Emotion classification

2 datasets for emotion classification. Most are inferred from the data: a label field, the title, or the graph itself.

2 datasets.

LiveJournal Emotions Dataset (Jin & Zafarani 2017) Analyzed

A LiveJournal (blogging social network) dataset of 12,404,868 emotion-labeled posts from 115,444 users with 106,577 nodes and 961,392 friendship/follower edges, collected by Jin & Zafarani via LiveJournal API, distributed by Data Lab SU. Each record contains user ID, emotion label (-1=negative/0=neutral/1=positive), timestamp, friendship edges, follower edges, community membership. 12,404,868 LiveJournal posts labeled with trichotomous emotion scores across 115,444 users — uniquely combines social network structure (friendship + follower graphs) with per-post emotion labels and community membership for studying emotion propagation in networks. Suitable for emotion propagation in social networks, community emotion analysis, sentiment diffusion, social influence on emotion, network-based mood modeling.

12.4M records106.6K nodes961.4K edges2017

graph

LiveJournal Blogging Emotion Sentiment+1

LiveJournal Rich Social Network (Emotion Dataset) Analyzed

A LiveJournal social network of 106,577 users and 961,392 directed edges (793,948 follower links and 246,164 friendship links), from a crawl of the users within three hops of the seed user krazykitkat, pre-processed for the paper 'Emotions in Social Networks: Distributions, Patterns, and Models' (Jin and Zafarani): only users with 10 or more posts carrying a predefined LiveJournal mood were kept, and user ids are anonymized. This copy holds the two network files (Friendship.txt, Followers.txt) and the release README; the per-post mood file (Posts.txt) and community memberships (Community.txt) that the README describes are not included. Suitable for social network structure, follower and friendship analysis, community detection, social influence modeling.

106.6K records106.6K nodes961.4K edges2017

graph

LiveJournal Emotion Sentiment Social graph