Network Models and Content Models for Social Media Users, Domains of Interest, and User Similarity

Dataset containing processed Twitter users information for content and graph analysis. Users are grouped in 4 manually-defined communities (fashion, finance, Australian writers and chess players) and an additional set of 160 random users is included in the collection. Content features are calculated with NLTK and Dandelion: each user is represented by a vector of words and entities occurrences, based on their tweets. Networks are built using SNAP, network features are extracted using node2vec: for each domain, the following, followers and mentions graph is calculated and stored in two separated files for edges and nodes; additionally, the embeddings are stored for each node in a separated file.