A Instagram (fashion influencers) dataset of 3,411,944 posts from fashion influencers, collected by Seungbae Kim and Jinyoung Han (2020). Each record contains post caption text, like count, comment count, video/ad flags, tagged user count, location presence, image dimensions, timestamp, and influencer follower count. The dataset is notable for its subset of 9.7M-post Instagram influencer dataset covering the fashion category, with rich per-post engagement signals and influencer-level metadata. Intended for research into influencer marketing analysis, engagement prediction, social media content analysis, and sponsorship detection. This package holds the post metadata (one JSON .info file per post) and influencer lists; the post images, which the authors distribute separately on request, are not included.
A YouTube dataset of 2,139,109 users with 38,047,355 uploads and 40,088,341 friend relationships, collected by Cheng et al. via YouTube API (2007-2008). Each record contains user ID, number of uploaded videos, number of friends/subscriptions. YouTube user network from 2008 capturing user activity and friendship patterns during the platform's early growth period. Suitable for social network analysis, user behavior analysis, influencer identification, community detection.
A Instagram (other influencers) dataset of 1,625,526 posts from other influencers, collected by Seungbae Kim and Jinyoung Han (2020). Each record contains post caption text, like count, comment count, video/ad flags, tagged user count, location presence, image dimensions, timestamp, and influencer follower count. The dataset is notable for its subset of 9.7M-post Instagram influencer dataset covering the other category, with rich per-post engagement signals and influencer-level metadata. Intended for research into influencer marketing analysis, engagement prediction, social media content analysis, and sponsorship detection. This package holds the post metadata (one JSON .info file per post) and influencer lists; the post images, which the authors distribute separately on request, are not included.
A snapshot of Truth Social from February to September 2022, released by Patrick Gerard, Nicholas Botzer and Tim Weninger (University of Notre Dame) on Zenodo (7522646). 454,458 users and 845,060 posts ('truths'), with 4,002,115 follow edges, 506,276 replies and 10,508 quotes, plus hashtags, media and external URLs, as 13 tab-separated tables. Suitable for social network analysis, political communication and misinformation research on an alternative platform.
A Instagram (travel influencers) dataset of 1,203,723 posts from travel influencers, collected by Seungbae Kim and Jinyoung Han (2020). Each record contains post caption text, like count, comment count, video/ad flags, tagged user count, location presence, image dimensions, timestamp, and influencer follower count. The dataset is notable for its subset of 9.7M-post Instagram influencer dataset covering the travel category, with rich per-post engagement signals and influencer-level metadata. Intended for research into influencer marketing analysis, engagement prediction, social media content analysis, and sponsorship detection. This package holds the post metadata (one JSON .info file per post) and influencer lists; the post images, which the authors distribute separately on request, are not included.
A Instagram (family influencers) dataset of 1,170,485 posts from family influencers, collected by Seungbae Kim and Jinyoung Han (2020). Each record contains post caption text, like count, comment count, video/ad flags, tagged user count, location presence, image dimensions, timestamp, and influencer follower count. The dataset is notable for its subset of 9.7M-post Instagram influencer dataset covering the family category, with rich per-post engagement signals and influencer-level metadata. Intended for research into influencer marketing analysis, engagement prediction, social media content analysis, and sponsorship detection. This package holds the post metadata (one JSON .info file per post) and influencer lists; the post images, which the authors distribute separately on request, are not included.
A YouTube dataset of 1,062,308 users with 11,757,702 uploads and 773,531,437 friend relationships, collected by Cheng et al. via YouTube API (2007-2008). Each record contains user ID, number of uploaded videos, number of friends/subscriptions, subscribers. YouTube user network from 2007 capturing user activity and friendship patterns during the platform's early growth period. Suitable for social network analysis, user behavior analysis, influencer identification, community detection.
A Instagram (food influencers) dataset of 1,039,706 posts from food influencers, collected by Seungbae Kim and Jinyoung Han (2020). Each record contains post caption text, like count, comment count, video/ad flags, tagged user count, location presence, image dimensions, timestamp, and influencer follower count. The dataset is notable for its subset of 9.7M-post Instagram influencer dataset covering the food category, with rich per-post engagement signals and influencer-level metadata. Intended for research into influencer marketing analysis, engagement prediction, social media content analysis, and sponsorship detection. This package holds the post metadata (one JSON .info file per post) and influencer lists; the post images, which the authors distribute separately on request, are not included.
A TikTok dataset of 1,035,817 videos, 4,920 creators, 27,169,858 engagement records, collected by Lingbo et al.. Each record contains video_id, author_id, create_time, description, hashtags, music, daily engagement (plays, likes, comments, shares, collects). Large-scale TikTok dataset with video metadata and 30-day engagement trajectories. Suitable for engagement analysis, creator behavior, viral prediction, social media dynamics.
A Instagram (beauty influencers) dataset of 436,843 posts from beauty influencers, collected by Seungbae Kim and Jinyoung Han (2020). Each record contains post caption text, like count, comment count, video/ad flags, tagged user count, location presence, image dimensions, timestamp, and influencer follower count. The dataset is notable for its subset of 9.7M-post Instagram influencer dataset covering the beauty category, with rich per-post engagement signals and influencer-level metadata. Intended for research into influencer marketing analysis, engagement prediction, social media content analysis, and sponsorship detection. This package holds the post metadata (one JSON .info file per post) and influencer lists; the post images, which the authors distribute separately on request, are not included.
A Instagram (interior influencers) dataset of 348,559 posts from interior influencers, collected by Seungbae Kim and Jinyoung Han (2020). Each record contains post caption text, like count, comment count, video/ad flags, tagged user count, location presence, image dimensions, timestamp, and influencer follower count. The dataset is notable for its subset of 9.7M-post Instagram influencer dataset covering the interior category, with rich per-post engagement signals and influencer-level metadata. Intended for research into influencer marketing analysis, engagement prediction, social media content analysis, and sponsorship detection. This package holds the post metadata (one JSON .info file per post) and influencer lists; the post images, which the authors distribute separately on request, are not included.
A Instagram (fitness influencers) dataset of 319,414 posts from fitness influencers, collected by Seungbae Kim and Jinyoung Han (2020). Each record contains post caption text, like count, comment count, video/ad flags, tagged user count, location presence, image dimensions, timestamp, and influencer follower count. The dataset is notable for its subset of 9.7M-post Instagram influencer dataset covering the fitness category, with rich per-post engagement signals and influencer-level metadata. Intended for research into influencer marketing analysis, engagement prediction, social media content analysis, and sponsorship detection. This package holds the post metadata (one JSON .info file per post) and influencer lists; the post images, which the authors distribute separately on request, are not included.
A TikTok dataset of 209,543 videos from 1,872 creators, collected by Lingbo et al.. Each record contains video_id, author_id, create_time, description, transcript, GPT summary, topic, emotion scores, music metadata. Large-scale TikTok dataset with video metadata, GPT-generated summaries, emotion scores, and 30-day engagement trajectories. Suitable for engagement analysis, creator behavior, viral prediction, social media dynamics.
A Instagram (pet influencers) dataset of 169,998 posts from pet influencers, collected by Seungbae Kim and Jinyoung Han (2020). Each record contains post caption text, like count, comment count, video/ad flags, tagged user count, location presence, image dimensions, timestamp, and influencer follower count. The dataset is notable for its subset of 9.7M-post Instagram influencer dataset covering the pet category, with rich per-post engagement signals and influencer-level metadata. Intended for research into influencer marketing analysis, engagement prediction, social media content analysis, and sponsorship detection. This package holds the post metadata (one JSON .info file per post) and influencer lists; the post images, which the authors distribute separately on request, are not included.
A Facebook dataset of 11,565 pages, 67,114 mutual likes, collected by Rozemberczki et al. via the Facebook Graph API (November 2017). Each record contains undirected edges between blue verified Facebook public figure pages that mutually liked each other, with node attributes including page name and raw Facebook ID. Graph of blue verified Facebook public figure pages, where edges represent mutual likes between pages, capturing how audiences engage with celebrities, influencers, and notable personalities. Suitable for community detection in social networks, graph embedding, celebrity influence analysis, public figure engagement studies, social media personality network analysis.
A TUAW (The Unofficial Apple Weblog) dataset of 10,485 blog posts, collected by Agarwal, Tang, Liu, and Yu via TUAW crawl (2004-2007). Each record contains blog post title, text, author, categories, publication date, inlinks, outlinks, and comment counts. The dataset is notable for its blog dataset with rich metadata, representing a closed community of Apple-focused bloggers. Intended for research into influential blogger identification, blog post analysis, text classification, and influence modeling.
A Instagram (user engagement patterns) dataset of 1,000 posts with engagement metrics and multimedia, collected by Hanmante via Instagram API (2024). Each record contains post captions, hashtags, images, videos, likes, comments, user metadata, and tagged users. The dataset is notable for its dataset of Instagram political content from Uruguay, with rich multimodal data. Intended for research into social media engagement analysis, political communication, multimodal NLP, and user behavior modeling.
A Twitch dataset of 1,000 top streamers with viewership and engagement metrics, collected by Aayush Mishra via Twitch API (2020). Each record contains channel name, watch time, streaming time, max/average viewers, followers, followers gained, views gained, partnership status, age-restriction status, and language. A comprehensive snapshot of the top 1,000 Twitch streamers, capturing key engagement and growth metrics for the platform's most influential content creators. Suitable for influencer analysis, engagement prediction, platform dynamics, streamer growth modeling, gaming community studies.
Instagram
YouTube
Truth Social
TikTok
Facebook
Twitch