Jingyi77 commited on
Commit
311bc5d
·
verified ·
1 Parent(s): 16ac448

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +82 -80
README.md CHANGED
@@ -1,77 +1,73 @@
1
  ---
2
  language:
3
- - zh
4
  license: mit
5
  pretty_name: RedNote Covert Advertisement Detection Dataset
6
  tags:
7
- - covert advertisement detection
8
- - social-media
9
- - image-text
10
- - multimodal
11
- - RedNote
12
- - Xiaohongshu
13
  datasets:
14
- - Jingyi77/CHASM-Covert_Advertisement_on_RedNote
15
  dataset_info:
16
- - config_name: default
17
- features:
18
- - name: id
19
- dtype: string
20
- - name: title
21
- dtype: string
22
- - name: description
23
- dtype: string
24
- - name: date
25
- dtype: string
26
- - name: comments
27
- sequence:
28
- dtype: string
29
- - name: images
30
- sequence:
31
- dtype: binary
32
- - name: image_count
33
- dtype: int32
34
- - name: label
35
- dtype: int8
36
- - name: split
37
- dtype: string
38
  configs:
39
- - config_name: default
40
- data_files:
41
- - split: test
42
- path: test.parquet
43
- - split: train
44
- path: train.parquet
45
- - split: validation
46
- path: validation.parquet
47
 
48
  size_categories:
49
- - 1K<n<10K
50
  ---
51
 
 
 
52
  # RedNote Covert Advertisement Detection Dataset
53
 
54
  This dataset contains posts from the RedNote platform for covert advertisement detection tasks.
55
 
56
  ## Dataset Overview
57
 
58
- | Split | Posts | Ad Posts | Non-Ad Posts | Total Images |
59
- |-------|-------|----------|--------------|-------------|
60
- | Train | 3493 | 426 | 3067 | 18543 |
61
- | Validation | 499 | 57 | 442 | 2678 |
62
- | Test | 1000 | 130 | 870 | 5103 |
63
- | **Total** | **4992** | **613** | **4379** | **26324** |
64
-
65
- ## Dataset Structure
66
-
67
- The dataset is divided into three parts:
68
- - `train.parquet`: Training set
69
- - `validation.parquet`: Validation set
70
- - `test.parquet`: Test set
71
 
72
  ## Field Descriptions
73
 
74
  Each parquet file contains the following fields:
 
75
  - `id`: Unique identifier for each post
76
  - `title`: Post title
77
  - `description`: Post description content
@@ -82,43 +78,49 @@ Each parquet file contains the following fields:
82
  - `label`: Label (0=non-advertisement, 1=advertisement)
83
  - `split`: Data split (train/validation/test)
84
 
85
- ## Usage
86
 
87
- ```python
88
- import pandas as pd
 
89
 
90
- # Load the training set
91
- train_df = pd.read_parquet("train.parquet")
92
 
93
- # View dataset statistics
94
- print(f"Number of training samples: {len(train_df)}")
95
- print(f"Label distribution: {train_df['label'].value_counts()}")
96
- ```
97
 
98
- ## Image Processing
 
 
 
 
 
 
 
99
 
100
- Images are stored as base64 encoded strings. You can decode them using:
101
 
102
  ```python
103
- import base64
104
- import io
105
- from PIL import Image
106
 
107
- # Get the first image from the first sample
108
- base64_str = train_df.iloc[0]['images'][0]
109
 
110
- # Decode base64 and display the image
111
- img_data = base64.b64decode(base64_str)
112
- img = Image.open(io.BytesIO(img_data))
113
- img.show()
114
- ```
115
 
116
- ## Dataset Features
117
-
118
- - **Multimodal Data**: Each post contains both text (title, description, comments) and images
119
- - **Real-world Data**: Collected from actual social media posts on the RedNote platform
120
- - **Multiple Images**: Each post may contain multiple images (average of 5.27 images per post)
121
 
 
 
 
 
 
 
 
 
 
 
122
 
123
  ## Citation
124
 
@@ -126,11 +128,11 @@ If you use this dataset in your research, please cite:
126
 
127
  ```
128
  @dataset{CHASM,
129
- author = {Jingyi Zheng, Tianyi Hu, Yule Liu, Zhen Sun, Zongmin Zhang, Wenhan Dong, Zifan Peng},
130
  title = {CHASM: Unveiling Covert Advertisements on Chinese Social Media},
131
  year = {2025},
132
  publisher = {Hugging Face},
133
  journal = {Hugging Face Hub},
134
  howpublished = {\url{https://huggingface.co/datasets/Jingyi77/CHASM-Covert_Advertisement_on_RedNote}}
135
  }
136
- ```
 
1
  ---
2
  language:
3
+ - zh
4
  license: mit
5
  pretty_name: RedNote Covert Advertisement Detection Dataset
6
  tags:
7
+ - covert advertisement detection
8
+ - social-media
9
+ - image-text
10
+ - multimodal
11
+ - RedNote
12
+ - Xiaohongshu
13
  datasets:
14
+ - Jingyi77/CHASM-Covert_Advertisement_on_RedNote
15
  dataset_info:
16
+ - config_name: default
17
+ features:
18
+ - name: id
19
+ dtype: string
20
+ - name: title
21
+ dtype: string
22
+ - name: description
23
+ dtype: string
24
+ - name: date
25
+ dtype: string
26
+ - name: comments
27
+ sequence:
28
+ dtype: string
29
+ - name: images
30
+ sequence:
31
+ dtype: binary
32
+ - name: image_count
33
+ dtype: int32
34
+ - name: label
35
+ dtype: int8
36
+ - name: split
37
+ dtype: string
38
  configs:
39
+ - config_name: default
40
+ data_files:
41
+ - split: test
42
+ path: test.parquet
43
+ - split: train
44
+ path: train.parquet
45
+ - split: validation
46
+ path: validation.parquet
47
 
48
  size_categories:
49
+ - 1K<n<10K
50
  ---
51
 
52
+ <!-- @format -->
53
+
54
  # RedNote Covert Advertisement Detection Dataset
55
 
56
  This dataset contains posts from the RedNote platform for covert advertisement detection tasks.
57
 
58
  ## Dataset Overview
59
 
60
+ | Split | Posts | Ad Posts | Non-Ad Posts | Total Images |
61
+ | ---------- | -------- | -------- | ------------ | ------------ |
62
+ | Train | 3493 | 426 | 3067 | 18543 |
63
+ | Validation | 499 | 57 | 442 | 2678 |
64
+ | Test | 1000 | 130 | 870 | 5103 |
65
+ | **Total** | **4992** | **613** | **4379** | **26324** |
 
 
 
 
 
 
 
66
 
67
  ## Field Descriptions
68
 
69
  Each parquet file contains the following fields:
70
+
71
  - `id`: Unique identifier for each post
72
  - `title`: Post title
73
  - `description`: Post description content
 
78
  - `label`: Label (0=non-advertisement, 1=advertisement)
79
  - `split`: Data split (train/validation/test)
80
 
81
+ ## Dataset Features
82
 
83
+ - **Multimodal Data**: Each post contains both text (title, description, comments) and images
84
+ - **Real-world Data**: Collected from actual social media posts on the RedNote platform
85
+ - **Multiple Images**: Each post may contain multiple images (average of 5.27 images per post)
86
 
87
+ ## Data Format
 
88
 
89
+ The dataset is stored in WebDataset format, with each sample containing:
 
 
 
90
 
91
+ 1. One or more image files (.jpg format)
92
+ 2. A JSON metadata file with the following fields:
93
+ - `id`: Sample ID
94
+ - `title`: Title
95
+ - `description`: Description
96
+ - `date`: Date
97
+ - `comments`: List of comments
98
+ - `label`: Label (0: non-advertisement, 1: advertisement)
99
 
100
+ ## Loading the Dataset
101
 
102
  ```python
103
+ from datasets import load_dataset
 
 
104
 
105
+ # Load training set
106
+ train_dataset = load_dataset("Jingyi77/CHASM-Covert_Advertisement_on_RedNote", split="train")
107
 
108
+ # Load validation set
109
+ val_dataset = load_dataset("Jingyi77/CHASM-Covert_Advertisement_on_RedNote", split="validation")
 
 
 
110
 
111
+ # Load test set
112
+ test_dataset = load_dataset("Jingyi77/CHASM-Covert_Advertisement_on_RedNote", split="test")
 
 
 
113
 
114
+ # Access a sample
115
+ example = train_dataset[0]
116
+ metadata = {
117
+ "id": example["id"],
118
+ "title": example["title"],
119
+ "description": example["description"],
120
+ "label": example["label"]
121
+ }
122
+ images = example["images"] # List of images
123
+ ```
124
 
125
  ## Citation
126
 
 
128
 
129
  ```
130
  @dataset{CHASM,
131
+ author = {Jingyi Zheng, Tianyi Hu, Yule Liu, Zhen Sun, Zongmin Zhang, Wenhan Dong, Zifan Peng, Xinlei He},
132
  title = {CHASM: Unveiling Covert Advertisements on Chinese Social Media},
133
  year = {2025},
134
  publisher = {Hugging Face},
135
  journal = {Hugging Face Hub},
136
  howpublished = {\url{https://huggingface.co/datasets/Jingyi77/CHASM-Covert_Advertisement_on_RedNote}}
137
  }
138
+ ```