TOMOHA

Abstract

On Twitter, hashtags are used to summarize topics of the tweet content and to help to categorize and search tweets. However, hashtags are created in a free style and thus heterogeneous, increasing difficulty of their usage. We propose TOMOHA, a supervised TOpic MOdel-based solution for HAshtag recommendation on Twitter. We treat hashtags as labels of topics, and develop a supervised topic model to discover relationship among words, hashtags and topics of tweets. We also novelly add user following relationship into the model. We infer the probability that a hashtag will be contained in a new tweet, and recommend k most probable ones. We propose parallel computing and pruning techniques to speed up model training and recommendation process. Experiments show that our method can properly and efficiently recommend hashtags.