A Huawei telecom network (anonymized carrier data) dataset of 170,380 nodes (100,000 users, 1,000 apps, 380 packages) and 2,000,000 directed edges, collected by Huawei Global Technology Services (anonymized, 2021). Each record contains user ID, app ID, package ID, cell ID; edge types: user-buy-package and user-use-app; normalized float feature vectors per node [0,1]. Heterogeneous graph from a real telecom carrier with 4 node types (user, app, package, cell) and 2 edge types — one of the few publicly available large-scale telecom behavioral graphs for research.. Suitable for heterogeneous graph neural networks, telecom network analysis, graph algorithm benchmarking, user behavior modeling in carrier networks.
A Douban (Chinese social network) dataset of 154,908 users and 327,162 friendship edges, collected by Long Qiu via Douban crawl (December 2010). Each record contains user-user friendship edges and user IDs. The dataset is notable for its friendship network from Douban, China's largest online community for book, movie, and music reviews and recommendations. Intended for research into social network analysis, community detection, Chinese social media research, and content-sharing platform structure studies.
A Stack Exchange (chinese.stackexchange.com) dataset of 7,358 users, 19,445 answer-collaboration edges, 11,570 questions, 28,885 answers, collected by Stack Exchange, Inc. via the annual Stack Exchange Data Dump. Each record contains user nodes (reputation, badge counts, vote history), directed weighted edges (answerer→asker, weight=answer score, count=interactions, accepted=bool), post text, tag distribution. User collaboration network from chinese.stackexchange.com — directed weighted graph where edges represent knowledge-sharing relationships (A answered B's question), with edge weight reflecting answer quality. Enables prestige hierarchy, community detection, and expertise flow analysis.. Suitable for expertise network analysis, reputation systems, community detection, knowledge flow, PageRank-based authority ranking, Q&A platform dynamics..
A TextDetox 2024/2025 shared task (15 languages) dataset of 71,374 posts across 15 languages with binary toxicity labels, collected by Dementieva et al. for PAN/CLEF TextDetox 2024. Each record contains post text, binary toxicity label (0=non-toxic, 1=toxic), language. Combined multilingual toxicity benchmark covering 15 languages including low-resource languages (Amharic, Tatar, Hindglish) — 71,374 balanced samples total. Suitable for multilingual toxicity detection, cross-lingual transfer learning, low-resource NLP, content moderation benchmarking.
A TextDetox 2024 shared task (Chinese) dataset of 5,000 Chinese posts (2,500 toxic, 2,500 non-toxic), collected by Dementieva et al. for PAN/CLEF TextDetox 2024. Each record contains post text, binary toxicity label (0=non-toxic, 1=toxic). TextDetox Chinese provides 5,000 balanced binary toxicity annotations — part of a 15-language multilingual benchmark for cross-lingual toxicity detection research. Suitable for toxicity detection, multilingual content moderation, cross-lingual NLP, hate speech classification, text detoxification.
A 4chan, Twitter, and Weibo (cross-platform) dataset of 2,455 multimodal posts (1,318 hate, 1,137 non-hate), collected by Mira AI Lab (2024). Each record contains image filename, image OCR text, post text, hate/non-hate binary label, hate category (gender/race/religion/politics/violence/health/etc), and human-annotated reason for classification. Multimodal cross-platform collection hate speech dataset combining image OCR text and post text with fine-grained category labels across 8 hate dimensions — 54% of posts are hate speech with human-annotated reasoning. Suitable for multimodal hate speech detection, cross-platform analysis, hate category classification, image-text hate speech, content moderation research. The package includes the 2,455 meme images (img/) that the records name.
A Weibo (CHECKED dataset) dataset of 2,104 microblogs with engagement and expert analysis, collected by Yang, Zhou, and Zafarani via Weibo (2020-2021). Each record contains microblog text, images, videos, engagement metrics (comments, reposts, likes), and expert analysis for fake news. The dataset is notable for its first Chinese COVID-19 fake news dataset with expert analysis, anonymized user IDs, and rich social interaction data. Intended for research into fake news detection, misinformation analysis, Chinese NLP, social network analysis, and public health communication.
A 4chan dataset of 1,400 multimodal posts (1,062 hate, 338 non-hate), collected by Mira AI Lab (2024). Each record contains image filename, image OCR text, post text, hate/non-hate binary label, hate category (gender/race/religion/politics/violence/health/etc), and human-annotated reason for classification. Multimodal 4chan subset hate speech dataset combining image OCR text and post text with fine-grained category labels across 8 hate dimensions — 76% of posts are hate speech with human-annotated reasoning. Suitable for multimodal hate speech detection, cross-platform analysis, hate category classification, image-text hate speech, content moderation research. The package includes the 1,400 meme images (img/) that the records name.
A Cross-platform meme corpus (4chan, Twitter, Weibo) dataset of 1,318 hate-speech labeled meme records from 3 platforms, collected by Mira AI Lab via cross-platform meme collection. Each record contains image filename, OCR-extracted image text, post caption text, hate speech label, category list (politics/race/religion/etc.), and human-readable annotation reason per record. M3-Hate is the hate-speech filtered subset of the M3 meme dataset, with 1,318 hateful memes across 4chan (1062), Twitter (132), and Weibo (124) — multimodal with both image OCR text and post captions. Suitable for hate speech detection, multimodal meme analysis, cross-platform content moderation, image-text classification, harmful content research. The package includes the 1,318 meme images (img/) alongside the JSON records.
A Weibo dataset of 529 multimodal posts (124 hate, 405 non-hate), collected by Mira AI Lab (2024). Each record contains image filename, image OCR text, post text, hate/non-hate binary label, hate category (gender/race/religion/politics/violence/health/etc), and human-annotated reason for classification. Multimodal Weibo subset hate speech dataset combining image OCR text and post text with fine-grained category labels across 8 hate dimensions — 23% of posts are hate speech with human-annotated reasoning. Suitable for multimodal hate speech detection, cross-platform analysis, hate category classification, image-text hate speech, content moderation research. The package includes the 529 meme images (img/) that the records name.
A Twitter/X dataset of 526 multimodal posts (132 hate, 394 non-hate), collected by Mira AI Lab (2024). Each record contains image filename, image OCR text, post text, hate/non-hate binary label, hate category (gender/race/religion/politics/violence/health/etc), and human-annotated reason for classification. Multimodal Twitter subset hate speech dataset combining image OCR text and post text with fine-grained category labels across 8 hate dimensions — 25% of posts are hate speech with human-annotated reasoning. Suitable for multimodal hate speech detection, cross-platform analysis, hate category classification, image-text hate speech, content moderation research. The package includes the 526 meme images (img/) that the records name.
Douban
Stack Exchange
4chan
Twitter (X)
Weibo