
"sony",
"psp",
"nintendo",
"console",
"xbox",
"halo",
"launch",
"san"
],
[
"its",
"oil",
"yukos",
"russian",
"deutsche",
"bid",
"bankruptcy",
"london",
"khodorkovsky",
"deal"
],
[
"show",
"confirmed",
"tour",
"broadcast",
"scottish",
"singer",
"green",
"include",
"black",
"fight"
],
[
"elvis",
"box",
"album",
"places",
"single",
"chart",
"number",
"band",
"including",
"top"
],
[
"election",
"blair",
"labour",
"mr",
"campaigning",
"was",
"howard",
"he",
"party",
"said"
],
[
"age",
"label",
"children",
"mtv",
"living",
"income",
"pension",
"students",
"opinions",
"35"
],
[
"drugs",
"tests",
"athens",
"suspended",
"doping",
"charges",
"greek",
"test",
"decision",
"guilty"
],
[
"smith",
"job",
"players",
"referee",
"football",
"club",
"rugby",
"lions",
"manager",
"england"
],
[
"six",
"nations",
"zealand",
"lions",
"france",
"england",
"woodward",
"wales",
"saturday",
"white"
],
[
"eu",
"workers",
"foreign",
"government",
"savings",
"local",
"tax",
"council",
"straw",
"economic"
],
[
"rises",
"inflation",
"economists",
"bank",
"policy",
"interest",
"rates",
"economy",
"fed",
"us"
],
[
"site",
"net",
"attacks",
"sites",
"traffic",
"said",
"attack",
"are",
"criminals",
"target"
],
[
"figures",
"strong",
"overall",
"2004",
"economic",
"rose",
"growth",
"dollar",
"mortgage",
"lending"
],
[
"and",
"it",
"mouse",
"the",
"of",
"to",
"screen",
"children",
"that",
"mac"
],
[
"unions",
"members",
"merger",
"society",
"schools",
"super",
"age",
"social",
"transport",
"union"
],
[
"jewish",
"mr",
"he",
"muslim",
"prime",
"minister",
"blair",
"mp",
"howard",
"told"
],
[
"said",
"they",
"not",
"are",
"that",
"is",
"of",
"online",
"to",
"sites"
],
[
"set",
"seed",
"match",
"open",
"beat",
"serve",
"australian",
"matches",
"federer",
"round"
],
[
"in",
"was",
"he",
"have",
"and",
"but",
"his",
"the",
"as",
"to"
],
[
"trade",
"2004",
"sales",
"year",
"growth",
"strong",
"rose",
"economic",
"figures",
"value"
],
[
"that",
"it",
"3g",
"to",
"thinks",
"is",
"and",
"will",
"systems",
"trusted"
],
[
"dvd",
"high",
"and",
"definition",
"dvds",
"games",
"wars",
"studios",
"time",
"are"
],
[
"language",
"spanish",
"and",
"hip",
"radio",
"the",
"rap",
"hop",
"to",
"is"
],
[
"rose",
"exports",
"figures",
"2004",
"january",
"rate",
"growth",
"overall",
"rise",
"manufacturing"
],
[
"mr",
"brown",
"he",
"blair",
"tax",
"labour",
"spending",
"taxes",
"election",
"manifesto"
],
[
"mr",
"party",
"election",
"brown",
"elections",
"campaign",
"labour",
"minister",
"leader",
"howard"
],
[
"screens",
"show",
"ces",
"will",
"new",
"gamers",
"industry",
"which",
"devices",
"gaming"
],
[
"china",
"oil",
"russian",
"its",
"yukos",
"london",
"deutsche",
"us",
"gazprom",
"russia"
],
[
"sites",
"said",
"were",
"they",
"people",
"online",
"file",
"sharing",
"using",
"research"
],
[
"would",
"on",
"be",
"law",
"said",
"local",
"police",
"plans",
"ban",
"new"
],
[
"band",
"chart",
"number",
"single",
"box",
"top",
"release",
"office",
"charity",
"director"
],
[
"lord",
"should",
"inquiry",
"murder",
"law",
"judges",
"justice",
"human",
"said",
"judge"
],
[
"software",
"versions",
"users",
"viruses",
"programs",
"security",
"linux",
"google",
"windows",
"microsoft"
],
[
"broadband",
"mobile",
"uk",
"million",
"bt",
"tv",
"phone",
"operators",
"mobiles",
"services"
],
[
"liverpool",
"arsenal",
"chelsea",
"real",
"madrid",
"club",
"league",
"barcelona",
"manager",
"champions"
],
[
"play",
"boss",
"arsenal",
"champions",
"mourinho",
"liverpool",
"league",
"jose",
"real",
"chelsea"
],
[
"000m",
"indoor",
"gold",
"race",
"world",
"triple",
"norwich",
"sheffield",
"here",
"prix"
],
[
"25",
"the",
"that",
"song",
"music",
"is",
"it",
"robbie",
"of",
"years"
],
[
"minutes",
"ball",
"half",
"after",
"ireland",
"from",
"lead",
"gara",
"first",
"penalty"
],
[
"represent",
"votes",
"unions",
"strategy",
"meetings",
"campaign",
"agenda",
"elections",
"call",
"meant"
],
[
"anti",
"virus",
"mails",
"mail",
"itself",
"infected",
"social",
"program",
"message",
"messages"
]
] | 310.730993 | average_word_embeddings_glove.6B.300d | 0.724 | 0.023967 | 0.173695 | 0.833458 |
42 | ArXiv ML Papers | GMM | 10 | [
[
"representation",
"space",
"quantum",
"embedding",
"graph",
"function",
"neural",
"networks",
"node",
"graphs"
],
[
"were",
"dataset",
"detection",
"images",
"cancer",
"prediction",
"compared",
"medical",
"patients",
"segmentation"
],
[
"bounds",
"bound",
"regret",
"sample",
"convergence",
"variance",
"error",
"probability",
"convex",
"optimal"
],
[
"adversarial",
"agent",
"attacks",
"human",
"research",
"policy",
"their",
"agents",
"reinforcement",
"knowledge"
],
[
"network",
"accuracy",
"devices",
"traffic",
"deep",
"networks",
"neural",
"design",
"machine",
"techniques"
],
[
"methods",
"prediction",
"approach",
"method",
"classification",
"algorithms",
"tasks",
"uncertainty",
"approaches",
"proposed"
],
[
"optimization",
"stochastic",
"bayesian",
"variational",
"gradient",
"posterior",
"algorithm",
"approximation",
"inference",
"convergence"
],
[
"adversarial",
"speech",
"end",
"training",
"datasets",
"propose",
"gan",
"neural",
"generative",
"convolutional"
],
[
"text",
"visual",
"image",
"language",
"representations",
"features",
"word",
"images",
"semantic",
"natural"
],
[
"tensor",
"graphs",
"graph",
"matrix",
"rank",
"nodes",
"decomposition",
"gnns",
"spectral",
"laplacian"
]
] | 1.53068 | average_word_embeddings_glove.6B.300d | 0.91 | 0.011302 | 0.162483 | 0.795347 |
42 | ArXiv ML Papers | GMM | 20 | [
[
"loss",
"activation",
"uncertainty",
"functions",
"relu",
"network",
"networks",
"neurons",
"neural",
"coding"
],
[
"cell",
"brain",
"cancer",
"disease",
"patients",
"were",
"ct",
"healthy",
"patient",
"images"
],
[
"function",
"probability",
"bound",
"variance",
"convergence",
"bounds",
"sample",
"varepsilon",
"regression",
"distribution"
],
[
"social",
"ml",
"research",
"privacy",
"questions",
"software",
"information",
"recommendation",
"challenges",
"users"
],
[
"user",
"service",
"machine",
"design",
"devices",
"mobile",
"communication",
"techniques",
"distributed",
"systems"
],
[
"driving",
"performance",
"accuracy",
"ensemble",
"deep",
"training",
"search",
"test",
"art",
"compared"
],
[
"problems",
"linear",
"algorithm",
"quantum",
"optimization",
"algorithms",
"pooling",
"clustering",
"function",
"problem"
],
[
"generative",
"classifier",
"latent",
"adversarial",
"datasets",
"interpretable",
"perturbations",
"examples",
"robustness",
"trained"
],
[
"video",
"object",
"image",
"images",
"audio",
"scene",
"features",
"objects",
"feature",
"visual"
],
[
"stochastic",
"gradient",
"variational",
"posterior",
"optimization",
"inference",
"bayesian",
"order",
"approximation",
"gaussian"
],
[
"metric",
"quantum",
"representation",
"representations",
"relational",
"classical",
"causal",
"embedding",
"feature",
"transformation"
],
[
"method",
"datasets",
"methods",
"selection",
"samples",
"class",
"classification",
"sample",
"metric",
"prediction"
],
[
"graph",
"manifold",
"tensor",
"rank",
"decomposition",
"topological",
"matrix",
"low",
"laplacian",
"subspaces"
],
[
"gnns",
"nodes",
"graph",
"graphs",
"node",
"link",
"embeddings",
"embedding",
"representation",
"vertex"
],
[
"power",
"prediction",
"demand",
"forecasts",
"forecasting",
"traffic",
"weather",
"forecast",
"time",
"energy"
],
[
"word",
"languages",
"representations",
"language",
"visual",
"text",
"semantic",
"words",
"natural",
"task"
],
[
"action",
"policy",
"rl",
"reward",
"policies",
"fairness",
"reinforcement",
"exploration",
"fair",
"optimal"
],
[
"tasks",
"attacks",
"agents",
"agent",
"human",
"reinforcement",
"adversarial",
"policy",
"robot",
"environment"
],
[
"segmentation",
"speech",
"accuracy",
"convolutional",
"image",
"cnn",
"end",
"layers",
"convolution",
"feature"
],
[
"regret",
"bounds",
"bound",
"agent",
"sqrt",
"points",
"agents",
"day",
"online",
"algorithm"
]
] | 2.288565 | average_word_embeddings_glove.6B.300d | 0.875 | -0.035698 | 0.161252 | 0.842745 |
42 | ArXiv ML Papers | GMM | 30 | [
[
"loss",
"gradient",
"neural",
"activation",
"functions",
"networks",
"relu",
"bounds",
"network",
"function"
],
[
"diagnosis",
"ct",
"clinical",
"patient",
"health",
"medical",
"patients",
"diseases",
"cancer",
"disease"
],
[
"gradient",
"convergence",
"convex",
"regression",
"variance",
"descent",
"stochastic",
"bounds",
"linear",
"error"
],
[
"social",
"research",
"methods",
"techniques",
"applications",
"challenges",
"health",
"patient",
"clinical",
"analysis"
],
[
"deep",
"nodes",
"structure",
"network",
"networks",
"layer",
"neural",
"neurons",
"architectures",
"flow"
],
[
"speech",
"signals",
"channel",
"enhancement",
"signal",
"source",
"transmission",
"end",
"hardware",
"coding"
],
[
"representation",
"representations",
"embedding",
"metric",
"concepts",
"causal",
"between",
"objects",
"sequences",
"learn"
],
[
"software",
"quality",
"engineering",
"process",
"deep",
"machine",
"been",
"actions",
"user",
"feature"
],
[
"tensor",
"rank",
"kernel",
"algorithms",
"algorithm",
"matrix",
"decomposition",
"convex",
"quantum",
"optimization"
],
[
"against",
"detection",
"adversarial",
"attack",
"defense",
"attacker",
"attacks",
"box",
"examples",
"robustness"
],
[
"optimization",
"tensor",
"order",
"gradient",
"stochastic",
"kernels",
"compositional",
"rank",
"estimator",
"convex"
],
[
"object",
"gan",
"dataset",
"visual",
"segmentation",
"samples",
"images",
"image",
"detection",
"imaging"
],
[
"datasets",
"segmentation",
"text",
"classifier",
"image",
"adversarial",
"transformer",
"speech",
"audio",
"encoder"
],
[
"inference",
"bayesian",
"variational",
"stochastic",
"posterior",
"estimator",
"approximate",
"distribution",
"latent",
"distributions"
],
[
"distribution",
"given",
"bounds",
"bound",
"sample",
"regret",
"lower",
"probability",
"algorithm",
"number"
],
[
"feature",
"sensor",
"was",
"signals",
"temperature",
"protein",
"3d",
"convolutional",
"eeg",
"energy"
],
[
"dynamics",
"nonlinear",
"optimization",
"systems",
"solver",
"linear",
"end",
"efficient",
"quantum",
"problems"
],
[
"graphs",
"nodes",
"graph",
"node",
"gnns",
"clustering",
"spectral",
"embeddings",
"laplacian",
"edges"
],
[
"reinforcement",
"agent",
"policies",
"policy",
"tasks",
"reward",
"human",
"environment",
"robot",
"agents"
],
[
"agents",
"regret",
"game",
"rl",
"players",
"games",
"agent",
"play",
"human",
"group"
],
[
"text",
"generation",
"music",
"image",
"images",
"style",
"video",
"audio",
"gans",
"realistic"
],
[
"user",
"users",
"fairness",
"ml",
"recommendation",
"media",
"systems",
"social",
"information",
"research"
],
[
"forecasting",
"predictions",
"uncertainty",
"predictive",
"prediction",
"models",
"forecasts",
"series",
"forecast",
"generative"
],
[
"12",
"explainable",
"2019",
"interpretable",
"vision",
"person",
"periods",
"personal",
"machine",
"period"
],
[
"accuracy",
"unlabeled",
"weather",
"ensemble",
"forecasting",
"prediction",
"test",
"dataset",
"demand",
"forecast"
],
[
"design",
"search",
"federated",
"cloud",
"distributed",
"hardware",
"devices",
"framework",
"distillation",
"architecture"
],
[
"method",
"algorithms",
"linear",
"algorithm",
"problems",
"quantum",
"methods",
"clustering",
"proposed",
"dimensional"
],
[
"languages",
"word",
"embeddings",
"natural",
"text",
"words",
"representations",
"sentences",
"translation",
"language"
],
[
"system",
"detection",
"vehicles",
"driving",
"vehicle",
"traffic",
"demand",
"time",
"detect",
"speed"
],
[
"scaling",
"training",
"efficient",
"loss",
"day",
"cost",
"rl",
"state",
"pruning",
"overfitting"
]
] | 2.712271 | average_word_embeddings_glove.6B.300d | 0.78 | -0.035137 | 0.170366 | 0.843608 |
42 | ArXiv ML Papers | GMM | 40 | [
[
"sample",
"marginal",
"bayesian",
"out",
"size",
"evidence",
"posterior",
"individual",
"variational",
"family"
],
[
"health",
"medical",
"predict",
"patients",
"patient",
"ct",
"diagnosis",
"disease",
"diseases",
"cancer"
],
[
"convergence",
"descent",
"gradient",
"convex",
"diffusion",
"stochastic",
"variance",
"linear",
"subspace",
"rank"
],
[
"class",
"strategies",
"prior",
"case",
"when",
"bounds",
"control",
"estimate",
"treatment",
"fairness"
],
[
"layer",
"networks",
"architectures",
"deep",
"structure",
"neural",
"network",
"neurons",
"flow",
"loss"
],
[
"hardware",
"accelerators",
"power",
"processing",
"memory",
"source",
"bandwidth",
"convolution",
"end",
"layer"
],
[
"reasoning",
"representations",
"visual",
"causal",
"representation",
"concepts",
"relational",
"neural",
"symbolic",
"continuous"
],
[
"engineering",
"quality",
"machine",
"deep",
"software",
"tests",
"feature",
"recognition",
"process",
"search"
],
[
"convex",
"log",
"matrix",
"algorithms",
"algorithm",
"quantum",
"optimization",
"exchange",
"decomposition",
"linear"
],
[
"attack",
"attacks",
"adversarial",
"against",
"defense",
"detection",
"box",
"attacker",
"security",
"iot"
],
[
"gradient",
"optimization",
"order",
"stochastic",
"convex",
"descent",
"proximal",
"solving",
"constrained",
"variational"
],
[
"images",
"detection",
"segmentation",
"image",
"dataset",
"training",
"samples",
"art",
"augmentation",
"techniques"
],
[
"decoder",
"encoder",
"speech",
"end",
"audio",
"segmentation",
"image",
"classifier",
"speaker",
"eeg"
],
[
"stochastic",
"regression",
"variational",
"inference",
"posterior",
"bayesian",
"approximate",
"distribution",
"uncertainty",
"distributions"
],
[
"sample",
"bounds",
"bound",
"regret",
"distribution",
"lower",
"delta",
"minimax",
"probability",
"optimal"
],
[
"was",
"signals",
"were",
"protein",
"energy",
"whole",
"spectrum",
"brain",
"convolutional",
"power"
],
[
"network",
"neural",
"dynamics",
"nonlinear",
"networks",
"protein",
"systems",
"activation",
"field",
"convolutional"
],
[
"graphs",
"laplacian",
"graph",
"nodes",
"vertices",
"filters",
"spectral",
"gnns",
"vertex",
"geometric"
],
[
"reward",
"actions",
"agents",
"agent",
"environment",
"reinforcement",
"robot",
"policy",
"rl",
"tasks"
],
[
"players",
"agents",
"game",
"games",
"agent",
"play",
"atari",
"environment",
"rl",
"environments"
],
[
"gans",
"frames",
"style",
"audio",
"music",
"generation",
"image",
"video",
"segment",
"generating"
],
[
"privacy",
"recommendation",
"media",
"users",
"research",
"social",
"ml",
"compliance",
"user",
"information"
],
[
"hierarchical",
"forecasting",
"market",
"federated",
"forecast",
"probabilistic",
"uncertainty",
"forecasts",
"prediction",
"relations"
],
[
"12",
"explainable",
"2019",
"interpretable",
"vision",
"person",
"periods",
"personal",
"machine",
"period"
],
[
"stream",
"unlabeled",
"classification",
"ensemble",
"samples",
"big",
"supervised",
"labeled",
"high",
"sets"
],
[
"framework",
"search",
"nas",
"devices",
"design",
"federated",
"server",
"cloud",
"resource",
"user"
],
[
"linear",
"clustering",
"classification",
"problems",
"algorithms",
"label",
"methods",
"method",
"algorithm",
"pooling"
],
[
"translation",
"sentences",
"natural",
"languages",
"text",
"language",
"words",
"word",
"embeddings",
"context"
],
[
"wireless",
"channel",
"driving",
"modes",
"demand",
"system",
"vehicle",
"speed",
"traffic",
"vehicles"
],
[
"imagenet",
"cost",
"day",
"scaling",
"pruning",
"training",
"score",
"rate",
"loss",
"state"
],
[
"series",
"models",
"modeling",
"predictive",
"ensemble",
"fidelity",
"generative",
"different",
"prediction",
"cnn"
],
[
"nodes",
"representation",
"graph",
"link",
"graphs",
"node",
"vertex",
"embeddings",
"gnns",
"embedding"
],
[
"measure",
"distance",
"feature",
"metric",
"euclidean",
"embedding",
"space",
"similarity",
"objects",
"dimension"
],
[
"lead",
"financial",
"market",
"approximations",
"media",
"membership",
"mixed",
"price",
"operators",
"credit"
],
[
"forecasts",
"forecast",
"demand",
"year",
"weather",
"forecasting",
"time",
"prediction",
"patterns",
"scale"
],
[
"scene",
"representations",
"image",
"text",
"objects",
"images",
"features",
"visual",
"object",
"human"
],
[
"biases",
"adversarial",
"fair",
"fairness",
"dataset",
"gender",
"bias",
"overfitting",
"confidence",
"groups"
],
[
"quantum",
"policies",
"optimal",
"stochastic",
"policy",
"problem",
"problems",
"inference",
"solving",
"optimization"
],
[
"machine",
"human",
"research",
"scientific",
"techniques",
"challenges",
"quantum",
"social",
"review",
"design"
],
[
"tensor",
"gaussian",
"kernels",
"kernel",
"divergence",
"separable",
"rank",
"convergence",
"fast",
"inducing"
]
] | 3.330639 | average_word_embeddings_glove.6B.300d | 0.775 | -0.067732 | 0.164781 | 0.868519 |
42 | ArXiv ML Papers | GMM | 50 | [
[
"sample",
"marginal",
"bayesian",
"out",
"size",
"evidence",
"posterior",
"individual",
"variational",
"family"
],
[
"diseases",
"was",
"patients",
"ct",
"predict",
"medical",
"disease",
"cancer",
"patient",
"diagnosis"
],
[
"minimax",
"gradient",
"descent",
"convergence",
"convex",
"strongly",
"function",
"stochastic",
"bounds",
"matrix"
],
[
"deep",
"detection",
"active",
"available",
"samples",
"datasets",
"class",
"patient",
"methods",
"supervised"
],
[
"temporal",
"deep",
"architecture",
"neurons",
"network",
"networks",
"architectures",
"neural",
"layer",
"structure"
],
[
"coding",
"digital",
"speech",
"mobile",
"end",
"source",
"hardware",
"channel",
"transmission",
"processing"
],
[
"reasoning",
"representations",
"relational",
"quantum",
"causal",
"concepts",
"representation",
"symbolic",
"processes",
"theory"
],
[
"hyperparameters",
"process",
"pipeline",
"goals",
"losses",
"quality",
"strategy",
"questions",
"expert",
"driving"
],
[
"matrix",
"algorithm",
"convex",
"decomposition",
"algorithms",
"exchange",
"rank",
"kernel",
"matrices",
"approximation"
],
[
"attacks",
"attack",
"adversarial",
"threat",
"detection",
"box",
"iot",
"attacker",
"against",
"defense"
],
[
"gradient",
"optimization",
"order",
"stochastic",
"convex",
"descent",
"proximal",
"solving",
"constrained",
"variational"
],
[
"segmentation",
"images",
"detection",
"reconstruction",
"augmentation",
"image",
"medical",
"dl",
"noise",
"object"
],
[
"image",
"encoder",
"decoder",
"classifier",
"perturbations",
"convolution",
"channel",
"speech",
"eeg",
"end"
],
[
"distributions",
"variational",
"uncertainty",
"inference",
"approximate",
"bayesian",
"regression",
"posterior",
"distribution",
"variables"
],
[
"delta",
"lower",
"distribution",
"frac",
"bound",
"varepsilon",
"probability",
"bounds",
"sample",
"number"
],
[
"were",
"protein",
"convolutional",
"fusion",
"brain",
"detect",
"whole",
"segmentation",
"detection",
"was"
],
[
"gan",
"shape",
"dynamics",
"local",
"quantum",
"nonlinear",
"linear",
"protein",
"theory",
"carlo"
],
[
"spectral",
"laplacian",
"filters",
"graphs",
"graph",
"vertices",
"nodes",
"geometric",
"vertex",
"gnns"
],
[
"reinforcement",
"reward",
"policy",
"control",
"policies",
"safety",
"optimal",
"constraints",
"measurement",
"action"
],
[
"look",
"game",
"simulation",
"environment",
"play",
"players",
"agent",
"human",
"atari",
"games"
],
[
"video",
"generating",
"image",
"style",
"music",
"generator",
"generative",
"generation",
"gans",
"audio"
],
[
"user",
"recommendation",
"media",
"group",
"information",
"view",
"users",
"domain",
"ml",
"privacy"
],
[
"probabilistic",
"series",
"distribution",
"federated",
"forecast",
"calibrated",
"forecasts",
"relations",
"hierarchical",
"forecasting"
],
[
"12",
"explainable",
"2019",
"interpretable",
"vision",
"person",
"periods",
"personal",
"machine",
"period"
],
[
"stream",
"big",
"classification",
"online",
"high",
"database",
"000",
"memory",
"algorithms",
"size"
],
[
"generation",
"search",
"design",
"architecture",
"nas",
"distillation",
"generator",
"trained",
"language",
"bert"
],
[
"clustering",
"algorithms",
"classification",
"pooling",
"methods",
"linear",
"label",
"problems",
"method",
"algorithm"
],
[
"language",
"word",
"text",
"natural",
"translation",
"languages",
"source",
"cross",
"code",
"nlp"
],
[
"prediction",
"modes",
"vehicle",
"traffic",
"service",
"social",
"sharing",
"demand",
"internet",
"driving"
],
[
"pruning",
"scaling",
"day",
"imagenet",
"rate",
"weights",
"imputation",
"costs",
"precision",
"10"
],
[
"cnn",
"generative",
"ensemble",
"predictive",
"models",
"modeling",
"fidelity",
"prediction",
"parameters",
"high"
],
[
"representation",
"nodes",
"node",
"gnns",
"graph",
"graphs",
"vertex",
"link",
"edges",
"embeddings"
],
[
"class",
"metric",
"distance",
"embedding",
"measure",
"space",
"objects",
"dimensional",
"classical",
"classes"
],
[
"mixed",
"credit",
"market",
"financial",
"membership",
"lead",
"operators",
"approximations",
"media",
"social"
],
[
"weather",
"forecasting",
"traffic",
"forecasts",
"forecast",
"series",
"time",
"temporal",
"speed",
"prediction"
],
[
"video",
"feature",
"scene",
"3d",
"features",
"visual",
"attention",
"gan",
"images",
"image"
],
[
"samples",
"uncertainty",
"estimator",
"bias",
"estimation",
"prediction",
"gender",
"adversarial",
"predictions",
"causal"
],
[
"optimization",
"stochastic",
"optimal",
"policy",
"problem",
"games",
"algorithms",
"algorithm",
"inference",
"guarantees"
],
[
"software",
"ml",
"research",
"machine",
"challenges",
"scientific",
"techniques",
"engineering",
"review",
"quantum"
],
[
"convergence",
"fast",
"rank",
"gaussian",
"noise",
"tensor",
"inducing",
"divergence",
"optimization",
"kernel"
],
[
"meta",
"rl",
"regret",
"agents",
"setting",
"agent",
"problem",
"bandit",
"optimal",
"where"
],
[
"shift",
"demand",
"performance",
"energy",
"forecast",
"forecasting",
"computing",
"transformers",
"federated",
"cost"
],
[
"fairness",
"robust",
"rl",
"reward",
"adversarial",
"fair",
"robustness",
"biases",
"policy",
"explanations"
],
[
"distributed",
"communication",
"server",
"devices",
"power",
"hardware",
"computation",
"memory",
"execution",
"cost"
],
[
"gradient",
"layers",
"activation",
"convolutional",
"network",
"networks",
"neural",
"relu",
"layer",
"signals"
],
[
"tracking",
"detection",
"driving",
"scenarios",
"autonomous",
"detect",
"audio",
"vehicle",
"object",
"profiles"
],
[
"words",
"representations",
"semantic",
"visual",
"objects",
"speech",
"word",
"scene",
"style",
"language"
],
[
"documents",
"day",
"multi",
"medical",
"successful",
"was",
"sets",
"text",
"evaluated",
"used"
],
[
"human",
"privacy",
"questions",
"attacks",
"fairness",
"bias",
"influence",
"how",
"use",
"groups"
],
[
"actions",
"agent",
"goal",
"policy",
"reinforcement",
"task",
"tasks",
"robot",
"human",
"agents"
]
] | 3.846268 | average_word_embeddings_glove.6B.300d | 0.728 | -0.081105 | 0.14901 | 0.871859 |
42 | ArXiv ML Papers | KeyNMF | 10 | [
[
"demonstrate",
"work",
"framework",
"tasks",
"learning",
"learn",
"information",
"problems",
"existing",
"performance"
],
[
"applications",
"various",
"standard",
"existing",
"simple",
"methods",
"problems",
"important",
"structure",
"approaches"
],
[
"techniques",
"time",
"using",
"network",
"networks",
"technique",
"results",
"developed",
"input",
"present"
],
[
"models",
"better",
"based",
"parameters",
"modeling",
"performance",
"model",
"design",
"demonstrate",
"language"
],
[
"provide",
"analysis",
"process",
"large",
"available",
"data",
"applications",
"techniques",
"specific",
"information"
],
[
"problems",
"algorithm",
"problem",
"algorithms",
"optimal",
"function",
"linear",
"provide",
"given",
"functions"
],
[
"able",
"study",
"multiple",
"examples",
"various",
"different",
"types",
"specific",
"similar",
"components"
],
[
"proposed",
"efficient",
"simple",
"demonstrate",
"based",
"experiments",
"effective",
"method",
"process",
"analysis"
],
[
"used",
"use",
"time",
"work",
"information",
"study",
"analysis",
"important",
"based",
"provide"
],
[
"systems",
"approach",
"given",
"approaches",
"problem",
"provide",
"specific",
"requires",
"use",
"value"
]
] | 3.895604 | average_word_embeddings_glove.6B.300d | 0.7 | -0.013505 | 0.175113 | 0.673608 |
42 | ArXiv ML Papers | KeyNMF | 20 | [
[
"tasks",
"learn",
"learning",
"demonstrate",
"framework",
"problems",
"knowledge",
"allows",
"environment",
"process"
],
[
"existing",
"methods",
"standard",
"structure",
"important",
"various",
"applications",
"techniques",
"problems",
"experiments"
],
[
"types",
"present",
"technique",
"shown",
"techniques",
"developed",
"using",
"input",
"detection",
"able"
],
[
"models",
"demonstrate",
"design",
"modeling",
"parameters",
"model",
"language",
"distribution",
"existing",
"simple"
],
[
"types",
"large",
"available",
"quality",
"process",
"data",
"possible",
"distribution",
"specific",
"furthermore"
],
[
"given",
"problem",
"problems",
"solve",
"solution",
"consider",
"knowledge",
"model",
"terms",
"result"
],
[
"different",
"multiple",
"similar",
"examples",
"various",
"study",
"components",
"specific",
"able",
"types"
],
[
"demonstrate",
"simple",
"proposed",
"method",
"effective",
"experiments",
"framework",
"process",
"efficient",
"better"
],
[
"based",
"better",
"results",
"performance",
"study",
"experimental",
"recent",
"training",
"improve",
"achieve"
],
[
"approach",
"value",
"processes",
"approaches",
"requires",
"simple",
"given",
"instead",
"necessary",
"standard"
],
[
"important",
"improve",
"image",
"study",
"developed",
"used",
"effective",
"difficult",
"uses",
"quality"
],
[
"use",
"specific",
"example",
"requires",
"uses",
"available",
"allows",
"important",
"make",
"language"
],
[
"better",
"practical",
"example",
"theoretical",
"particular",
"provide",
"functions",
"research",
"results",
"provides"
],
[
"network",
"networks",
"neural",
"image",
"various",
"efficient",
"input",
"large",
"multiple",
"structure"
],
[
"recent",
"improve",
"able",
"human",
"work",
"knowledge",
"available",
"make",
"possible",
"various"
],
[
"features",
"based",
"information",
"existing",
"important",
"user",
"feature",
"order",
"research",
"present"
],
[
"process",
"time",
"large",
"number",
"cost",
"efficient",
"uses",
"applications",
"study",
"current"
],
[
"applications",
"application",
"based",
"analysis",
"theoretical",
"developed",
"study",
"applied",
"social",
"human"
],
[
"techniques",
"systems",
"better",
"various",
"existing",
"research",
"user",
"applications",
"important",
"multiple"
],
[
"optimization",
"function",
"algorithm",
"algorithms",
"optimal",
"functions",
"linear",
"complexity",
"parameters",
"efficient"
]
] | 4.101509 | average_word_embeddings_glove.6B.300d | 0.58 | -0.063003 | 0.176383 | 0.728601 |
42 | ArXiv ML Papers | KeyNMF | 30 | [
[
"learn",
"tasks",
"learning",
"framework",
"process",
"environment",
"demonstrate",
"knowledge",
"training",
"allows"
],
[
"methods",
"standard",
"important",
"structure",
"simple",
"experiments",
"theoretical",
"various",
"suggest",
"available"
],
[
"types",
"object",
"technique",
"shown",
"detection",
"present",
"available",
"using",
"able",
"developed"
],
[
"model",
"demonstrate",
"modeling",
"models",
"parameters",
"design",
"language",
"multiple",
"distribution",
"experiments"
],
[
"distribution",
"possible",
"available",
"quality",
"identify",
"data",
"process",
"furthermore",
"types",
"addition"
],
[
"problem",
"knowledge",
"solution",
"consider",
"model",
"practical",
"terms",
"solve",
"optimal",
"value"
],
[
"different",
"types",
"able",
"specific",
"multiple",
"various",
"similar",
"examples",
"identify",
"components"
],
[
"efficient",
"simple",
"process",
"method",
"effective",
"demonstrate",
"proposed",
"framework",
"experiments",
"accuracy"
],
[
"improve",
"performance",
"results",
"achieve",
"quality",
"tasks",
"significantly",
"training",
"accuracy",
"recent"
],
[
"approaches",
"approach",
"requires",
"necessary",
"processes",
"instead",
"difficult",
"value",
"standard",
"simple"
],
[
"effective",
"used",
"difficult",
"make",
"important",
"improve",
"uses",
"developed",
"quality",
"mean"
],
[
"analysis",
"theoretical",
"standard",
"developed",
"application",
"applied",
"social",
"type",
"component",
"human"
],
[
"results",
"studies",
"study",
"research",
"effective",
"application",
"multiple",
"related",
"various",
"knowledge"
],
[
"structure",
"network",
"neural",
"multiple",
"various",
"input",
"networks",
"image",
"communication",
"standard"
],
[
"work",
"able",
"possible",
"make",
"improve",
"recent",
"human",
"way",
"knowledge",
"available"
],
[
"improve",
"information",
"important",
"specific",
"feature",
"furthermore",
"user",
"research",
"useful",
"order"
],
[
"uses",
"step",
"time",
"possible",
"process",
"making",
"required",
"cost",
"current",
"efficiency"
],
[
"practical",
"functions",
"theoretical",
"provide",
"research",
"provides",
"way",
"results",
"example",
"addition"
],
[
"enables",
"developed",
"technique",
"applications",
"techniques",
"research",
"applied",
"useful",
"processing",
"various"
],
[
"optimization",
"objective",
"functions",
"distribution",
"function",
"optimal",
"linear",
"parameters",
"continuous",
"theory"
],
[
"algorithm",
"algorithms",
"linear",
"efficient",
"optimal",
"optimization",
"matrix",
"complexity",
"theoretical",
"dimensional"
],
[
"results",
"present",
"proposed",
"based",
"set",
"experimental",
"new",
"model",
"network",
"recent"
],
[
"critical",
"provides",
"significant",
"research",
"physical",
"systems",
"important",
"user",
"furthermore",
"current"
],
[
"existing",
"specifically",
"require",
"features",
"multiple",
"available",
"approaches",
"usually",
"framework",
"make"
],
[
"size",
"number",
"scale",
"large",
"available",
"need",
"small",
"simple",
"limited",
"parameters"
],
[
"example",
"input",
"given",
"important",
"present",
"image",
"result",
"images",
"available",
"number"
],
[
"understanding",
"efficient",
"technique",
"suggest",
"understand",
"limited",
"better",
"achieve",
"knowledge",
"furthermore"
],
[
"allows",
"various",
"solving",
"specific",
"problems",
"solve",
"solutions",
"result",
"context",
"number"
],
[
"specific",
"use",
"allows",
"important",
"requires",
"make",
"uses",
"available",
"language",
"impact"
],
[
"focus",
"number",
"simple",
"specific",
"particular",
"certain",
"functions",
"structure",
"example",
"ability"
]
] | 4.639518 | average_word_embeddings_glove.6B.300d | 0.546667 | -0.082642 | 0.179665 | 0.765961 |
42 | ArXiv ML Papers | KeyNMF | 40 | [
[
"tasks",
"shown",
"context",
"environment",
"training",
"learning",
"common",
"learn",
"policy",
"making"
],
[
"important",
"suggest",
"standard",
"simple",
"experiments",
"methods",
"structure",
"various",
"theoretical",
"empirical"
],
[
"types",
"detection",
"technique",
"developed",
"using",
"result",
"continuous",
"shown",
"object",
"objects"
],
[
"model",
"models",
"parameters",
"language",
"distribution",
"context",
"design",
"modeling",
"accurate",
"simple"
],
[
"data",
"require",
"possible",
"distribution",
"types",
"identify",
"underlying",
"means",
"furthermore",
"quality"
],
[
"optimal",
"solve",
"consider",
"problem",
"practical",
"solution",
"value",
"model",
"terms",
"solutions"
],
[
"components",
"able",
"object",
"various",
"furthermore",
"examples",
"types",
"different",
"similar",
"individual"
],
[
"resulting",
"simple",
"proposed",
"method",
"dynamic",
"efficient",
"experiments",
"effective",
"accuracy",
"requires"
],
[
"present",
"image",
"images",
"able",
"language",
"visual",
"generate",
"related",
"current",
"features"
],
[
"instead",
"value",
"simple",
"standard",
"difficult",
"approaches",
"requires",
"necessary",
"approach",
"terms"
],
[
"important",
"used",
"improve",
"developed",
"difficult",
"uses",
"effective",
"commonly",
"make",
"mean"
],
[
"standard",
"analysis",
"applied",
"theoretical",
"social",
"developed",
"component",
"application",
"type",
"human"
],
[
"recent",
"experimental",
"results",
"set",
"significant",
"test",
"larger",
"testing",
"indicate",
"models"
],
[
"networks",
"network",
"neural",
"input",
"image",
"various",
"structure",
"standard",
"communication",
"training"
],
[
"work",
"improve",
"shown",
"recent",
"way",
"make",
"human",
"possible",
"able",
"difficult"
],
[
"useful",
"user",
"research",
"important",
"improve",
"information",
"order",
"feature",
"context",
"furthermore"
],
[
"cost",
"time",
"uses",
"making",
"efficiency",
"step",
"current",
"control",
"setting",
"efficient"
],
[
"theoretical",
"provide",
"functions",
"practical",
"way",
"provides",
"research",
"addition",
"empirical",
"example"
],
[
"critical",
"physical",
"important",
"systems",
"current",
"provides",
"user",
"control",
"research",
"significant"
],
[
"functions",
"function",
"optimal",
"parameters",
"linear",
"distribution",
"objective",
"optimization",
"theory",
"continuous"
],
[
"complexity",
"efficient",
"algorithms",
"algorithm",
"optimal",
"linear",
"optimization",
"theoretical",
"matrix",
"dimensional"
],
[
"new",
"model",
"network",
"proposed",
"research",
"support",
"set",
"case",
"based",
"terms"
],
[
"performance",
"improve",
"tasks",
"accuracy",
"training",
"achieve",
"significantly",
"quality",
"recent",
"design"
],
[
"approaches",
"features",
"specifically",
"require",
"usually",
"existing",
"efficiency",
"make",
"addition",
"certain"
],
[
"large",
"need",
"number",
"size",
"parameters",
"scale",
"small",
"simple",
"larger",
"limited"
],
[
"studies",
"research",
"effective",
"application",
"various",
"study",
"subject",
"human",
"related",
"factors"
],
[
"terms",
"context",
"solving",
"number",
"solutions",
"allows",
"solve",
"complexity",
"problems",
"result"
],
[
"better",
"understanding",
"understand",
"achieve",
"suggest",
"efficient",
"technique",
"enables",
"limited",
"means"
],
[
"various",
"specific",
"new",
"way",
"behavior",
"identify",
"tasks",
"types",
"development",
"develop"
],
[
"generate",
"form",
"process",
"processes",
"image",
"quality",
"improve",
"applied",
"model",
"design"
],
[
"structure",
"consider",
"ability",
"simple",
"particular",
"certain",
"number",
"functions",
"focus",
"developed"
],
[
"reasoning",
"user",
"effect",
"knowledge",
"possible",
"useful",
"number",
"language",
"research",
"understanding"
],
[
"processing",
"example",
"computer",
"practical",
"require",
"input",
"applications",
"useful",
"various",
"significant"
],
[
"human",
"ability",
"improve",
"experiments",
"effective",
"images",
"demonstrate",
"image",
"furthermore",
"able"
],
[
"requires",
"impact",
"important",
"quality",
"allows",
"use",
"allow",
"make",
"uses",
"language"
],
[
"multiple",
"make",
"graph",
"structure",
"uses",
"types",
"allows",
"individual",
"representation",
"graphs"
],
[
"efficient",
"able",
"technique",
"techniques",
"furthermore",
"variety",
"applied",
"quality",
"models",
"images"
],
[
"structure",
"allows",
"tasks",
"framework",
"provides",
"complex",
"significant",
"evaluation",
"control",
"achieve"
],
[
"research",
"code",
"processing",
"developed",
"source",
"available",
"models",
"significant",
"addition",
"computer"
],
[
"given",
"probability",
"example",
"input",
"important",
"number",
"result",
"similar",
"accuracy",
"image"
]
] | 5.661899 | average_word_embeddings_glove.6B.300d | 0.5075 | -0.099641 | 0.17106 | 0.788403 |
42 | ArXiv ML Papers | KeyNMF | 50 | [
[
"environment",
"shown",
"training",
"policy",
"tasks",
"context",
"learning",
"learn",
"common",
"development"
],
[
"suggest",
"experiments",
"methods",
"standard",
"specifically",
"approaches",
"distribution",
"combination",
"complexity",
"direct"
],
[
"developed",
"using",
"types",
"detection",
"result",
"technique",
"shown",
"continuous",
"object",
"objects"
],
[
"modeling",
"generate",
"models",
"context",
"typically",
"approaches",
"design",
"language",
"size",
"experiments"
],
[
"data",
"distribution",
"types",
"possible",
"underlying",
"identify",
"devices",
"require",
"quality",
"access"
],
[
"consider",
"problem",
"solve",
"solution",
"optimal",
"value",
"practical",
"solutions",
"terms",
"distribution"
],
[
"able",
"different",
"types",
"examples",
"similar",
"components",
"object",
"individual",
"identify",
"furthermore"
],
[
"accuracy",
"proposed",
"dynamic",
"method",
"representation",
"experiments",
"combination",
"resulting",
"finding",
"detection"
],
[
"feature",
"uses",
"generated",
"image",
"similar",
"features",
"visual",
"images",
"input",
"experiments"
],
[
"allows",
"use",
"requires",
"allow",
"uses",
"impact",
"quality",
"make",
"language",
"difficult"
],
[
"developed",
"used",
"difficult",
"uses",
"commonly",
"mean",
"component",
"components",
"effective",
"possible"
],
[
"application",
"applied",
"standard",
"developed",
"social",
"component",
"analysis",
"theoretical",
"human",
"type"
],
[
"approaches",
"approach",
"requires",
"difficult",
"value",
"necessary",
"instead",
"terms",
"processes",
"control"
],
[
"communication",
"standard",
"network",
"networks",
"neural",
"image",
"training",
"layer",
"detection",
"graph"
],
[
"work",
"recent",
"possible",
"way",
"able",
"human",
"make",
"difficult",
"shown",
"new"
],
[
"order",
"information",
"user",
"feature",
"useful",
"research",
"features",
"social",
"need",
"context"
],
[
"possible",
"setting",
"current",
"cost",
"making",
"control",
"step",
"uses",
"time",
"way"
],
[
"provide",
"theoretical",
"functions",
"practical",
"research",
"addition",
"way",
"provides",
"empirical",
"necessary"
],
[
"control",
"research",
"critical",
"systems",
"current",
"provides",
"user",
"require",
"physical",
"future"
],
[
"functions",
"function",
"optimal",
"linear",
"objective",
"theory",
"distribution",
"continuous",
"probability",
"optimization"
],
[
"achieve",
"features",
"training",
"tasks",
"best",
"performance",
"good",
"design",
"recent",
"feature"
],
[
"based",
"set",
"proposed",
"new",
"network",
"terms",
"research",
"support",
"case",
"language"
],
[
"model",
"distribution",
"modeling",
"representation",
"design",
"instead",
"way",
"allows",
"type",
"source"
],
[
"approaches",
"existing",
"usually",
"require",
"features",
"make",
"specifically",
"contrast",
"certain",
"addition"
],
[
"large",
"scale",
"number",
"size",
"need",
"small",
"set",
"larger",
"limited",
"smaller"
],
[
"studies",
"application",
"study",
"subject",
"effective",
"research",
"related",
"factors",
"characteristics",
"human"
],
[
"solving",
"complexity",
"solutions",
"problems",
"solve",
"terms",
"allows",
"context",
"order",
"theory"
],
[
"better",
"understanding",
"understand",
"achieve",
"suggest",
"technique",
"limited",
"types",
"means",
"enables"
],
[
"algorithm",
"linear",
"optimal",
"algorithms",
"optimization",
"numerical",
"matrix",
"theoretical",
"complexity",
"dimensional"
],
[
"question",
"possible",
"given",
"instead",
"order",
"form",
"number",
"probability",
"result",
"set"
],
[
"terms",
"functions",
"focus",
"number",
"particular",
"certain",
"consider",
"ability",
"processes",
"developed"
],
[
"significant",
"processing",
"require",
"users",
"computer",
"applications",
"useful",
"practical",
"computational",
"computing"
],
[
"significant",
"set",
"larger",
"experimental",
"recent",
"results",
"testing",
"test",
"indicate",
"accuracy"
],
[
"able",
"ability",
"experiments",
"effectiveness",
"images",
"image",
"human",
"effective",
"demonstrate",
"learn"
],
[
"specific",
"way",
"tasks",
"new",
"behavior",
"identify",
"develop",
"development",
"approaches",
"significant"
],
[
"available",
"code",
"research",
"source",
"developed",
"processing",
"significant",
"computer",
"addition",
"especially"
],
[
"variety",
"furthermore",
"technique",
"techniques",
"applied",
"able",
"quality",
"developed",
"images",
"required"
],
[
"framework",
"tasks",
"control",
"proposed",
"allows",
"solution",
"provides",
"achieve",
"evaluation",
"complex"
],
[
"individual",
"make",
"uses",
"multiple",
"allows",
"types",
"graph",
"enables",
"experiments",
"graphs"
],
[
"parameter",
"optimization",
"parameters",
"computational",
"allows",
"cost",
"required",
"processes",
"dimensional",
"similar"
],
[
"useful",
"user",
"effect",
"possible",
"knowledge",
"reasoning",
"language",
"number",
"research",
"experience"
],
[
"experiments",
"theoretical",
"simple",
"effective",
"functions",
"empirical",
"standard",
"requires",
"tasks",
"ability"
],
[
"variety",
"ways",
"similar",
"training",
"various",
"human",
"effective",
"applied",
"tasks",
"extensive"
],
[
"structures",
"structural",
"structure",
"complex",
"dimensional",
"variables",
"graph",
"defined",
"underlying",
"representation"
],
[
"quality",
"significantly",
"improve",
"making",
"increase",
"efficiency",
"significant",
"accuracy",
"increasing",
"terms"
],
[
"generate",
"process",
"form",
"image",
"applied",
"processes",
"allows",
"quality",
"set",
"design"
],
[
"focus",
"providing",
"similar",
"examples",
"example",
"object",
"theory",
"possible",
"allows",
"concept"
],
[
"especially",
"question",
"technique",
"furthermore",
"common",
"significant",
"making",
"number",
"important",
"make"
],
[
"present",
"language",
"able",
"current",
"new",
"related",
"visual",
"including",
"images",
"image"
],
[
"efficiency",
"solutions",
"functions",
"efficient",
"cost",
"result",
"design",
"develop",
"significant",
"communication"
]
] | 4.755701 | average_word_embeddings_glove.6B.300d | 0.474 | -0.110715 | 0.17296 | 0.805725 |
42 | ArXiv ML Papers | FASTopic | 10 | [
[
"em",
"posterior",
"mutual",
"equation",
"estimators",
"variational",
"carlo",
"kernels",
"monte",
"mcmc"
],
[
"optimal",
"stochastic",
"linear",
"gradient",
"convergence",
"function",
"optimization",
"algorithm",
"sample",
"matrix"
],
[
"federated",
"environment",
"research",
"agent",
"traffic",
"online",
"agents",
"user",
"users",
"reinforcement"
],
[
"examples",
"biases",
"perturbations",
"bias",
"predictors",
"attacks",
"fairness",
"adversarial",
"causal",
"adversaries"
],
[
"was",
"segmentation",
"were",
"images",
"audio",
"speech",
"medical",
"video",
"imaging",
"disease"
],
[
"feature",
"representation",
"neural",
"networks",
"network",
"end",
"image",
"architecture",
"convolutional",
"generative"
],
[
"model",
"machine",
"data",
"datasets",
"models",
"classification",
"task",
"from",
"tasks",
"different"
],
[
"graphs",
"geometric",
"link",
"gnns",
"node",
"nodes",
"objects",
"clusters",
"graph",
"clustering"
],
[
"semantic",
"word",
"visual",
"language",
"text",
"natural",
"translation",
"reasoning",
"style",
"languages"
],
[
"quantum",
"efficiency",
"memory",
"computation",
"scaling",
"hardware",
"quantization",
"overhead",
"latency",
"bit"
]
] | 56.874147 | average_word_embeddings_glove.6B.300d | 1 | -0.079675 | 0.180779 | 0.857534 |
42 | ArXiv ML Papers | FASTopic | 20 | [
[
"nonlinear",
"link",
"gnns",
"graph",
"graphs",
"nodes",
"local",
"node",
"flows",
"diffusion"
],
[
"vehicle",
"driving",
"iot",
"traffic",
"sensors",
"coverage",
"smart",
"wireless",
"vehicles",
"predefined"
],
[
"linguistic",
"contexts",
"proposals",
"explanations",
"sentence",
"boxes",
"meaning",
"relevance",
"assigned",
"document"
],
[
"bayesian",
"posterior",
"variational",
"inference",
"carlo",
"nonparametric",
"dirichlet",
"mcmc",
"nonnegative",
"monte"
],
[
"width",
"molecular",
"pooling",
"euclidean",
"orthogonal",
"activation",
"separable",
"tensors",
"tensor",
"spectrum"
],
[
"generation",
"gan",
"generative",
"enhancement",
"music",
"character",
"frame",
"acoustic",
"composed",
"mismatch"
],
[
"object",
"representations",
"representation",
"features",
"visual",
"feature",
"objects",
"specific",
"human",
"detection"
],
[
"image",
"images",
"3d",
"segmentation",
"convolutional",
"imaging",
"augmentation",
"eeg",
"mri",
"localization"
],
[
"privacy",
"users",
"user",
"service",
"media",
"item",
"federated",
"internet",
"concerns",
"services"
],
[
"power",
"memory",
"energy",
"forecasting",
"forecast",
"forecasts",
"consumption",
"demand",
"temperature",
"latency"
],
[
"bounds",
"sqrt",
"delta",
"bound",
"frac",
"mathbb",
"tilde",
"mathcal",
"arms",
"ucb"
],
[
"techniques",
"datasets",
"dataset",
"accuracy",
"available",
"art",
"scale",
"classification",
"supervised",
"compared"
],
[
"programs",
"reasoning",
"logic",
"abstraction",
"enforce",
"formalism",
"symbolic",
"inefficient",
"easier",
"add"
],
[
"distribution",
"error",
"parameters",
"estimation",
"loss",
"samples",
"generalization",
"regression",
"sample",
"class"
],
[
"in",
"and",
"of",
"on",
"as",
"the",
"to",
"this",
"that",
"are"
],
[
"language",
"attacks",
"speech",
"adversarial",
"source",
"text",
"target",
"pre",
"domain",
"against"
],
[
"methodologies",
"investigation",
"centric",
"project",
"materials",
"becoming",
"ml",
"research",
"practices",
"discusses"
],
[
"policy",
"agent",
"reinforcement",
"decision",
"rl",
"control",
"agents",
"reward",
"strategies",
"goal"
],
[
"curve",
"diagnosis",
"cancer",
"detector",
"patients",
"patient",
"diseases",
"healthy",
"97",
"94"
],
[
"optimization",
"algorithm",
"stochastic",
"convergence",
"matrix",
"sparse",
"problems",
"linear",
"algorithms",
"gradient"
]
] | 416.58248 | average_word_embeddings_glove.6B.300d | 1 | -0.129604 | 0.166474 | 0.873438 |
42 | ArXiv ML Papers | FASTopic | 30 | [
[
"demands",
"resnet",
"configurations",
"modes",
"normalization",
"transportation",
"200",
"updated",
"pruned",
"managing"
],
[
"attack",
"adversarial",
"defense",
"malware",
"against",
"attacks",
"detected",
"malicious",
"threat",
"vulnerable"
],
[
"root",
"periods",
"deviations",
"exists",
"outliers",
"vanishing",
"relu",
"recovery",
"boundary",
"recover"
],
[
"reward",
"control",
"action",
"policies",
"policy",
"agents",
"rl",
"agent",
"reinforcement",
"environments"
],
[
"mobile",
"devices",
"device",
"60",
"wireless",
"throughput",
"enhancement",
"centralized",
"server",
"lightweight"
],
[
"quantum",
"extensions",
"logic",
"straightforward",
"shelf",
"correctness",
"formalism",
"formally",
"branch",
"classical"
],
[
"uncertainty",
"dropout",
"bayesian",
"streams",
"insufficient",
"uncertainties",
"observational",
"imputation",
"measurement",
"uncertain"
],
[
"user",
"users",
"twitter",
"name",
"recommendation",
"website",
"personal",
"media",
"item",
"recommender"
],
[
"function",
"algorithm",
"algorithms",
"optimal",
"bounds",
"linear",
"problem",
"case",
"epsilon",
"bound"
],
[
"gradient",
"iterations",
"convexity",
"proximal",
"convergence",
"descent",
"readily",
"accelerated",
"solvers",
"sgd"
],
[
"to",
"data",
"model",
"and",
"on",
"in",
"from",
"by",
"are",
"of"
],
[
"structure",
"domain",
"feature",
"features",
"representations",
"representation",
"information",
"learn",
"unsupervised",
"source"
],
[
"regret",
"bandit",
"sqrt",
"bandits",
"entries",
"armed",
"delta",
"ucb",
"arms",
"uniformly"
],
[
"dimensional",
"matrix",
"kernel",
"sparse",
"low",
"computational",
"dimensionality",
"method",
"clustering",
"rank"
],
[
"music",
"sentences",
"audio",
"acoustic",
"languages",
"word",
"speaker",
"english",
"corpus",
"speech"
],
[
"inform",
"humans",
"human",
"reasoning",
"explanations",
"explanation",
"explainable",
"perception",
"manipulation",
"robotics"
],
[
"nas",
"segmentation",
"image",
"pixel",
"images",
"patch",
"imaging",
"optical",
"reconstructed",
"patches"
],
[
"regression",
"monte",
"em",
"varepsilon",
"mcmc",
"logistic",
"square",
"computes",
"lasso",
"estimator"
],
[
"some",
"ml",
"many",
"systems",
"research",
"applications",
"techniques",
"machine",
"how",
"several"
],
[
"ideal",
"1d",
"convolutions",
"tensor",
"tensors",
"coordinates",
"activation",
"spectrum",
"accelerators",
"pooling"
],
[
"nodes",
"graph",
"graphs",
"node",
"clusters",
"gnns",
"vertex",
"directed",
"distinguishing",
"passing"
],
[
"tasks",
"task",
"level",
"attention",
"language",
"text",
"natural",
"knowledge",
"trained",
"bert"
],
[
"prediction",
"dataset",
"detection",
"were",
"classification",
"system",
"was",
"test",
"sensors",
"monitoring"
],
[
"label",
"labels",
"classifier",
"privacy",
"private",
"perturbations",
"fair",
"class",
"treatment",
"bias"
],
[
"discriminative",
"generative",
"inductive",
"localization",
"intuition",
"impressive",
"svhn",
"discovering",
"invariance",
"performances"
],
[
"neural",
"networks",
"layer",
"architecture",
"memory",
"architectures",
"network",
"loss",
"power",
"convolutional"
],
[
"forecasting",
"series",
"modeling",
"variables",
"relational",
"causal",
"rnns",
"inferring",
"forecasts",
"phenomena"
],
[
"price",
"formalize",
"imbalance",
"fairness",
"interventions",
"credit",
"assessing",
"compliance",
"criteria",
"biases"
],
[
"disease",
"clinical",
"health",
"cancer",
"papers",
"methodologies",
"medicine",
"diseases",
"databases",
"countries"
],
[
"inducing",
"variational",
"equations",
"distributions",
"symmetric",
"equation",
"posterior",
"hilbert",
"analytically",
"covariance"
]
] | 397.196201 | average_word_embeddings_glove.6B.300d | 1 | -0.175521 | 0.149348 | 0.882518 |
42 | ArXiv ML Papers | FASTopic | 40 | [
[
"centralized",
"server",
"engine",
"accelerators",
"federated",
"library",
"service",
"devices",
"execute",
"latency"
],
[
"rl",
"policy",
"agent",
"reinforcement",
"agents",
"forgetting",
"robot",
"actions",
"reward",
"action"
],
[
"urban",
"year",
"records",
"spread",
"market",
"weather",
"city",
"reached",
"days",
"day"
],
[
"cost",
"scaling",
"healthy",
"bit",
"memory",
"efficiency",
"reduced",
"efficient",
"filter",
"precision"
],
[
"thorough",
"explaining",
"qualitative",
"explanation",
"relational",
"strengths",
"explanations",
"causal",
"concepts",
"papers"
],
[
"nodes",
"node",
"graph",
"link",
"graphs",
"gnns",
"vertices",
"edges",
"gnn",
"vertex"
],
[
"minutes",
"xgboost",
"par",
"accuracies",
"82",
"streams",
"percentage",
"stream",
"anomalies",
"boosting"
],
[
"quantum",
"mathematical",
"theorem",
"correctness",
"logic",
"formal",
"extensions",
"equally",
"package",
"occur"
],
[
"generative",
"models",
"prediction",
"model",
"modeling",
"series",
"predictions",
"forecasts",
"forecasting",
"forecast"
],
[
"systems",
"machine",
"research",
"ml",
"process",
"some",
"social",
"how",
"applications",
"recent"
],
[
"heuristic",
"communication",
"distributed",
"hundreds",
"impractical",
"messages",
"search",
"reach",
"searching",
"round"
],
[
"convergence",
"problem",
"gradient",
"optimization",
"convex",
"optimal",
"function",
"non",
"problems",
"stochastic"
],
[
"gpus",
"transmission",
"layer",
"backbone",
"convolutional",
"1d",
"wireless",
"enhancement",
"modules",
"architecture"
],
[
"et",
"dirichlet",
"em",
"clustering",
"nonnegative",
"clusters",
"al",
"2015",
"provable",
"families"
],
[
"images",
"image",
"object",
"segmentation",
"3d",
"scene",
"pixel",
"video",
"videos",
"camera"
],
[
"item",
"preferences",
"users",
"user",
"personal",
"recommendation",
"ranking",
"recommender",
"interactive",
"online"
],
[
"fairness",
"gender",
"biases",
"bias",
"interventions",
"fair",
"outcomes",
"treatment",
"criteria",
"affected"
],
[
"regression",
"estimator",
"logistic",
"interval",
"covariates",
"decay",
"squared",
"deviation",
"error",
"estimators"
],
[
"anomaly",
"system",
"traffic",
"detection",
"time",
"proposed",
"was",
"data",
"high",
"power"
],
[
"bandits",
"optimality",
"tilde",
"armed",
"regret",
"sqrt",
"bandit",
"arms",
"ucb",
"strictly"
],
[
"operates",
"pruning",
"backpropagation",
"operate",
"reconstructing",
"pruned",
"nas",
"paths",
"cifar10",
"removing"
],
[
"2021",
"rarely",
"credit",
"compliance",
"experiences",
"imitation",
"difficulties",
"drive",
"counterfactual",
"consequences"
],
[
"word",
"sentences",
"language",
"bert",
"languages",
"text",
"sentence",
"name",
"bleu",
"linguistic"
],
[
"protein",
"cognitive",
"resonance",
"brain",
"biological",
"artifacts",
"rnns",
"recognizing",
"observing",
"eeg"
],
[
"distance",
"label",
"metric",
"features",
"feature",
"nearest",
"classifier",
"classification",
"set",
"correlation"
],
[
"disease",
"clinical",
"patients",
"patient",
"diagnosis",
"health",
"diseases",
"healthcare",
"wearable",
"cancer"
],
[
"unlabeled",
"art",
"training",
"trained",
"supervised",
"datasets",
"methods",
"large",
"dataset",
"labeled"
],
[
"control",
"trajectory",
"controller",
"delay",
"equations",
"dynamics",
"safe",
"transitions",
"informed",
"transition"
],
[
"adversarial",
"attacks",
"robustness",
"against",
"robust",
"perturbations",
"attack",
"defense",
"attacker",
"vulnerable"
],
[
"distributions",
"inference",
"bayesian",
"variational",
"estimation",
"distribution",
"parameter",
"posterior",
"variables",
"gaussian"
],
[
"hilbert",
"coordinates",
"width",
"curse",
"functions",
"infinite",
"vanishing",
"activation",
"relu",
"kernels"
],
[
"extracts",
"invariance",
"normalization",
"remove",
"augment",
"pooling",
"vae",
"auxiliary",
"mixture",
"patch"
],
[
"domain",
"representations",
"representation",
"source",
"knowledge",
"task",
"tasks",
"target",
"transfer",
"learn"
],
[
"project",
"discusses",
"functionality",
"technologies",
"projects",
"becoming",
"materials",
"easier",
"software",
"engineering"
],
[
"tensor",
"hessian",
"squares",
"subspace",
"proximal",
"thresholding",
"newton",
"sgd",
"solvers",
"principal"
],
[
"audio",
"visual",
"speech",
"recognition",
"media",
"speaker",
"voice",
"twitter",
"participants",
"attention"
],
[
"with",
"on",
"as",
"by",
"can",
"of",
"neural",
"networks",
"network",
"to"
],
[
"algorithm",
"algorithms",
"linear",
"bounds",
"sample",
"matrix",
"complexity",
"random",
"number",
"theoretical"
],
[
"ideas",
"latent",
"formalism",
"dimensional",
"enforce",
"topological",
"transformation",
"coding",
"geometric",
"transformations"
],
[
"quality",
"driving",
"dataset",
"test",
"human",
"study",
"system",
"their",
"best",
"not"
]
] | 396.602338 | average_word_embeddings_glove.6B.300d | 0.995 | -0.161884 | 0.158562 | 0.875394 |
42 | ArXiv ML Papers | FASTopic | 50 | [
[
"market",
"credit",
"price",
"compliance",
"financial",
"drive",
"service",
"spread",
"demand",
"managing"
],
[
"subspaces",
"squares",
"normalizing",
"approximates",
"laplacian",
"newton",
"solvers",
"approximating",
"approximated",
"subspace"
],
[
"these",
"been",
"deep",
"tasks",
"machine",
"how",
"systems",
"their",
"task",
"have"
],
[
"conclude",
"qualitative",
"quantitative",
"assessing",
"methodologies",
"biases",
"explanations",
"know",
"thorough",
"bias"
],
[
"attack",
"attacker",
"adversarial",
"adversaries",
"against",
"attacks",
"defense",
"robots",
"vulnerable",
"perturbation"
],
[
"accuracies",
"25",
"60",
"82",
"94",
"resnet",
"xgboost",
"18",
"day",
"91"
],
[
"matrix",
"order",
"linear",
"dimensional",
"tensor",
"rank",
"optimization",
"problems",
"sparse",
"kernel"
],
[
"combinatorial",
"add",
"heuristic",
"encourages",
"appealing",
"search",
"heuristics",
"predefined",
"programming",
"inefficient"
],
[
"forces",
"discovery",
"relational",
"functional",
"phenomena",
"causal",
"inferring",
"dynamics",
"biological",
"capturing"
],
[
"variance",
"distribution",
"error",
"sample",
"regression",
"estimation",
"estimator",
"bound",
"sampling",
"gaussian"
],
[
"research",
"project",
"discusses",
"industry",
"media",
"workflow",
"ml",
"name",
"software",
"twitter"
],
[
"perturbations",
"noise",
"generalization",
"training",
"loss",
"normalization",
"improve",
"robustness",
"robust",
"overfitting"
],
[
"dictionary",
"nonnegative",
"curse",
"metric",
"orthogonal",
"factorization",
"manifold",
"distance",
"euclidean",
"color"
],
[
"dirichlet",
"observational",
"modification",
"dropout",
"multivariate",
"calibrated",
"svhn",
"additive",
"probabilistic",
"latent"
],
[
"volume",
"streams",
"stream",
"24",
"measurement",
"measurements",
"trends",
"spatio",
"anomalies",
"minutes"
],
[
"speedups",
"configurations",
"whilst",
"retaining",
"optimizes",
"surrogate",
"runtime",
"complexities",
"oracle",
"complement"
],
[
"transformers",
"vae",
"sequence",
"character",
"programs",
"asr",
"scratch",
"suite",
"python",
"hope"
],
[
"lipschitz",
"delta",
"frac",
"varepsilon",
"arms",
"omega",
"mathbb",
"mathcal",
"ucb",
"regime"
],
[
"positive",
"views",
"occurs",
"categorical",
"boxes",
"relation",
"proposal",
"distinguishing",
"correspond",
"adaptively"
],
[
"are",
"in",
"by",
"on",
"of",
"to",
"this",
"and",
"that",
"the"
],
[
"clustering",
"supervised",
"available",
"samples",
"datasets",
"synthetic",
"methods",
"data",
"dataset",
"large"
],
[
"sentence",
"translate",
"sentences",
"bidirectional",
"names",
"corpora",
"linguistic",
"bleu",
"english",
"corpus"
],
[
"points",
"properties",
"class",
"theoretical",
"provide",
"functions",
"general",
"function",
"non",
"classes"
],
[
"stochastic",
"momentum",
"descent",
"sgd",
"gradient",
"proximal",
"convergence",
"optimisation",
"convexity",
"nonconvex"
],
[
"recommendation",
"feature",
"source",
"domain",
"item",
"target",
"users",
"knowledge",
"user",
"domains"
],
[
"federated",
"decentralized",
"strict",
"centralized",
"server",
"exhaustive",
"distillation",
"distributed",
"overhead",
"heterogeneity"
],
[
"algorithm",
"online",
"bandit",
"queries",
"regret",
"optimal",
"algorithms",
"decision",
"problem",
"setting"
],
[
"models",
"series",
"modeling",
"prediction",
"forecasting",
"model",
"predictions",
"forecast",
"uncertainty",
"predictive"
],
[
"outliers",
"spectrum",
"spatially",
"double",
"count",
"sketch",
"consequence",
"em",
"localized",
"temperature"
],
[
"distributions",
"bayesian",
"wasserstein",
"variational",
"bernoulli",
"nonparametric",
"carlo",
"posterior",
"imply",
"pac"
],
[
"best",
"feature",
"was",
"classifier",
"classification",
"ct",
"features",
"were",
"accuracy",
"test"
],
[
"speech",
"text",
"speaker",
"language",
"style",
"music",
"representations",
"acoustic",
"semantic",
"attention"
],
[
"extracts",
"label",
"labels",
"labeling",
"patches",
"codes",
"exploited",
"cells",
"weak",
"selects"
],
[
"fourier",
"informed",
"quantum",
"classical",
"decades",
"nonlinear",
"differentiation",
"tensors",
"equations",
"derivatives"
],
[
"covid",
"disease",
"cancer",
"patients",
"diagnosis",
"patient",
"protein",
"weather",
"curve",
"diseases"
],
[
"imputation",
"nearest",
"square",
"screening",
"calibration",
"largest",
"interval",
"neighbor",
"root",
"hyperparameters"
],
[
"quantized",
"lightweight",
"enhancement",
"1d",
"bandwidth",
"transmission",
"channels",
"compress",
"accelerators",
"channel"
],
[
"graphs",
"nodes",
"link",
"gnns",
"graph",
"node",
"gnn",
"vertex",
"vertices",
"edges"
],
[
"imbalance",
"augmentation",
"annotated",
"annotation",
"annotations",
"database",
"transferring",
"sound",
"textual",
"actively"
],
[
"scene",
"images",
"object",
"3d",
"objects",
"video",
"shape",
"image",
"segmentation",
"resolution"
],
[
"urban",
"city",
"device",
"demands",
"infrastructure",
"mobile",
"transportation",
"rapid",
"cloud",
"consumption"
],
[
"enabling",
"manipulation",
"mri",
"artifacts",
"optical",
"creates",
"modal",
"catastrophic",
"camera",
"reconstruct"
],
[
"outlier",
"detection",
"detector",
"iot",
"sensor",
"sensors",
"driving",
"vehicle",
"detect",
"malware"
],
[
"explainable",
"human",
"experiences",
"humans",
"goals",
"force",
"expert",
"players",
"robotic",
"imitation"
],
[
"contextual",
"answers",
"fairness",
"fair",
"counterfactual",
"interventions",
"preferences",
"identity",
"notion",
"learner"
],
[
"concepts",
"reasoning",
"learners",
"abstraction",
"materials",
"cognitive",
"forgetting",
"reason",
"facilitating",
"logic"
],
[
"networks",
"architecture",
"neural",
"layer",
"convolutional",
"network",
"layers",
"cnns",
"architectures",
"pruning"
],
[
"efficiency",
"efficient",
"memory",
"cost",
"accuracy",
"energy",
"computational",
"scaling",
"achieve",
"performance"
],
[
"feed",
"activations",
"architectural",
"activation",
"backpropagation",
"neuron",
"neurons",
"paths",
"recursive",
"movement"
],
[
"control",
"agents",
"agent",
"rl",
"reward",
"policy",
"reinforcement",
"policies",
"action",
"games"
]
] | 401.022002 | average_word_embeddings_glove.6B.300d | 0.996 | -0.197069 | 0.154347 | 0.893485 |
42 | ArXiv ML Papers | S³_angular | 10 | [
[
"expression",
"experiences",
"similarities",
"characterized",
"belief",
"nature",
"differences",
"relationship",
"explain",
"understood"
],
[
"producing",
"equipped",
"combine",
"navigation",
"audio",
"robot",
"digital",
"memory",
"transformers",
"enabled"
],
[
"topological",
"vertices",
"symmetric",
"manifold",
"graphs",
"graph",
"orthogonal",
"topology",
"vertex",
"tensor"
],
[
"prediction",
"forecasting",
"projections",
"forecasts",
"forecast",
"models",
"model",
"comparable",
"predictions",
"robust"
],
[
"particle",
"artificial",
"equations",
"geometry",
"synthesis",
"quantum",
"technique",
"nonlinear",
"experiment",
"differential"
],
[
"foundation",
"include",
"library",
"science",
"technical",
"focuses",
"established",
"languages",
"engineering",
"dedicated"
],
[
"given",
"point",
"number",
"times",
"small",
"far",
"time",
"numbers",
"large",
"initial"
],
[
"introduce",
"compact",
"essentially",
"simplicity",
"platform",
"dynamic",
"mode",
"dual",
"traditional",
"attractive"
],
[
"max",
"vs",
"semi",
"dl",
"ml",
"99",
"min",
"f1",
"par",
"88"
],
[
"accelerate",
"transition",
"extend",
"extending",
"accelerated",
"controlling",
"rapid",
"expansion",
"contribute",
"tremendous"
]
] | 0.934259 | average_word_embeddings_glove.6B.300d | 1 | -0.338978 | 0.226506 | 0.890888 |
42 | ArXiv ML Papers | S³_angular | 20 | [
[
"choosing",
"typical",
"combinations",
"selecting",
"accordingly",
"vary",
"overlap",
"differ",
"desirable",
"depending"
],
[
"flows",
"catastrophic",
"forecasts",
"predicted",
"magnitude",
"forecast",
"weather",
"worst",
"expected",
"traffic"
],
[
"armed",
"attacks",
"targeting",
"forces",
"regime",
"prevent",
"deployed",
"attack",
"bandit",
"threat"
],
[
"module",
"gps",
"parameters",
"validation",
"parameter",
"calibration",
"calibrated",
"computed",
"configuration",
"initialization"
],
[
"added",
"given",
"mean",
"zero",
"order",
"sum",
"right",
"adding",
"instance",
"means"
],
[
"network",
"neural",
"networks",
"connections",
"channels",
"channel",
"backpropagation",
"neurons",
"propagation",
"posterior"
],
[
"loss",
"lack",
"poor",
"despite",
"performance",
"weak",
"losses",
"worse",
"handling",
"attention"
],
[
"population",
"families",
"branch",
"applicable",
"classifying",
"centers",
"apply",
"poorly",
"communities",
"classes"
],
[
"implementing",
"normalization",
"consideration",
"framework",
"policies",
"policy",
"stability",
"integration",
"ensuring",
"recommendation"
],
[
"consensus",
"correct",
"follow",
"target",
"possible",
"date",
"conclusion",
"scenario",
"decisions",
"suggested"
],
[
"cancer",
"patients",
"diseases",
"disease",
"prevalence",
"clinical",
"brain",
"diagnosis",
"genetic",
"researchers"
],
[
"graph",
"vertices",
"vertex",
"nodes",
"node",
"graphs",
"convex",
"closely",
"topology",
"edges"
],
[
"special",
"includes",
"unit",
"currently",
"consists",
"program",
"addition",
"consisting",
"form",
"new"
],
[
"textual",
"paper",
"usage",
"derived",
"incomplete",
"statistical",
"extraction",
"estimates",
"standard",
"definition"
],
[
"quality",
"excellent",
"boost",
"high",
"competitive",
"efficiency",
"low",
"improved",
"better",
"improve"
],
[
"convergence",
"obtaining",
"facilitate",
"filtering",
"gradient",
"relaxation",
"intra",
"advance",
"directions",
"sensitive"
],
[
"strongly",
"viewed",
"nature",
"independent",
"observed",
"positive",
"biased",
"described",
"describes",
"relationship"
],
[
"web",
"internet",
"users",
"online",
"twitter",
"google",
"website",
"server",
"distributed",
"offline"
],
[
"optimization",
"stochastic",
"variational",
"optimisation",
"iterative",
"minimization",
"nonlinear",
"dirichlet",
"deterministic",
"multivariate"
],
[
"sensitive",
"specific",
"processes",
"certain",
"inherent",
"characteristics",
"particular",
"minimize",
"risks",
"mechanisms"
]
] | 1.008063 | average_word_embeddings_glove.6B.300d | 0.995 | -0.318467 | 0.232512 | 0.89284 |
42 | ArXiv ML Papers | S³_angular | 30 | [
[
"16",
"19",
"28",
"13",
"17",
"12",
"14",
"18",
"71",
"24"
],
[
"challenging",
"demands",
"elements",
"purely",
"context",
"compositional",
"dimension",
"constraints",
"interpretation",
"unprecedented"
],
[
"proof",
"explanation",
"clear",
"valid",
"details",
"answer",
"explanations",
"make",
"reason",
"making"
],
[
"region",
"city",
"leading",
"highest",
"represented",
"ranking",
"prominent",
"rank",
"known",
"high"
],
[
"research",
"studies",
"science",
"explored",
"focused",
"scientific",
"literature",
"focusing",
"topics",
"study"
],
[
"music",
"interesting",
"sampled",
"combinations",
"entries",
"shows",
"turns",
"times",
"variations",
"typical"
],
[
"incorporating",
"introducing",
"approaches",
"formulation",
"designs",
"aspects",
"design",
"style",
"geometric",
"designing"
],
[
"neurons",
"connections",
"backpropagation",
"neural",
"networks",
"convolutional",
"network",
"channels",
"programming",
"connected"
],
[
"candidates",
"countries",
"participants",
"transition",
"parallel",
"basis",
"responses",
"simultaneously",
"measured",
"proportional"
],
[
"leveraged",
"outperform",
"consumption",
"price",
"market",
"share",
"trend",
"globally",
"fidelity",
"rates"
],
[
"estimator",
"regression",
"bayesian",
"iterative",
"parametric",
"multivariate",
"nonparametric",
"interpolation",
"bayes",
"stochastic"
],
[
"practice",
"overhead",
"poor",
"manual",
"mean",
"conditioning",
"running",
"driving",
"fine",
"trains"
],
[
"maps",
"predicting",
"forecast",
"short",
"forecasting",
"projections",
"prediction",
"forecasts",
"weather",
"traffic"
],
[
"detector",
"magnetic",
"transformer",
"physics",
"particle",
"chemical",
"molecular",
"power",
"light",
"fusion"
],
[
"parameter",
"variance",
"length",
"width",
"bounds",
"approximate",
"maximum",
"square",
"posterior",
"extent"
],
[
"critical",
"success",
"production",
"performance",
"performs",
"improvement",
"produced",
"role",
"successful",
"absence"
],
[
"efficacy",
"basis",
"validated",
"provided",
"uniformly",
"conditional",
"conditioned",
"strongly",
"hessian",
"aggregate"
],
[
"conversion",
"nets",
"whilst",
"relying",
"account",
"sources",
"assumed",
"retaining",
"temporal",
"delta"
],
[
"realized",
"tremendous",
"surprising",
"shed",
"enormous",
"accelerate",
"importantly",
"confidence",
"deeper",
"remarkable"
],
[
"interface",
"compatible",
"software",
"server",
"hardware",
"functionality",
"implementations",
"embedded",
"computing",
"platforms"
],
[
"model",
"modeling",
"models",
"modeled",
"fit",
"car",
"compact",
"fashion",
"modelling",
"version"
],
[
"words",
"english",
"language",
"languages",
"word",
"speech",
"translation",
"speaker",
"texts",
"text"
],
[
"attacks",
"attack",
"targeting",
"threat",
"attacker",
"forces",
"targeted",
"attempts",
"adversaries",
"armed"
],
[
"catastrophic",
"anomalies",
"consequence",
"arising",
"consequences",
"resulting",
"affecting",
"occur",
"caused",
"arise"
],
[
"suggested",
"desirable",
"actual",
"determining",
"sampling",
"optimal",
"date",
"reasonable",
"likelihood",
"population"
],
[
"consist",
"primary",
"classes",
"predictors",
"interval",
"intermediate",
"typical",
"auxiliary",
"factors",
"vary"
],
[
"fourier",
"mathematical",
"algorithms",
"computer",
"quantum",
"computation",
"asymptotic",
"advances",
"computational",
"contributions"
],
[
"amenable",
"various",
"appear",
"perform",
"computationally",
"augment",
"approximations",
"suboptimal",
"pairwise",
"ensembles"
],
[
"substantially",
"scaling",
"compression",
"resulting",
"substantial",
"throughput",
"reduced",
"processing",
"reduction",
"costs"
],
[
"block",
"eventually",
"called",
"extension",
"transfer",
"allowing",
"later",
"allow",
"delta",
"blocks"
]
] | 1.071836 | average_word_embeddings_glove.6B.300d | 0.986667 | -0.358635 | 0.215587 | 0.906856 |
42 | ArXiv ML Papers | S³_angular | 40 | [
[
"comparable",
"remarkably",
"consistently",
"impressive",
"estimators",
"explainable",
"sufficiently",
"rarely",
"grained",
"scalability"
],
[
"unknown",
"temporal",
"consequently",
"assumed",
"conversion",
"delta",
"assuming",
"sources",
"widely",
"simpler"
],
[
"formalize",
"exploitation",
"combines",
"utilizing",
"utilizes",
"dependency",
"pooling",
"mechanisms",
"involves",
"modelling"
],
[
"reach",
"proven",
"labels",
"aim",
"viable",
"aiming",
"reliable",
"alternatives",
"feasible",
"alternative"
],
[
"modeled",
"independently",
"supervised",
"overfitting",
"learnable",
"denoising",
"updating",
"investigates",
"gans",
"annotated"
],
[
"forces",
"threat",
"targeting",
"attack",
"attacks",
"armed",
"attempts",
"targeted",
"attempt",
"malicious"
],
[
"model",
"models",
"mini",
"modeling",
"autoregressive",
"car",
"surrogate",
"built",
"hybrid",
"drive"
],
[
"arms",
"superior",
"requiring",
"count",
"obtaining",
"block",
"allowing",
"received",
"recognition",
"procedure"
],
[
"functionality",
"compatible",
"interface",
"developers",
"software",
"server",
"mobile",
"hardware",
"95",
"embedded"
],
[
"wise",
"valued",
"storage",
"valuable",
"redundant",
"purpose",
"utility",
"metrics",
"defense",
"pac"
],
[
"genetic",
"synthesis",
"protein",
"molecular",
"differentiation",
"activation",
"cells",
"domains",
"encodes",
"interactions"
],
[
"formalize",
"supported",
"validated",
"conditional",
"normalizing",
"strongly",
"activations",
"basis",
"place",
"validation"
],
[
"vehicle",
"passing",
"bridge",
"traffic",
"transport",
"transportation",
"service",
"car",
"nearest",
"train"
],
[
"surprisingly",
"extremely",
"makes",
"kind",
"prone",
"intelligent",
"inherently",
"compact",
"classifier",
"proved"
],
[
"extension",
"extended",
"exploration",
"short",
"discovery",
"sequences",
"conversion",
"space",
"search",
"expansion"
],
[
"distances",
"configurations",
"runtime",
"suitable",
"dataset",
"contexts",
"variability",
"locations",
"variances",
"exhibits"
],
[
"cnn",
"propagation",
"backpropagation",
"neural",
"cells",
"networks",
"network",
"programming",
"convolutional",
"neurons"
],
[
"quantum",
"classical",
"age",
"theory",
"modern",
"example",
"authors",
"literature",
"life",
"phenomena"
],
[
"condition",
"guarantees",
"present",
"necessary",
"provided",
"compliance",
"relevant",
"complete",
"guarantee",
"return"
],
[
"policy",
"policies",
"aim",
"formulate",
"aims",
"flexible",
"adopt",
"appropriate",
"optimal",
"objectives"
],
[
"objects",
"2019",
"70",
"magnitude",
"2017",
"scale",
"90",
"relation",
"object",
"80"
],
[
"observe",
"survey",
"analyzed",
"experts",
"examine",
"characterize",
"believe",
"closely",
"analyze",
"suggested"
],
[
"twitter",
"feedback",
"online",
"offline",
"personalized",
"recommendations",
"users",
"user",
"responses",
"options"
],
[
"infrastructure",
"substantially",
"incremental",
"constructing",
"redundant",
"suggests",
"construction",
"extraction",
"augmentation",
"inefficient"
],
[
"color",
"vision",
"camera",
"eye",
"image",
"viewed",
"images",
"pixels",
"create",
"3d"
],
[
"features",
"feature",
"includes",
"identifying",
"profile",
"unique",
"attributes",
"highlights",
"character",
"defining"
],
[
"research",
"biology",
"science",
"scientific",
"contributions",
"advances",
"studies",
"engineering",
"center",
"developments"
],
[
"changes",
"class",
"fail",
"requirements",
"change",
"meet",
"program",
"code",
"accelerate",
"failure"
],
[
"translation",
"meaning",
"language",
"texts",
"languages",
"words",
"word",
"english",
"derived",
"dictionary"
],
[
"vehicles",
"codes",
"kinds",
"boxes",
"items",
"covariates",
"sensors",
"devices",
"labels",
"sets"
],
[
"exposure",
"entropy",
"scheme",
"provable",
"incorporates",
"rl",
"distributional",
"sparsity",
"coverage",
"distortion"
],
[
"design",
"aspects",
"aspect",
"max",
"architectural",
"designing",
"characteristics",
"critic",
"improvements",
"wasserstein"
],
[
"cancer",
"patients",
"disease",
"diagnosis",
"heart",
"diseases",
"treat",
"clinical",
"patient",
"health"
],
[
"training",
"learned",
"guidance",
"trained",
"calibration",
"learnt",
"learn",
"evaluations",
"assigned",
"background"
],
[
"ones",
"groups",
"families",
"especially",
"locally",
"mainly",
"counterparts",
"communities",
"notably",
"benefit"
],
[
"unclear",
"date",
"area",
"early",
"short",
"far",
"overlap",
"location",
"likely",
"locations"
],
[
"composed",
"style",
"consists",
"squares",
"shape",
"scales",
"layers",
"alternating",
"frame",
"stacked"
],
[
"material",
"feed",
"caused",
"coarse",
"materials",
"additive",
"remove",
"mixture",
"filter",
"flow"
],
[
"constrain",
"importantly",
"realized",
"learns",
"forgetting",
"discovering",
"discover",
"unseen",
"aware",
"wasserstein"
],
[
"algorithms",
"algorithm",
"iteratively",
"polynomial",
"heuristics",
"help",
"implementations",
"compute",
"reconstruct",
"recursive"
]
] | 1.261012 | average_word_embeddings_glove.6B.300d | 0.975 | -0.363008 | 0.204249 | 0.902059 |
42 | ArXiv ML Papers | S³_angular | 50 | [
[
"dual",
"stationary",
"multi",
"facilitating",
"providing",
"proportional",
"distribution",
"non",
"soft",
"direct"
],
[
"identify",
"affected",
"developers",
"fail",
"generalize",
"communities",
"integrate",
"downstream",
"classify",
"operate"
],
[
"applicability",
"applicable",
"predictors",
"validated",
"intermediate",
"applies",
"vae",
"outlier",
"learnable",
"pass"
],
[
"initialization",
"leverage",
"expectation",
"probabilities",
"attain",
"fold",
"demonstrates",
"python",
"exponentially",
"leveraging"
],
[
"considerable",
"mining",
"activities",
"elements",
"element",
"drawn",
"numerical",
"superiority",
"actions",
"generated"
],
[
"framework",
"basis",
"elements",
"solution",
"forms",
"phases",
"principles",
"possible",
"present",
"yield"
],
[
"share",
"weaker",
"price",
"market",
"gains",
"exchange",
"leveraged",
"derivatives",
"auto",
"financial"
],
[
"languages",
"language",
"english",
"word",
"meaning",
"linguistic",
"translate",
"words",
"dictionary",
"translation"
],
[
"express",
"gpus",
"gps",
"recognition",
"expressed",
"respect",
"recognizing",
"temporal",
"extending",
"distance"
],
[
"verification",
"process",
"processes",
"continuous",
"inspection",
"processing",
"automated",
"stochastic",
"semi",
"monitoring"
],
[
"measurement",
"association",
"universal",
"collect",
"color",
"eeg",
"applies",
"criteria",
"preference",
"patterns"
],
[
"preference",
"block",
"discrimination",
"transferred",
"involving",
"taken",
"involve",
"transferring",
"class",
"involved"
],
[
"uses",
"distinguishing",
"characteristic",
"attributes",
"particular",
"characteristics",
"techniques",
"certain",
"good",
"combination"
],
[
"signal",
"signals",
"acoustic",
"audio",
"message",
"voice",
"speech",
"speaker",
"addressing",
"sensing"
],
[
"stored",
"data",
"unstructured",
"databases",
"evidence",
"database",
"static",
"compressed",
"single",
"storage"
],
[
"actual",
"person",
"similarity",
"allows",
"reasonable",
"comparable",
"modal",
"ability",
"simpler",
"identical"
],
[
"regression",
"patch",
"supports",
"mapping",
"survey",
"root",
"partitioning",
"showing",
"recovery",
"smoothing"
],
[
"rules",
"modified",
"limits",
"generalization",
"modification",
"modifications",
"synthetic",
"extensions",
"definition",
"generic"
],
[
"generating",
"transformers",
"solely",
"binary",
"purposes",
"power",
"consumption",
"instead",
"non",
"energy"
],
[
"optimized",
"scalable",
"hardware",
"gpu",
"server",
"platforms",
"architectures",
"implementations",
"scalability",
"gpus"
],
[
"utilization",
"enhance",
"beneficial",
"enhancement",
"augment",
"enhancing",
"boosting",
"rich",
"efficient",
"promising"
],
[
"paradigm",
"converge",
"distinct",
"coherent",
"shape",
"directions",
"direction",
"unified",
"closely",
"broadly"
],
[
"enforce",
"code",
"rule",
"rules",
"imposes",
"codes",
"compliance",
"absolute",
"labeling",
"strict"
],
[
"fairness",
"diversity",
"individual",
"regardless",
"gender",
"fundamental",
"equal",
"respect",
"balance",
"perturbations"
],
[
"evolution",
"defines",
"change",
"defining",
"taxonomy",
"changing",
"current",
"protocol",
"classification",
"evolving"
],
[
"classical",
"view",
"interpretation",
"art",
"newton",
"overhead",
"deviation",
"perspective",
"aware",
"maximization"
],
[
"distortion",
"negative",
"difference",
"perception",
"bias",
"tradeoff",
"reasoning",
"biases",
"cognitive",
"hypothesis"
],
[
"mcmc",
"easy",
"comparisons",
"costly",
"hard",
"expensive",
"labeling",
"labels",
"easier",
"way"
],
[
"investigation",
"retrieval",
"analyses",
"analysis",
"verification",
"assessment",
"technical",
"evaluation",
"module",
"augmentation"
],
[
"internet",
"online",
"mobile",
"phenomena",
"sgd",
"wireless",
"videos",
"99",
"unseen",
"interactive"
],
[
"coherent",
"semantically",
"convexity",
"magnitude",
"visual",
"characterization",
"gan",
"incomplete",
"elements",
"assessment"
],
[
"approximation",
"background",
"overview",
"count",
"abstraction",
"sketch",
"consistency",
"epsilon",
"usually",
"indicate"
],
[
"available",
"maps",
"practical",
"distances",
"navigation",
"applicability",
"divergence",
"kl",
"beneficial",
"helpful"
],
[
"variance",
"sampling",
"probability",
"sample",
"random",
"beta",
"bernoulli",
"error",
"distributions",
"deviation"
],
[
"match",
"play",
"players",
"games",
"game",
"matches",
"super",
"plays",
"scoring",
"best"
],
[
"graphs",
"vertices",
"graph",
"vertex",
"edges",
"node",
"nodes",
"embedding",
"undesirable",
"embeddings"
],
[
"solely",
"interventions",
"action",
"potential",
"impact",
"strategy",
"successful",
"significant",
"feasible",
"consequence"
],
[
"notably",
"mainly",
"region",
"origin",
"previously",
"captured",
"classic",
"distinct",
"bandits",
"superiority"
],
[
"ranking",
"contrast",
"highest",
"highly",
"comparison",
"compared",
"intelligent",
"superior",
"far",
"margin"
],
[
"network",
"channels",
"neural",
"programming",
"backpropagation",
"networks",
"channel",
"connections",
"cnn",
"cells"
],
[
"issue",
"question",
"documents",
"regarding",
"confirm",
"revisit",
"unclear",
"discussed",
"release",
"reasons"
],
[
"threat",
"targeting",
"detect",
"detecting",
"detection",
"malware",
"detector",
"vulnerability",
"malicious",
"identification"
],
[
"requires",
"category",
"comes",
"given",
"kind",
"brings",
"grows",
"event",
"takes",
"arises"
],
[
"hessian",
"matrix",
"vector",
"tensors",
"tensor",
"symmetric",
"vectors",
"matrices",
"convolution",
"gpu"
],
[
"solving",
"tasks",
"task",
"obstacles",
"integrating",
"difficulty",
"challenges",
"involves",
"capturing",
"crucial"
],
[
"proving",
"believe",
"belief",
"confidence",
"prove",
"superiority",
"motivation",
"claim",
"implicit",
"proof"
],
[
"inputs",
"quantization",
"perceptual",
"invariance",
"conditioned",
"selection",
"max",
"shed",
"gained",
"select"
],
[
"rank",
"actually",
"uniform",
"achieved",
"appear",
"furthermore",
"rarely",
"identical",
"consequently",
"simultaneously"
],
[
"scores",
"current",
"points",
"textual",
"scoring",
"outputs",
"score",
"variable",
"primary",
"transformer"
],
[
"ascent",
"accelerated",
"rate",
"rapid",
"descent",
"gradient",
"rates",
"gradients",
"curve",
"convergence"
]
] | 1.450473 | average_word_embeddings_glove.6B.300d | 0.954 | -0.361875 | 0.209012 | 0.904924 |
42 | ArXiv ML Papers | S³_combined | 10 | [
[
"expression",
"similarities",
"phenomena",
"relationship",
"influence",
"distinct",
"belief",
"nature",
"understanding",
"differences"
],
[
"sensor",
"sensors",
"audio",
"digital",
"camera",
"equipped",
"video",
"robot",
"optical",
"devices"
],
[
"vertices",
"graph",
"topological",
"symmetric",
"vertex",
"graphs",
"orthogonal",
"topology",
"tensor",
"nodes"
],
[
"forecast",
"models",
"model",
"forecasts",
"forecasting",
"projections",
"prediction",
"growth",
"statistical",
"predictions"
],
[
"geometry",
"nonlinear",
"technique",
"method",
"techniques",
"quantum",
"equations",
"physics",
"synthesis",
"linear"
],
[
"science",
"engineering",
"languages",
"published",
"biology",
"literature",
"library",
"research",
"foundation",
"physics"
],
[
"time",
"number",
"points",
"times",
"10",
"high",
"rate",
"point",
"half",
"small"
],
[
"simplicity",
"mode",
"style",
"compact",
"version",
"platform",
"graphical",
"image",
"sound",
"video"
],
[
"vs",
"82",
"91",
"min",
"71",
"88",
"rl",
"94",
"83",
"averaged"
],
[
"infrastructure",
"stability",
"accelerate",
"alleviate",
"extending",
"expansion",
"forces",
"energy",
"transition",
"operations"
]
] | 0.926694 | average_word_embeddings_glove.6B.300d | 0.98 | -0.290118 | 0.266428 | 0.910757 |
42 | ArXiv ML Papers | S³_combined | 20 | [
[
"depending",
"vary",
"sizes",
"settings",
"selecting",
"different",
"pairwise",
"combinations",
"predefined",
"desirable"
],
[
"catastrophic",
"forecasts",
"forecast",
"weather",
"traffic",
"forecasting",
"predicting",
"worst",
"uncertainties",
"flows"
],
[
"prevent",
"bandit",
"forces",
"agents",
"deployed",
"attacks",
"attacker",
"armed",
"enforce",
"attack"
],
[
"gps",
"controller",
"parameter",
"module",
"parameters",
"calibration",
"evaluation",
"configuration",
"classifier",
"test"
],
[
"polynomial",
"equation",
"sum",
"integer",
"zero",
"right",
"factor",
"sentence",
"cause",
"left"
],
[
"network",
"neurons",
"networks",
"neural",
"connections",
"channels",
"channel",
"wireless",
"backpropagation",
"activation"
],
[
"losses",
"problems",
"difficulties",
"suffer",
"performance",
"lack",
"poor",
"loss",
"training",
"worse"
],
[
"training",
"centers",
"branch",
"priori",
"communities",
"classes",
"families",
"classification",
"population",
"operate"
],
[
"policy",
"framework",
"stability",
"relations",
"implementation",
"implementing",
"compliance",
"comprehensive",
"policies",
"normalization"
],
[
"correct",
"algorithm",
"target",
"goal",
"consensus",
"conclusion",
"date",
"scenario",
"prediction",
"strategy"
],
[
"brain",
"cancer",
"disease",
"diseases",
"patients",
"protein",
"clinical",
"cells",
"genetic",
"diagnosis"
],
[
"nodes",
"graphs",
"nonconvex",
"graph",
"vertices",
"node",
"convex",
"edges",
"vertex",
"topological"
],
[
"consisting",
"functions",
"unit",
"joint",
"special",
"consists",
"named",
"units",
"introduced",
"new"
],
[
"paper",
"textual",
"statistical",
"derived",
"estimate",
"layer",
"compression",
"estimates",
"usage",
"standard"
],
[
"quality",
"technology",
"excellent",
"competitive",
"efficient",
"low",
"efficiency",
"imaging",
"maximize",
"enhancing"
],
[
"signals",
"https",
"gradient",
"relaxation",
"filtering",
"signal",
"facilitate",
"frequencies",
"denoising",
"subspace"
],
[
"biased",
"positive",
"observations",
"negative",
"conducted",
"evidence",
"bias",
"relationship",
"results",
"views"
],
[
"web",
"twitter",
"online",
"internet",
"server",
"users",
"user",
"website",
"databases",
"google"
],
[
"iterative",
"nonlinear",
"variational",
"multivariate",
"optimization",
"stochastic",
"dirichlet",
"minimization",
"deterministic",
"optimisation"
],
[
"behaviors",
"minimize",
"variables",
"processes",
"factors",
"mechanisms",
"specific",
"inherent",
"sensitive",
"risks"
]
] | 1.04701 | average_word_embeddings_glove.6B.300d | 0.995 | -0.28966 | 0.243583 | 0.885332 |
42 | ArXiv ML Papers | S³_combined | 30 | [
[
"19",
"preliminary",
"2019",
"2018",
"28",
"2017",
"14",
"13",
"16",
"autoregressive"
],
[
"constraints",
"ensemble",
"demands",
"elements",
"initialization",
"dimension",
"compositional",
"element",
"architectural",
"interpretation"
],
[
"false",
"valid",
"answer",
"proof",
"answers",
"explanation",
"explanations",
"documents",
"questions",
"details"
],
[
"city",
"rank",
"regions",
"region",
"highest",
"prominent",
"ranking",
"areas",
"area",
"gained"
],
[
"science",
"scientific",
"research",
"survey",
"studies",
"topics",
"literature",
"biology",
"physics",
"studied"
],
[
"interesting",
"combinations",
"mixture",
"music",
"videos",
"sampled",
"random",
"sample",
"samples",
"acoustic"
],
[
"style",
"design",
"architectural",
"designs",
"modifications",
"introducing",
"formulation",
"techniques",
"series",
"geometric"
],
[
"channels",
"networks",
"backpropagation",
"connections",
"network",
"neural",
"channel",
"convolutional",
"programming",
"neurons"
],
[
"proportional",
"candidates",
"resolution",
"responses",
"process",
"parallel",
"measured",
"coordinate",
"tasks",
"simultaneous"
],
[
"market",
"consumption",
"outperform",
"price",
"rates",
"share",
"leveraged",
"trend",
"weaker",
"yields"
],
[
"estimator",
"regression",
"bayesian",
"iterative",
"stochastic",
"bayes",
"multivariate",
"inference",
"parametric",
"probabilistic"
],
[
"conditioning",
"practice",
"overhead",
"poor",
"trains",
"safety",
"driving",
"fine",
"manual",
"unsupervised"
],
[
"traffic",
"forecast",
"prediction",
"weather",
"maps",
"forecasting",
"forecasts",
"flows",
"predicting",
"predictions"
],
[
"molecular",
"detector",
"particle",
"physics",
"chemical",
"magnetic",
"energy",
"quantum",
"transformer",
"generator"
],
[
"width",
"posterior",
"length",
"parameter",
"density",
"variance",
"approximate",
"maximum",
"bounds",
"square"
],
[
"success",
"production",
"stability",
"improvement",
"role",
"performs",
"performances",
"performance",
"successful",
"excellent"
],
[
"efficacy",
"conditional",
"performances",
"uniformly",
"positive",
"observation",
"material",
"vs",
"false",
"exhibits"
],
[
"nets",
"temporal",
"inefficient",
"relying",
"agent",
"modes",
"vary",
"sources",
"averaging",
"methods"
],
[
"remarkable",
"extracting",
"confidence",
"deeper",
"deep",
"tremendous",
"gains",
"yields",
"impressive",
"enormous"
],
[
"functionality",
"interface",
"software",
"server",
"hardware",
"platforms",
"user",
"computing",
"kernel",
"implementations"
],
[
"version",
"models",
"modeling",
"model",
"car",
"fit",
"built",
"compact",
"fashion",
"designs"
],
[
"languages",
"words",
"english",
"language",
"word",
"translation",
"texts",
"speech",
"text",
"dictionary"
],
[
"attacks",
"attack",
"targeting",
"forces",
"malicious",
"attacker",
"malware",
"threat",
"armed",
"targeted"
],
[
"arising",
"occur",
"consequences",
"uncertainties",
"catastrophic",
"events",
"severe",
"suffers",
"caused",
"effects"
],
[
"population",
"optimal",
"genetic",
"optimum",
"desirable",
"variability",
"location",
"randomly",
"locations",
"sampling"
],
[
"intermediate",
"classes",
"requirements",
"predictors",
"auxiliary",
"classification",
"vary",
"languages",
"consisting",
"consists"
],
[
"quantum",
"algorithms",
"computation",
"computational",
"cognitive",
"mathematical",
"computer",
"fourier",
"asymptotic",
"physics"
],
[
"computationally",
"perform",
"pairwise",
"approximations",
"performances",
"ensembles",
"various",
"numerous",
"tasks",
"datasets"
],
[
"compression",
"throughput",
"substantial",
"costs",
"extraction",
"processing",
"widespread",
"increased",
"reduction",
"considerable"
],
[
"construction",
"block",
"process",
"transfer",
"extension",
"allow",
"prevent",
"thousands",
"closed",
"aid"
]
] | 1.178767 | average_word_embeddings_glove.6B.300d | 0.956667 | -0.331822 | 0.226281 | 0.899027 |
42 | ArXiv ML Papers | S³_combined | 40 | [
[
"estimators",
"remarkably",
"explainable",
"scalability",
"impressive",
"improvements",
"surprisingly",
"margin",
"accurate",
"grained"
],
[
"nets",
"speaker",
"sources",
"inefficient",
"unknown",
"temporal",
"delta",
"vary",
"deep",
"encoding"
],
[
"mcmc",
"exploitation",
"classification",
"dependency",
"pooling",
"method",
"technique",
"monte",
"clustering",
"formalize"
],
[
"viable",
"alternatives",
"alternative",
"proven",
"feasible",
"reliable",
"labels",
"reach",
"prove",
"producing"
],
[
"overfitting",
"supervised",
"modeled",
"gans",
"learnable",
"investigates",
"updating",
"annotated",
"denoising",
"adapts"
],
[
"forces",
"targeting",
"armed",
"malware",
"malicious",
"attack",
"threat",
"attacks",
"attacker",
"targeted"
],
[
"engine",
"models",
"modeling",
"autoregressive",
"model",
"mini",
"car",
"surrogate",
"hybrid",
"built"
],
[
"class",
"mri",
"obtaining",
"procedure",
"inspection",
"recognition",
"superior",
"requiring",
"identification",
"resonance"
],
[
"server",
"software",
"mobile",
"interface",
"developers",
"compatible",
"implementations",
"functionality",
"applications",
"computing"
],
[
"defense",
"space",
"wise",
"storage",
"valued",
"valuable",
"metrics",
"utility",
"quantized",
"robot"
],
[
"molecular",
"cells",
"genetic",
"encoded",
"encodes",
"protein",
"synthesis",
"activation",
"domains",
"neurons"
],
[
"activations",
"formalize",
"kernel",
"normalization",
"normalizing",
"vanilla",
"conditional",
"covariance",
"supported",
"hypothesis"
],
[
"traffic",
"transportation",
"transport",
"train",
"nearest",
"trains",
"vehicles",
"car",
"bridge",
"navigation"
],
[
"classifier",
"intelligent",
"prone",
"surprisingly",
"compact",
"reliable",
"extremely",
"strong",
"proved",
"noisy"
],
[
"exploration",
"extension",
"space",
"short",
"sequences",
"extended",
"search",
"operations",
"training",
"discovery"
],
[
"distances",
"dataset",
"runtime",
"locations",
"configurations",
"tests",
"suitable",
"implementations",
"variability",
"densities"
],
[
"programming",
"convolutional",
"backpropagation",
"network",
"networks",
"cnn",
"neural",
"neurons",
"cells",
"propagation"
],
[
"mathematical",
"age",
"music",
"quantum",
"classical",
"theory",
"modern",
"literature",
"physics",
"phenomena"
],
[
"guarantees",
"necessary",
"condition",
"obtain",
"guarantee",
"provided",
"compliance",
"stability",
"provide",
"verify"
],
[
"policy",
"policies",
"formulate",
"adopt",
"flexible",
"objectives",
"coordinate",
"optimal",
"aims",
"implement"
],
[
"objects",
"2015",
"pruning",
"2017",
"magnitude",
"classification",
"2018",
"2019",
"object",
"taxonomy"
],
[
"survey",
"explore",
"analyze",
"analyzed",
"observe",
"closely",
"experts",
"conducted",
"examine",
"detected"
],
[
"users",
"twitter",
"personalized",
"feedback",
"online",
"user",
"responses",
"recommendations",
"messages",
"offline"
],
[
"usage",
"pruning",
"infrastructure",
"inefficient",
"incremental",
"construction",
"substantially",
"extraction",
"building",
"augmentation"
],
[
"image",
"3d",
"pixels",
"vision",
"eye",
"images",
"color",
"pixel",
"landscape",
"camera"
],
[
"features",
"feature",
"videos",
"highlights",
"identifying",
"privacy",
"differentiable",
"character",
"includes",
"profile"
],
[
"foundation",
"scientific",
"engineering",
"center",
"biology",
"science",
"research",
"studies",
"physics",
"papers"
],
[
"codes",
"classes",
"structural",
"changes",
"fail",
"requirements",
"class",
"code",
"detect",
"accelerate"
],
[
"translation",
"word",
"text",
"words",
"languages",
"english",
"dictionary",
"language",
"texts",
"meaning"
],
[
"codes",
"vehicles",
"sensors",
"labels",
"machines",
"items",
"automatic",
"boxes",
"devices",
"contrastive"
],
[
"rl",
"coverage",
"compression",
"provable",
"textual",
"codes",
"entropy",
"sparsity",
"bias",
"distributional"
],
[
"improvements",
"max",
"design",
"critic",
"architectural",
"efficiency",
"aspects",
"characteristics",
"classical",
"designing"
],
[
"cancer",
"disease",
"patient",
"patients",
"diseases",
"diagnosis",
"treat",
"clinical",
"heart",
"medical"
],
[
"learnt",
"learn",
"classes",
"training",
"trained",
"gps",
"guidance",
"evaluations",
"calibration",
"unbiased"
],
[
"neighboring",
"areas",
"families",
"countries",
"diverse",
"groups",
"locally",
"regions",
"communities",
"counterparts"
],
[
"area",
"population",
"location",
"date",
"range",
"neighborhood",
"reasonable",
"target",
"locations",
"remote"
],
[
"composed",
"layers",
"style",
"layer",
"mixture",
"shape",
"upper",
"squares",
"consists",
"dense"
],
[
"coarse",
"additive",
"remove",
"mixture",
"materials",
"feed",
"material",
"caused",
"temperature",
"chemical"
],
[
"constrain",
"discover",
"discovering",
"wasserstein",
"unseen",
"learns",
"bottleneck",
"forgetting",
"finds",
"revisit"
],
[
"algorithms",
"iteratively",
"polynomial",
"algorithm",
"reconstruct",
"heuristics",
"implementations",
"memory",
"help",
"compute"
]
] | 1.224885 | average_word_embeddings_glove.6B.300d | 0.94 | -0.344065 | 0.209929 | 0.898738 |
42 | ArXiv ML Papers | S³_combined | 50 | [
[
"soft",
"stationary",
"distribution",
"margin",
"dual",
"proportional",
"providing",
"boundary",
"facilitating",
"continuous"
],
[
"identify",
"regions",
"countries",
"detect",
"developers",
"discriminate",
"latent",
"downstream",
"integrate",
"communities"
],
[
"outlier",
"intermediate",
"learnable",
"benchmark",
"predictors",
"bounds",
"applicability",
"threshold",
"criteria",
"ensembles"
],
[
"outperforming",
"latent",
"initialization",
"leverage",
"python",
"vanilla",
"attain",
"exploit",
"achieve",
"probabilities"
],
[
"architectural",
"mining",
"considerable",
"activities",
"superiority",
"construction",
"numerical",
"elements",
"coarse",
"huge"
],
[
"elements",
"scenarios",
"basis",
"solutions",
"solution",
"yield",
"samples",
"modalities",
"framework",
"feasibility"
],
[
"share",
"exchange",
"derivatives",
"leveraged",
"price",
"federated",
"weaker",
"benchmark",
"gains",
"market"
],
[
"language",
"words",
"translation",
"english",
"languages",
"word",
"meaning",
"translate",
"linguistic",
"dictionary"
],
[
"expressed",
"recognition",
"express",
"gps",
"gpus",
"temporal",
"relations",
"respect",
"normalized",
"distance"
],
[
"processes",
"process",
"automated",
"monitoring",
"stochastic",
"gaussian",
"inspection",
"continuous",
"markov",
"processing"
],
[
"association",
"color",
"universal",
"measurement",
"criteria",
"collect",
"architectural",
"patterns",
"eeg",
"shelf"
],
[
"mri",
"discrimination",
"class",
"posterior",
"transferred",
"preference",
"classes",
"facial",
"involving",
"superior"
],
[
"attributes",
"uses",
"distinguishing",
"characteristic",
"techniques",
"characteristics",
"strengths",
"quality",
"simplicity",
"methods"
],
[
"speech",
"signal",
"signals",
"speaker",
"audio",
"addressing",
"message",
"acoustic",
"voice",
"noise"
],
[
"stored",
"evidence",
"data",
"unstructured",
"database",
"compressed",
"databases",
"storage",
"information",
"lightweight"
],
[
"conditional",
"reasonable",
"similarity",
"accurately",
"modal",
"person",
"classifier",
"simpler",
"partial",
"execute"
],
[
"logistic",
"regression",
"patch",
"survey",
"tree",
"root",
"recovery",
"supports",
"vanilla",
"partitioning"
],
[
"modified",
"modifications",
"rules",
"synthetic",
"bounds",
"limits",
"modification",
"generalization",
"generic",
"extensions"
],
[
"power",
"binary",
"generating",
"solely",
"consumption",
"energy",
"pruning",
"transformers",
"randomly",
"alternating"
],
[
"server",
"gpu",
"scalable",
"scalability",
"platforms",
"atari",
"architectures",
"hardware",
"implementations",
"gpus"
],
[
"enhance",
"utilization",
"enhancing",
"efficient",
"rich",
"rl",
"enhancement",
"augment",
"efficiency",
"boosting"
],
[
"formulate",
"unified",
"directions",
"agnostic",
"broadly",
"shape",
"coherent",
"converge",
"paradigm",
"direction"
],
[
"codes",
"labeling",
"absolute",
"strict",
"safety",
"rules",
"code",
"imposes",
"privacy",
"compliance"
],
[
"perturbations",
"individual",
"respect",
"gender",
"fundamental",
"fairness",
"denoising",
"diversity",
"equal",
"challenges"
],
[
"classification",
"defines",
"et",
"protocol",
"entropy",
"evolving",
"changing",
"taxonomy",
"defining",
"change"
],
[
"classical",
"overhead",
"view",
"art",
"maximization",
"deviation",
"settings",
"dictionary",
"definition",
"music"
],
[
"hypothesis",
"biases",
"negative",
"distortion",
"reasoning",
"perception",
"bias",
"cognitive",
"poisoning",
"perceptual"
],
[
"labels",
"costly",
"accurate",
"pairwise",
"expensive",
"reliable",
"mcmc",
"distances",
"convolutional",
"labeling"
],
[
"training",
"investigation",
"evaluation",
"verification",
"retrieval",
"inspection",
"analysis",
"insight",
"intelligence",
"module"
],
[
"internet",
"online",
"mobile",
"videos",
"wireless",
"interactive",
"phenomena",
"dirichlet",
"google",
"streaming"
],
[
"coding",
"aggregating",
"malicious",
"visual",
"magnitude",
"coherent",
"similarity",
"results",
"errors",
"sequences"
],
[
"sketch",
"results",
"approximation",
"correction",
"overview",
"consistency",
"epsilon",
"statistical",
"abstraction",
"autoregressive"
],
[
"navigation",
"kl",
"distances",
"ml",
"applicability",
"available",
"practical",
"logistic",
"distribution",
"sizes"
],
[
"sampling",
"variance",
"random",
"beta",
"sample",
"bernoulli",
"gaussian",
"error",
"estimator",
"distributions"
],
[
"super",
"scoring",
"matches",
"match",
"games",
"play",
"players",
"game",
"goals",
"plays"
],
[
"embeddings",
"edges",
"graph",
"graphs",
"benchmark",
"nodes",
"node",
"vertex",
"vertices",
"embedding"
],
[
"strategy",
"interventions",
"outcomes",
"successful",
"hypothesis",
"statistically",
"feasible",
"effects",
"programming",
"programs"
],
[
"regions",
"region",
"notably",
"origin",
"classic",
"superiority",
"central",
"captured",
"mainly",
"logistic"
],
[
"compared",
"margin",
"favorably",
"superior",
"ranking",
"highest",
"highly",
"intelligent",
"biased",
"weighted"
],
[
"channels",
"backpropagation",
"programming",
"networks",
"neural",
"network",
"channel",
"convolutional",
"cnn",
"cells"
],
[
"issue",
"intractable",
"revisit",
"question",
"confirm",
"details",
"documents",
"release",
"discussed",
"regret"
],
[
"malicious",
"detector",
"detection",
"detecting",
"detect",
"malware",
"threat",
"targeting",
"identification",
"vulnerability"
],
[
"comes",
"event",
"mild",
"occurs",
"severe",
"brings",
"grows",
"category",
"magnitude",
"phase"
],
[
"vectors",
"vector",
"tensor",
"hessian",
"tensors",
"algebra",
"matrices",
"symmetric",
"rank",
"gradient"
],
[
"obstacles",
"task",
"solving",
"tasks",
"challenges",
"difficulty",
"capturing",
"goals",
"crucial",
"involves"
],
[
"usefulness",
"empirically",
"proving",
"belief",
"reliable",
"superiority",
"proof",
"confidence",
"prove",
"predictive"
],
[
"inputs",
"pruning",
"perceptual",
"trees",
"invariance",
"quantization",
"selection",
"shed",
"label",
"rl"
],
[
"rank",
"pattern",
"achieve",
"achieved",
"uniform",
"papers",
"paper",
"frequency",
"appear",
"identical"
],
[
"outputs",
"scoring",
"goals",
"score",
"points",
"textual",
"scores",
"output",
"transformer",
"variable"
],
[
"rate",
"descent",
"gradient",
"rates",
"accelerated",
"ascent",
"rapid",
"gradients",
"convergence",
"stochastic"
]
] | 1.334867 | average_word_embeddings_glove.6B.300d | 0.906 | -0.34613 | 0.198195 | 0.89588 |
42 | BBC News | GMM | 10 | [
[
"prices",
"rates",
"economy",
"growth",
"rise",
"dollar",
"rate",
"economic",
"spending",
"deficit"
],
[
"oscar",
"awards",
"film",
"actor",
"award",
"director",
"comedy",
"actress",
"best",
"aviator"
],
[
"liverpool",
"club",
"wales",
"coach",
"side",
"chelsea",
"rugby",
"england",
"ireland",
"players"
],
[
"brown",
"prime",
"tory",
"party",
"howard",
"election",
"minister",
"labour",
"blair",
"leader"
],
[
"users",
"net",
"security",
"software",
"search",
"virus",
"internet",
"site",
"microsoft",
"computer"
],
[
"yukos",
"shares",
"company",
"firm",
"deutsche",
"stock",
"oil",
"financial",
"deal",
"state"
],
[
"broadband",
"phones",
"technology",
"apple",
"sony",
"digital",
"mobile",
"tv",
"video",
"phone"
],
[
"olympic",
"championships",
"champion",
"60m",
"athletics",
"indoor",
"race",
"holmes",
"athens",
"gold"
],
[
"music",
"band",
"album",
"rock",
"hop",
"hip",
"singer",
"song",
"chart",
"artists"
],
[
"open",
"australian",
"roddick",
"seed",
"match",
"round",
"henman",
"federer",
"beat",
"champion"
]
] | 1.504865 | average_word_embeddings_glove.6B.300d | 0.99 | 0.110742 | 0.225802 | 0.933737 |
42 | BBC News | GMM | 20 | [
[
"oil",
"countries",
"economy",
"economic",
"budget",
"china",
"india",
"debt",
"president",
"bush"
],
[
"actor",
"film",
"award",
"actress",
"oscar",
"director",
"awards",
"best",
"aviator",
"comedy"
],
[
"ireland",
"italy",
"capt",
"france",
"edinburgh",
"try",
"penalty",
"wasps",
"wing",
"gara"
],
[
"blair",
"party",
"howard",
"labour",
"election",
"prime",
"tory",
"leader",
"brown",
"chancellor"
],
[
"sites",
"security",
"virus",
"software",
"search",
"users",
"microsoft",
"spam",
"site",
"programs"
],
[
"doping",
"drugs",
"trial",
"athens",
"charged",
"banned",
"arrested",
"court",
"charges",
"greek"
],
[
"online",
"net",
"phone",
"technology",
"bt",
"broadband",
"mobile",
"service",
"data",
"internet"
],
[
"host",
"africa",
"italy",
"japan",
"final",
"winners",
"cup",
"august",
"republic",
"rugby"
],
[
"50",
"cent",
"box",
"show",
"films",
"series",
"film",
"festival",
"office",
"million"
],
[
"seed",
"henman",
"match",
"roddick",
"australian",
"round",
"open",
"federer",
"beat",
"champion"
],
[
"album",
"rock",
"robbie",
"band",
"pop",
"singer",
"song",
"chart",
"songs",
"single"
],
[
"rights",
"human",
"police",
"law",
"lord",
"laws",
"terror",
"bill",
"lords",
"parliament"
],
[
"indoor",
"championships",
"champion",
"race",
"60m",
"olympic",
"holmes",
"gold",
"jump",
"seconds"
],
[
"prices",
"economy",
"rates",
"dollar",
"growth",
"rise",
"rate",
"sales",
"economic",
"figures"
],
[
"minutes",
"chelsea",
"everton",
"scored",
"united",
"liverpool",
"ball",
"newcastle",
"goal",
"striker"
],
[
"dvd",
"apple",
"digital",
"sony",
"mac",
"gadget",
"mini",
"games",
"gadgets",
"mobile"
],
[
"club",
"rugby",
"coach",
"england",
"players",
"wales",
"cup",
"team",
"liverpool",
"league"
],
[
"deutsche",
"shares",
"company",
"yukos",
"firm",
"shareholders",
"bid",
"stock",
"exchange",
"deal"
],
[
"hip",
"hop",
"game",
"mtv",
"spanish",
"music",
"black",
"language",
"halo",
"stone"
],
[
"workers",
"council",
"000",
"unions",
"local",
"tax",
"minimum",
"pay",
"pension",
"increase"
]
] | 1.856538 | average_word_embeddings_glove.6B.300d | 0.955 | 0.047522 | 0.203056 | 0.937627 |
42 | BBC News | GMM | 30 | [
[
"budget",
"eu",
"economy",
"poverty",
"deficit",
"debt",
"countries",
"bush",
"economic",
"imf"
],
[
"film",
"actress",
"award",
"oscar",
"best",
"actor",
"awards",
"aviator",
"nominated",
"director"
],
[
"robinson",
"lions",
"nations",
"rugby",
"england",
"coach",
"wales",
"six",
"ireland",
"squad"
],
[
"net",
"speeds",
"broadband",
"3g",
"calls",
"mobile",
"bt",
"users",
"service",
"networks"
],
[
"human",
"terror",
"rights",
"lords",
"lord",
"law",
"police",
"id",
"hunting",
"clarke"
],
[
"australian",
"roddick",
"match",
"henman",
"federer",
"beat",
"seed",
"serve",
"final",
"open"
],
[
"euros",
"profits",
"sales",
"shares",
"company",
"firm",
"profit",
"deal",
"market",
"stock"
],
[
"music",
"band",
"song",
"album",
"hop",
"chart",
"rock",
"stone",
"hip",
"rap"
],
[
"sky",
"broadcast",
"mtv",
"television",
"channel",
"tv",
"broadcaster",
"stations",
"viewers",
"satellite"
],
[
"shareholders",
"exchange",
"bid",
"takeover",
"deutsche",
"stock",
"offer",
"talks",
"london",
"frankfurt"
],
[
"indoor",
"race",
"holmes",
"gold",
"champion",
"olympic",
"60m",
"championships",
"jump",
"seconds"
],
[
"growth",
"rates",
"economy",
"prices",
"rate",
"economic",
"exports",
"dollar",
"figures",
"bank"
],
[
"election",
"labour",
"tory",
"blair",
"howard",
"brown",
"party",
"prime",
"leader",
"chancellor"
],
[
"cup",
"nations",
"rugby",
"coach",
"lions",
"welsh",
"woodward",
"ireland",
"players",
"wales"
],
[
"league",
"manager",
"club",
"football",
"chelsea",
"liverpool",
"arsenal",
"champions",
"madrid",
"team"
],
[
"users",
"programs",
"virus",
"websites",
"microsoft",
"spam",
"security",
"windows",
"software",
"program"
],
[
"singer",
"brother",
"star",
"50",
"cent",
"her",
"show",
"arrested",
"she",
"police"
],
[
"bankruptcy",
"russian",
"khodorkovsky",
"gazprom",
"auction",
"fraud",
"yukos",
"sec",
"court",
"company"
],
[
"technology",
"net",
"information",
"use",
"site",
"firms",
"computer",
"data",
"attacks",
"using"
],
[
"doping",
"drugs",
"greek",
"athens",
"banned",
"tests",
"collins",
"test",
"olympics",
"athletes"
],
[
"xbox",
"tickets",
"sold",
"sales",
"films",
"potter",
"ticket",
"harry",
"san",
"halo"
],
[
"cup",
"tennis",
"henman",
"davis",
"open",
"wimbledon",
"murray",
"roddick",
"australian",
"round"
],
[
"box",
"film",
"series",
"films",
"comedy",
"book",
"lee",
"movie",
"story",
"office"
],
[
"minimum",
"public",
"children",
"tax",
"workers",
"local",
"council",
"pension",
"health",
"plans"
],
[
"search",
"programmes",
"tv",
"web",
"content",
"internet",
"video",
"google",
"visual",
"net"
],
[
"capt",
"italy",
"gara",
"edinburgh",
"wasps",
"penalty",
"try",
"ireland",
"murphy",
"half"
],
[
"mobiles",
"phone",
"mobile",
"phones",
"digital",
"cameras",
"music",
"tv",
"handsets",
"gartner"
],
[
"minutes",
"goal",
"injury",
"united",
"newcastle",
"chelsea",
"everton",
"liverpool",
"ball",
"side"
],
[
"mini",
"gadget",
"games",
"sony",
"dvd",
"apple",
"mac",
"gadgets",
"nintendo",
"gaming"
],
[
"airline",
"india",
"oil",
"china",
"gas",
"air",
"state",
"airlines",
"owned",
"company"
]
] | 2.355865 | average_word_embeddings_glove.6B.300d | 0.9 | 0.072226 | 0.200504 | 0.933844 |
42 | BBC News | GMM | 40 | [
[
"countries",
"budget",
"economic",
"eu",
"debt",
"economy",
"poverty",
"imf",
"spending",
"deficit"
],
[
"best",
"actress",
"oscar",
"film",
"awards",
"award",
"actor",
"aviator",
"nominated",
"prize"
],
[
"try",
"england",
"ireland",
"half",
"penalty",
"side",
"gara",
"hodgson",
"robinson",
"ball"
],
[
"service",
"calls",
"broadband",
"networks",
"bt",
"speeds",
"3g",
"mobile",
"net",
"users"
],
[
"police",
"silk",
"labour",
"id",
"party",
"vote",
"voting",
"cards",
"blair",
"commissioner"
],
[
"australian",
"roddick",
"match",
"henman",
"federer",
"beat",
"seed",
"serve",
"final",
"open"
],
[
"euros",
"sales",
"profits",
"shares",
"profit",
"stock",
"firm",
"deal",
"company",
"stake"
],
[
"album",
"song",
"elvis",
"chart",
"best",
"band",
"rock",
"robbie",
"songs",
"pop"
],
[
"mtv",
"sky",
"tv",
"programmes",
"channel",
"broadcast",
"television",
"satellite",
"channels",
"viewers"
],
[
"shareholders",
"exchange",
"bid",
"takeover",
"deutsche",
"stock",
"offer",
"talks",
"london",
"frankfurt"
],
[
"60m",
"seconds",
"champion",
"holmes",
"championships",
"olympic",
"indoor",
"gold",
"jump",
"race"
],
[
"housing",
"dollar",
"bank",
"rates",
"prices",
"market",
"mortgage",
"euro",
"lending",
"rise"
],
[
"labour",
"party",
"blair",
"leader",
"election",
"prime",
"howard",
"minister",
"brown",
"silk"
],
[
"coach",
"england",
"wales",
"nations",
"lions",
"rugby",
"squad",
"ireland",
"cup",
"welsh"
],
[
"chelsea",
"club",
"liverpool",
"manager",
"arsenal",
"league",
"champions",
"madrid",
"football",
"united"
],
[
"microsoft",
"security",
"search",
"spam",
"software",
"users",
"programs",
"virus",
"google",
"windows"
],
[
"star",
"her",
"singer",
"cent",
"50",
"brother",
"she",
"show",
"celebrity",
"rapper"
],
[
"yukos",
"bankruptcy",
"auction",
"gazprom",
"russian",
"russia",
"unit",
"court",
"khodorkovsky",
"putin"
],
[
"data",
"computer",
"attacks",
"use",
"information",
"net",
"internet",
"technology",
"users",
"site"
],
[
"doping",
"tests",
"athens",
"olympics",
"drugs",
"federation",
"greek",
"athletics",
"charges",
"athletes"
],
[
"films",
"xbox",
"san",
"sales",
"sold",
"harry",
"tickets",
"potter",
"halo",
"japan"
],
[
"davis",
"open",
"cup",
"australian",
"grand",
"henman",
"roddick",
"round",
"tennis",
"slam"
],
[
"box",
"film",
"comedy",
"book",
"films",
"actor",
"lee",
"office",
"festival",
"movie"
],
[
"committee",
"children",
"minimum",
"social",
"parents",
"health",
"care",
"increase",
"child",
"report"
],
[
"black",
"hop",
"hip",
"music",
"game",
"spanish",
"games",
"stone",
"halo",
"rap"
],
[
"edinburgh",
"glasgow",
"squad",
"leeds",
"wing",
"capt",
"newcastle",
"cardiff",
"rugby",
"wasps"
],
[
"phone",
"phones",
"mobile",
"digital",
"tv",
"music",
"mobiles",
"video",
"cameras",
"handsets"
],
[
"chelsea",
"striker",
"newcastle",
"united",
"minutes",
"goal",
"mourinho",
"ball",
"everton",
"injury"
],
[
"dvd",
"mac",
"nintendo",
"mini",
"gadget",
"apple",
"sony",
"gadgets",
"machine",
"gaming"
],
[
"banks",
"bank",
"financial",
"firms",
"sec",
"china",
"company",
"state",
"executives",
"foreign"
],
[
"drugs",
"doping",
"drug",
"fraud",
"sullivan",
"collins",
"jones",
"banned",
"accounting",
"white"
],
[
"taxes",
"labour",
"election",
"council",
"tory",
"tax",
"tories",
"howard",
"cuts",
"local"
],
[
"rugby",
"lions",
"football",
"tennis",
"players",
"team",
"woodward",
"cup",
"sport",
"tour"
],
[
"car",
"vehicles",
"auto",
"models",
"recall",
"cars",
"plant",
"production",
"engines",
"company"
],
[
"terror",
"court",
"human",
"lords",
"lord",
"law",
"rights",
"trial",
"police",
"hunt"
],
[
"virus",
"mail",
"spam",
"messages",
"mails",
"attacks",
"contains",
"security",
"warning",
"infected"
],
[
"rate",
"economic",
"quarter",
"economy",
"exports",
"figures",
"sales",
"growth",
"consumer",
"china"
],
[
"flight",
"costs",
"airlines",
"india",
"compensation",
"carriers",
"airline",
"travel",
"air",
"passengers"
],
[
"energy",
"gazprom",
"oil",
"india",
"gas",
"russian",
"russia",
"owned",
"yukos",
"company"
],
[
"squad",
"slam",
"wales",
"france",
"capt",
"saturday",
"wing",
"replacement",
"damien",
"grand"
]
] | 2.746569 | average_word_embeddings_glove.6B.300d | 0.8375 | 0.048686 | 0.202768 | 0.936229 |
42 | BBC News | GMM | 50 | [
[
"poverty",
"eu",
"economic",
"budget",
"countries",
"spending",
"bush",
"economy",
"deficit",
"subsidies"
],
[
"best",
"actress",
"oscar",
"film",
"awards",
"award",
"actor",
"aviator",
"nominated",
"prize"
],
[
"england",
"robinson",
"lions",
"rugby",
"nations",
"coach",
"woodward",
"wales",
"six",
"ireland"
],
[
"service",
"calls",
"broadband",
"networks",
"bt",
"speeds",
"3g",
"mobile",
"net",
"users"
],
[
"ban",
"hunting",
"id",
"law",
"police",
"hunt",
"cards",
"bill",
"plans",
"rules"
],
[
"australian",
"roddick",
"match",
"henman",
"federer",
"beat",
"seed",
"serve",
"final",
"open"
],
[
"profits",
"shares",
"euros",
"profit",
"stock",
"firm",
"company",
"sales",
"deal",
"stake"
],
[
"music",
"hop",
"band",
"hip",
"rock",
"album",
"song",
"chart",
"stone",
"robbie"
],
[
"sky",
"broadcast",
"mtv",
"television",
"channel",
"tv",
"broadcaster",
"stations",
"viewers",
"satellite"
],
[
"shareholders",
"exchange",
"bid",
"takeover",
"deutsche",
"stock",
"offer",
"talks",
"london",
"frankfurt"
],
[
"olympic",
"indoor",
"race",
"jump",
"60m",
"champion",
"seconds",
"championships",
"title",
"000m"
],
[
"rates",
"interest",
"bank",
"housing",
"slowdown",
"inflation",
"rise",
"rate",
"hold",
"manufacturing"
],
[
"lib",
"seats",
"tories",
"election",
"kennedy",
"party",
"conservatives",
"vote",
"labour",
"howard"
],
[
"rugby",
"wales",
"cup",
"side",
"ireland",
"coach",
"slam",
"nations",
"welsh",
"final"
],
[
"league",
"real",
"mourinho",
"madrid",
"arsenal",
"chelsea",
"champions",
"goal",
"play",
"liverpool"
],
[
"software",
"microsoft",
"security",
"virus",
"programs",
"windows",
"program",
"users",
"files",
"viruses"
],
[
"her",
"cent",
"brother",
"singer",
"star",
"show",
"50",
"she",
"rapper",
"celebrity"
],
[
"yukos",
"bankruptcy",
"auction",
"gazprom",
"russian",
"russia",
"unit",
"court",
"khodorkovsky",
"putin"
],
[
"power",
"use",
"net",
"technology",
"information",
"computer",
"data",
"attacks",
"using",
"project"
],
[
"again",
"compete",
"sprinter",
"olympics",
"suspension",
"testing",
"collins",
"kim",
"chambers",
"learning"
],
[
"films",
"sales",
"xbox",
"tickets",
"potter",
"sold",
"harry",
"japan",
"san",
"halo"
],
[
"open",
"tennis",
"cup",
"davis",
"australian",
"roddick",
"henman",
"wimbledon",
"round",
"grand"
],
[
"actor",
"film",
"book",
"office",
"series",
"lee",
"box",
"movie",
"comedy",
"films"
],
[
"report",
"companies",
"aids",
"law",
"social",
"human",
"research",
"rights",
"file",
"elections"
],
[
"programmes",
"content",
"digital",
"video",
"tv",
"radio",
"definition",
"music",
"media",
"service"
],
[
"leicester",
"rugby",
"simon",
"squad",
"glasgow",
"edinburgh",
"leeds",
"training",
"wasps",
"robinson"
],
[
"phone",
"tv",
"gartner",
"mobile",
"mobiles",
"phones",
"digital",
"handsets",
"cameras",
"camera"
],
[
"chelsea",
"mourinho",
"injury",
"liverpool",
"united",
"fa",
"cup",
"everton",
"arsenal",
"league"
],
[
"sony",
"mini",
"gadget",
"mac",
"apple",
"dvd",
"games",
"nintendo",
"gaming",
"gadgets"
],
[
"banks",
"bank",
"firms",
"state",
"investors",
"china",
"argentina",
"foreign",
"companies",
"financial"
],
[
"sullivan",
"company",
"sec",
"attorney",
"insurance",
"drug",
"charges",
"fraud",
"prosecutors",
"accounting"
],
[
"tax",
"election",
"taxes",
"tory",
"cuts",
"labour",
"howard",
"tories",
"chancellor",
"spending"
],
[
"athletics",
"athens",
"race",
"collins",
"holmes",
"championships",
"francis",
"she",
"relay",
"lewis"
],
[
"car",
"vehicles",
"auto",
"models",
"recall",
"cars",
"plant",
"production",
"engines",
"company"
],
[
"lord",
"trial",
"rights",
"suspects",
"human",
"terror",
"lords",
"murder",
"law",
"police"
],
[
"virus",
"mail",
"spam",
"messages",
"mails",
"attacks",
"contains",
"security",
"warning",
"infected"
],
[
"growth",
"consumer",
"economic",
"unemployment",
"economy",
"rate",
"spending",
"quarter",
"exports",
"manufacturing"
],
[
"flight",
"costs",
"airlines",
"india",
"compensation",
"carriers",
"airline",
"travel",
"air",
"passengers"
],
[
"gas",
"gazprom",
"oil",
"energy",
"india",
"russian",
"yukos",
"russia",
"owned",
"company"
],
[
"damien",
"replacement",
"slam",
"france",
"grand",
"wales",
"wing",
"squad",
"jean",
"ireland"
],
[
"athletes",
"athens",
"doping",
"greek",
"drugs",
"tests",
"olympics",
"banned",
"test",
"charges"
],
[
"exports",
"currency",
"dollar",
"china",
"euro",
"trade",
"deficit",
"economic",
"prices",
"growth"
],
[
"silk",
"leadership",
"prince",
"mp",
"leader",
"party",
"robert",
"him",
"lord",
"lords"
],
[
"retail",
"lending",
"sales",
"figures",
"december",
"prices",
"mortgage",
"house",
"rose",
"rise"
],
[
"children",
"pension",
"local",
"councils",
"workers",
"council",
"pay",
"minimum",
"increase",
"tax"
],
[
"labour",
"election",
"tony",
"brown",
"minister",
"prime",
"blunkett",
"blair",
"iraq",
"howard"
],
[
"club",
"liverpool",
"football",
"manager",
"contract",
"arsenal",
"smith",
"united",
"league",
"players"
],
[
"league",
"newcastle",
"drawn",
"chelsea",
"everton",
"premiership",
"united",
"face",
"derby",
"west"
],
[
"users",
"google",
"web",
"sites",
"site",
"search",
"websites",
"engine",
"spam",
"engines"
],
[
"half",
"gara",
"murphy",
"ball",
"goal",
"penalty",
"ireland",
"minute",
"try",
"minutes"
]
] | 3.011143 | average_word_embeddings_glove.6B.300d | 0.8 | 0.056565 | 0.20117 | 0.932836 |