A twitter dataset of 2.7B+ tweet IDs across 9+ languages (ongoing collection since Jan 2020), collected by Chen, Lerman, and Ferrara (USC Information Sciences Institute). Each record contains tweet IDs only — full content requires hydration. Largest ongoing COVID-19 Twitter collection with multilingual coverage and continuous updates since pandemic onset. Suitable for pandemic discourse, public health surveillance, infodemic analysis, multilingual NLP.
A twitter dataset of 295,670,111 tweet IDs from the 2020 U.S. presidential election, collected by Wrubel and Kerchner via Social Feed Manager (GWU Libraries). Each record contains tweet IDs only — full content requires hydration. Complete tweet archive from the 2020 election including post-election period and January 6 lead-up. Suitable for election discourse, misinformation, political polarization, post-election unrest.
A Twitter/X dataset of 42,014,825 tweets from the 2024 US presidential election period across 39 data parts with full engagement metadata (41,561,521 records with engagement metrics), collected by USC Information Sciences Institute via Twitter/X API during the 2024 US election period. Each record contains tweet text, engagement metrics (reply count, retweet count, like count, quote count, view count), language, media attachments, reply chain metadata, user location, and publication timestamp. The dataset is notable for its one of the largest publicly available datasets from the 2024 US presidential election, with full engagement metadata per tweet enabling study of information amplification and political discourse dynamics. Intended for research into political opinion mining, election discourse analysis, misinformation detection, engagement prediction, and temporal analysis of political social media.
A Twitter (#NoMore hashtag campaign) dataset of 1,048,574 tweets from the Ethiopian #NoMore movement (2021-11 to 2022-08), collected by Zenodo contributors via Twitter API. Each record contains tweet text, language, timestamp; Amharic+English bilingual subset also available. 1,048,574 tweets covering the #NoMore solidarity movement in Ethiopia and Eritrea — one of few African social movement datasets with Amharic content, spanning 2021-11 to 2022-08. Suitable for African social movement analysis, multilingual NLP, Amharic text processing, political communication research, hashtag activism, Global South social media studies.
A Twitter dataset of 404,615 nodes, 434,963 edges, collected by Pierri et al. (OSoMe/Indiana University + Politecnico di Milano), 2020. Each record contains retweet network (uid1, uid2), user-domain bipartite network, user co-sharing network. Twitter news-sharing network revealing information polluters through retweet and co-sharing patterns.. Suitable for Information polluter identification, news sharing network analysis, source credibility..
A Twitter dataset of 346,573 nodes, 1,091,552 edges, collected by Shao et al. via Hoaxy API (OSoMe/Indiana University). Each record contains from_user_id, to_user_id, tweet_id, tweet_created_at, site_type (claim/fact-checking). Retweet diffusion network of misinformation and fact-checking content surrounding the 2016 US election; distinguishes claim vs. fact-checking retweets.. Suitable for Misinformation diffusion, fact-checking network analysis, network epidemiology..
A Twitter dataset of 263,555 tweets with binary event labels (14,129 disruptive events, 248,499 non-events) related to Bitcoin and cryptocurrency markets, collected by Mehta et al. via Twitter API, 2021. Each record contains tweet text, binary event/non-event label, follower count, location, username, Bitcoin price at time of tweet, and retweet count. one of few datasets explicitly linking social media content to financial market disruption events, with synchronized Bitcoin price data enabling direct study of social signal → market movement relationships. Suitable for event detection, financial market prediction from social media, cryptocurrency sentiment analysis, and disruptive event classification.
A X (Twitter) with Community Notes annotations dataset of 152,266 multimodal posts (21,845 AI-generated, 104,915 miscaptioned), collected by Chrysidis, Papadopoulos & Papadopoulos via X Community Notes (2026). Each record contains tweet ID, reply/favorite/view/retweet counts, Community Notes text and timestamp, and misinformation type label (miscaptioned / ai_generated / other). Large-scale dataset of 152,266 multimodal X posts with Community Notes — tracking the rise of AI-generated misinformation with 21,845 confirmed AI-generated posts. Includes engagement metrics showing that AI-generated content achieves disproportionate virality despite slower initial reporting.. Suitable for AI-generated content detection, misinformation classification, Community Notes effectiveness, multimodal misinformation analysis, virality and engagement modeling, temporal degradation of detectors.
A ReCOVery (news credibility repository) dataset of 2,029 news articles with reliability labels and social media engagement, collected by Zhou, Mulay, Ferrara, and Zafarani via news websites and Twitter (2020). Each record contains news article title and body, reliability label (1=reliable, 0=unreliable), publisher, political bias, country, author, and social media engagement data. The dataset is notable for its multimodal repository combining news content with social media engagement, designed for COVID-19 news credibility research. Intended for research into fake news detection, credibility assessment, political bias analysis, and multimodal misinformation research.
A Twitter dataset of 38,918 users and 32,523 directed, weighted reply edges (36,902 replies in all), collected by De Domenico et al. via the Twitter Streaming API around the Higgs boson announcement, 1-7 July 2012. Each record holds the replying user ID, the replied-to user ID and the number of replies between them. Complements the Higgs retweet and mention networks from the same collection. Suitable for information diffusion, scientific communication on social media and reply network structure.
Twitter conversations around the Charlie Hebdo shooting in Paris (January 2015), from the PHEME rumour dataset (Zubiaga, Liakata, Procter, Kochkina and colleagues; the complete release for rumour detection and veracity classification, figshare 6392078): 2,079 threads, each a source tweet and its replies, of which 458 were judged rumours by journalists and 1,621 not. Every rumour also carries its veracity as later established: 193 true, 116 false and 149 unverified. 38,268 tweets in all, with full tweet JSON. ECHO builds the reply network (36,189 reply links, each reply pointing to the tweet it answers) and labels each tweet with its thread's rumour status and, for source tweets, the veracity. Useful for rumour detection, rumour verification and modelling how conversations unfold. Annotations CC BY 4.0; Twitter retains the rights to tweet content. The archive PHEME_veracity.tar.bz2 is gzip-compressed despite its name (the name is the authors' own, kept so the checksum matches figshare): extract it with `tar xzf`.
A Twitter dataset of 25,327 nodes, 629,514 edges, collected by Shao et al. via Hoaxy + Botometer (OSoMe/Indiana University). Each record contains user centralities (betweenness, pagerank), bot scores, retweet graph, user-domain bipartite graph. Retweet network of low-credibility content with per-user bot scores; shows bots systematically amplify misinformation early in diffusion cascades.. Suitable for Bot-mediated misinformation diffusion, social bot impact measurement..
Twitter conversations around the Ferguson unrest after the shooting of Michael Brown (August 2014), from the PHEME rumour dataset (Zubiaga, Liakata, Procter, Kochkina and colleagues; the complete release for rumour detection and veracity classification, figshare 6392078): 1,143 threads, each a source tweet and its replies, of which 284 were judged rumours by journalists and 859 not. Every rumour also carries its veracity as later established: 10 true, 8 false and 266 unverified. 24,175 tweets in all, with full tweet JSON. ECHO builds the reply network (22,260 reply links, each reply pointing to the tweet it answers) and labels each tweet with its thread's rumour status and, for source tweets, the veracity. Useful for rumour detection, rumour verification and modelling how conversations unfold. Annotations CC BY 4.0; Twitter retains the rights to tweet content. The archive PHEME_veracity.tar.bz2 is gzip-compressed despite its name (the name is the authors' own, kept so the checksum matches figshare): extract it with `tar xzf`.
Twitter conversations around the Sydney Lindt café siege (December 2014), from the PHEME rumour dataset (Zubiaga, Liakata, Procter, Kochkina and colleagues; the complete release for rumour detection and veracity classification, figshare 6392078): 1,221 threads, each a source tweet and its replies, of which 522 were judged rumours by journalists and 699 not. Every rumour also carries its veracity as later established: 382 true, 86 false and 54 unverified. 23,996 tweets in all, with full tweet JSON. ECHO builds the reply network (22,775 reply links, each reply pointing to the tweet it answers) and labels each tweet with its thread's rumour status and, for source tweets, the veracity. Useful for rumour detection, rumour verification and modelling how conversations unfold. Annotations CC BY 4.0; Twitter retains the rights to tweet content. The archive PHEME_veracity.tar.bz2 is gzip-compressed despite its name (the name is the authors' own, kept so the checksum matches figshare): extract it with `tar xzf`.
A YouTube (with cross-platform spread on Twitter and Facebook) dataset of 7,659 removed YouTube videos with multilingual titles and descriptions, 249,346 associated tweets, and 18,897,277 Facebook shares, collected by Knuutila (2021) via YouTube Data API and CrowdTangle, tracking videos shared on public social media that were subsequently removed by YouTube for COVID-19 misinformation. Each record contains video titles and descriptions in multiple languages, view counts and subscriber counts where recoverable, Facebook reactions/comments/shares, Twitter and Facebook post ID lists for social spread tracking, removal and publication timestamps, and Wayback Machine archive URLs. The dataset is notable for its one of the few public datasets documenting the complete lifecycle of platform-removed health misinformation, linking video metadata to cross-platform social spread and enabling research into both content and dissemination patterns of COVID-19 misinformation. Intended for research into health misinformation detection, content moderation research, cross-platform information spread analysis, COVID-19 infodemic studies, multilingual misinformation narrative analysis, and platform policy impact assessment.
Twitter conversations around the Ottawa Parliament Hill shooting (October 2014), from the PHEME rumour dataset (Zubiaga, Liakata, Procter, Kochkina and colleagues; the complete release for rumour detection and veracity classification, figshare 6392078): 890 threads, each a source tweet and its replies, of which 470 were judged rumours by journalists and 420 not. Every rumour also carries its veracity as later established: 329 true, 72 false and 69 unverified. 12,284 tweets in all, with full tweet JSON. ECHO builds the reply network (11,394 reply links, each reply pointing to the tweet it answers) and labels each tweet with its thread's rumour status and, for source tweets, the veracity. Useful for rumour detection, rumour verification and modelling how conversations unfold. Annotations CC BY 4.0; Twitter retains the rights to tweet content. The archive PHEME_veracity.tar.bz2 is gzip-compressed despite its name (the name is the authors' own, kept so the checksum matches figshare): extract it with `tar xzf`.
A Twitter dataset of 32 nodes, 496 edges, collected by Yang, Ferrara, and Menczer (OSoMe/Indiana University). Each record contains user network (GML), suspicious tweet text, message cascades, targeted media accounts. Network and content evidence of a coordinated inauthentic influence operation targeting the 2020 US presidential election on Twitter.. Suitable for Coordinated inauthentic behavior detection, influence operations research, election integrity analysis..
A Twitter dataset of 4,952 tweets (1,238 rumour, 3,714 non-rumour) from the 2013 Boston Marathon bombings, collected by Zubiaga, Liakata, and Procter — Augmented PHEME (Zenodo, 2019). Each record contains tweet text, binary rumour/non-rumour label, timestamp, reaction thread structure linking source tweets to replies. Augmented extension of PHEME covering the Boston Marathon bombings event with both rumour threads and non-rumour threads, providing a balanced benchmark for rumour detection with reaction graph structure. Suitable for rumour detection, misinformation classification, crisis communication analysis, stance detection, conversational thread modelling.
A Twitter (state-sponsored IO accounts) dataset of 3,836 Russian IRA and 770 Iranian state-sponsored troll accounts with profile metadata, collected by Zannettou et al. via Twitter API (2019). Each record contains hashed user ID, display name, screen name, reported location, profile description, follower/following counts, account creation date, account language. Hashed and anonymized profiles of 4,606 verified state-sponsored troll accounts from Russian IRA and Iranian IO campaigns, enabling detection and characterization of coordinated inauthentic behavior. Suitable for information operations detection, state-sponsored troll identification, coordinated inauthentic behavior, influence campaign analysis, social media manipulation research.
Twitter conversations around the Germanwings plane crash in the French Alps (March 2015), from the PHEME rumour dataset (Zubiaga, Liakata, Procter, Kochkina and colleagues; the complete release for rumour detection and veracity classification, figshare 6392078): 469 threads, each a source tweet and its replies, of which 238 were judged rumours by journalists and 231 not. Every rumour also carries its veracity as later established: 94 true, 111 false and 33 unverified. 4,489 tweets in all, with full tweet JSON. ECHO builds the reply network (4,020 reply links, each reply pointing to the tweet it answers) and labels each tweet with its thread's rumour status and, for source tweets, the veracity. Useful for rumour detection, rumour verification and modelling how conversations unfold. Annotations CC BY 4.0; Twitter retains the rights to tweet content. The archive PHEME_veracity.tar.bz2 is gzip-compressed despite its name (the name is the authors' own, kept so the checksum matches figshare): extract it with `tar xzf`.
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 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 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.
Facebook
YouTube
4chan
Weibo