Evolutionary Personalized Hashtag Recommendation

Hashtags, starting with a symbol “#” ahead of terms, are widely used and inserted anywhere within posts as they present rich sentiment information on topics that people are really interested in. In this paper, we focus on the problem of hashtag recommendation considering its personalized and evolutionary aspects. We introduce three features to model personal user interest and its evolution, including (1) hashtag popularity; (2) hashtag textual information; and (3) hashtag time factor. We construct a hybrid model combining these features to learn user preference and recommend personalized hashtags consequently.