Datasets:
Update germandpr-beir.py
Browse files- germandpr-beir.py +56 -36
germandpr-beir.py
CHANGED
@@ -1,4 +1,5 @@
|
|
1 |
import json
|
|
|
2 |
|
3 |
import datasets
|
4 |
|
@@ -6,67 +7,73 @@ import datasets
|
|
6 |
_VERSION = "1.0.0"
|
7 |
|
8 |
_DESCRIPTION = "Deepset's germanDPR dataset made compatible with BEIR benchmark framework. One version contains " \
|
9 |
-
"the original dataset 1:1 and the other dataset is preprocessed.
|
10 |
-
"
|
11 |
|
12 |
-
_SUBSETS = ["queries
|
|
|
13 |
|
14 |
|
15 |
class GermanDPRBeir(datasets.GeneratorBasedBuilder):
|
16 |
BUILDER_CONFIGS = (
|
17 |
[
|
18 |
datasets.BuilderConfig(
|
19 |
-
name="queries
|
20 |
-
description=f"BEIR queries created 1:1 from deepset/germanDPR.",
|
21 |
version=_VERSION,
|
22 |
),
|
23 |
datasets.BuilderConfig(
|
24 |
-
name="corpus
|
25 |
-
description=f"BEIR corpus created 1:1 from deepset/germanDPR.",
|
26 |
version=_VERSION,
|
27 |
),
|
28 |
datasets.BuilderConfig(
|
29 |
-
name="
|
30 |
-
description=f"BEIR
|
31 |
version=_VERSION,
|
32 |
),
|
33 |
datasets.BuilderConfig(
|
34 |
-
name="
|
35 |
-
description=f"BEIR
|
36 |
version=_VERSION,
|
37 |
),
|
38 |
datasets.BuilderConfig(
|
39 |
-
name="
|
40 |
-
description=f"BEIR
|
|
|
|
|
|
|
|
|
|
|
41 |
version=_VERSION,
|
42 |
)
|
43 |
]
|
44 |
)
|
45 |
|
46 |
-
DEFAULT_CONFIG_NAME =
|
47 |
|
48 |
def _info(self):
|
49 |
name = self.config.name
|
50 |
-
|
51 |
-
|
52 |
-
if name.startswith("queries"):
|
53 |
features = {
|
54 |
"_id": datasets.Value("string"),
|
55 |
"text": datasets.Value("string")
|
56 |
}
|
57 |
-
elif name.
|
58 |
features = {
|
59 |
"_id": datasets.Value("string"),
|
60 |
"title": datasets.Value("string"),
|
61 |
"text": datasets.Value("string"),
|
62 |
}
|
63 |
-
|
64 |
# name == qrels
|
65 |
features = {
|
66 |
"query-id": datasets.Value("string"),
|
67 |
"corpus-id": datasets.Value("string"),
|
68 |
"score": datasets.Value("int32")
|
69 |
}
|
|
|
|
|
70 |
|
71 |
return datasets.DatasetInfo(
|
72 |
description=f"{_DESCRIPTION}\n{self.config.description}",
|
@@ -77,26 +84,38 @@ class GermanDPRBeir(datasets.GeneratorBasedBuilder):
|
|
77 |
|
78 |
def _split_generators(self, dl_manager):
|
79 |
"""Returns SplitGenerators."""
|
80 |
-
_SPLITS = ["queries-original", "corpus-original", "queries-processed", "corpus-processed", "qrels"]
|
81 |
-
|
82 |
name = self.config.name
|
83 |
-
if name
|
84 |
-
dl_path = dl_manager.
|
85 |
-
|
86 |
-
|
87 |
-
|
|
|
|
|
|
|
|
|
88 |
return [
|
89 |
-
datasets.SplitGenerator(
|
90 |
-
|
|
|
|
|
|
|
|
|
91 |
]
|
92 |
-
|
93 |
-
dl_path = dl_manager.download(f"https://huggingface.co/datasets/PM-AI/germandpr-beir/resolve/main/data/{name}.jsonl")
|
94 |
return [
|
95 |
-
datasets.SplitGenerator(
|
96 |
-
|
|
|
|
|
|
|
|
|
97 |
]
|
|
|
|
|
98 |
|
99 |
def _generate_queries_data(self, filepath):
|
|
|
100 |
with open(filepath, "r", encoding="utf-8") as in_file:
|
101 |
for idx, line in enumerate(in_file):
|
102 |
data = json.loads(line)
|
@@ -120,10 +139,11 @@ class GermanDPRBeir(datasets.GeneratorBasedBuilder):
|
|
120 |
def _generate_examples(self, filepath):
|
121 |
"""Yields examples."""
|
122 |
name = self.config.name
|
123 |
-
if name.
|
124 |
return self._generate_queries_data(filepath)
|
125 |
-
elif name.
|
126 |
return self._generate_corpus_data(filepath)
|
127 |
-
|
128 |
-
# name == qrels
|
129 |
return self._generate_qrel_data(filepath)
|
|
|
|
|
|
1 |
import json
|
2 |
+
import os.path
|
3 |
|
4 |
import datasets
|
5 |
|
|
|
7 |
_VERSION = "1.0.0"
|
8 |
|
9 |
_DESCRIPTION = "Deepset's germanDPR dataset made compatible with BEIR benchmark framework. One version contains " \
|
10 |
+
"the original dataset 1:1 (but deduplicated) and the other dataset is furhter preprocessed. " \
|
11 |
+
"See official dataset card for dataset usage with BEIR."
|
12 |
|
13 |
+
_SUBSETS = ["original-queries", "original-corpus", "original-qrels",
|
14 |
+
"processed-queries", "processed-corpus", "original-qrels"]
|
15 |
|
16 |
|
17 |
class GermanDPRBeir(datasets.GeneratorBasedBuilder):
|
18 |
BUILDER_CONFIGS = (
|
19 |
[
|
20 |
datasets.BuilderConfig(
|
21 |
+
name="original-queries",
|
22 |
+
description=f"BEIR queries created 1:1 but deduplicated from deepset/germanDPR.",
|
23 |
version=_VERSION,
|
24 |
),
|
25 |
datasets.BuilderConfig(
|
26 |
+
name="original-corpus",
|
27 |
+
description=f"BEIR corpus created 1:1 but deduplicated from deepset/germanDPR.",
|
28 |
version=_VERSION,
|
29 |
),
|
30 |
datasets.BuilderConfig(
|
31 |
+
name="original-qrels",
|
32 |
+
description=f"BEIR qrels for original version of deepset/germanDPR.",
|
33 |
version=_VERSION,
|
34 |
),
|
35 |
datasets.BuilderConfig(
|
36 |
+
name="processed-queries",
|
37 |
+
description=f"BEIR queries created, deduplicated and further text-processed from deepset/germanDPR.",
|
38 |
version=_VERSION,
|
39 |
),
|
40 |
datasets.BuilderConfig(
|
41 |
+
name="processed-corpus",
|
42 |
+
description=f"BEIR corpus created, deduplicated and further text-processed from deepset/germanDPR.",
|
43 |
+
version=_VERSION,
|
44 |
+
),
|
45 |
+
datasets.BuilderConfig(
|
46 |
+
name="processed-qrels",
|
47 |
+
description=f"BEIR qrels for processed version of deepset/germanDPR.",
|
48 |
version=_VERSION,
|
49 |
)
|
50 |
]
|
51 |
)
|
52 |
|
53 |
+
DEFAULT_CONFIG_NAME = _SUBSETS[0]
|
54 |
|
55 |
def _info(self):
|
56 |
name = self.config.name
|
57 |
+
if name.endswith("queries"):
|
|
|
|
|
58 |
features = {
|
59 |
"_id": datasets.Value("string"),
|
60 |
"text": datasets.Value("string")
|
61 |
}
|
62 |
+
elif name.endswith("corpus"):
|
63 |
features = {
|
64 |
"_id": datasets.Value("string"),
|
65 |
"title": datasets.Value("string"),
|
66 |
"text": datasets.Value("string"),
|
67 |
}
|
68 |
+
elif name.endswith("qrels"):
|
69 |
# name == qrels
|
70 |
features = {
|
71 |
"query-id": datasets.Value("string"),
|
72 |
"corpus-id": datasets.Value("string"),
|
73 |
"score": datasets.Value("int32")
|
74 |
}
|
75 |
+
else:
|
76 |
+
raise ValueError(f'Unknown subset, choose from: {", ".join(_SUBSETS)}')
|
77 |
|
78 |
return datasets.DatasetInfo(
|
79 |
description=f"{_DESCRIPTION}\n{self.config.description}",
|
|
|
84 |
|
85 |
def _split_generators(self, dl_manager):
|
86 |
"""Returns SplitGenerators."""
|
|
|
|
|
87 |
name = self.config.name
|
88 |
+
if name.startswith("original"):
|
89 |
+
dl_path = dl_manager.download_and_extract("https://huggingface.co/datasets/PM-AI/germandpr-beir/resolve/main/data/original.tar.gz")
|
90 |
+
elif name.startswith("processed"):
|
91 |
+
dl_path = dl_manager.download_and_extract("https://huggingface.co/datasets/PM-AI/germandpr-beir/resolve/main/data/processed.tar.gz")
|
92 |
+
else:
|
93 |
+
raise ValueError(f'Unknown subset, choose from: {", ".join(_SUBSETS)}')
|
94 |
+
|
95 |
+
type1, type2 = name.split("-")
|
96 |
+
if type2 in ["corpus", "queries"]:
|
97 |
return [
|
98 |
+
datasets.SplitGenerator(
|
99 |
+
name=datasets.Split.TRAIN,
|
100 |
+
gen_kwargs={"filepath": os.path.join(dl_path, f'{type1}/train/{type2}.jsonl')}),
|
101 |
+
datasets.SplitGenerator(
|
102 |
+
name=datasets.Split.TEST,
|
103 |
+
gen_kwargs={"filepath": os.path.join(dl_path, f'{type1}/test/{type2}.jsonl')})
|
104 |
]
|
105 |
+
elif type2 == "qrels":
|
|
|
106 |
return [
|
107 |
+
datasets.SplitGenerator(
|
108 |
+
name=datasets.Split.TRAIN,
|
109 |
+
gen_kwargs={"filepath": os.path.join(dl_path, f'{type1}/train/qrels/train.tsv')}),
|
110 |
+
datasets.SplitGenerator(
|
111 |
+
name=datasets.Split.TEST,
|
112 |
+
gen_kwargs={"filepath": os.path.join(dl_path, f'{type1}/test/qrels/test.tsv')})
|
113 |
]
|
114 |
+
else:
|
115 |
+
raise ValueError(f'Unknown subset, choose from: {", ".join(_SUBSETS)}')
|
116 |
|
117 |
def _generate_queries_data(self, filepath):
|
118 |
+
print("filepath: ", filepath)
|
119 |
with open(filepath, "r", encoding="utf-8") as in_file:
|
120 |
for idx, line in enumerate(in_file):
|
121 |
data = json.loads(line)
|
|
|
139 |
def _generate_examples(self, filepath):
|
140 |
"""Yields examples."""
|
141 |
name = self.config.name
|
142 |
+
if name.endswith("queries"):
|
143 |
return self._generate_queries_data(filepath)
|
144 |
+
elif name.endswith("corpus"):
|
145 |
return self._generate_corpus_data(filepath)
|
146 |
+
elif name.endswith("qrels"):
|
|
|
147 |
return self._generate_qrel_data(filepath)
|
148 |
+
else:
|
149 |
+
raise ValueError(f'Unknown subset, choose from: {", ".join(_SUBSETS)}')
|