What to tag your microblog: Hashtag recommendation based on topic analysis and collaborative filtering

Abstract

Hashtags are often utilized as metadata tags to mark messages for user-defined topics in a microblogging environment. However, difficulties in providing or selecting appropriate hashtags often force users giving up using them. In this paper, we propose a personalized method for hashtag recommendation that combines advantages of both topical information and collaborative intelligence. On one hand, we characterize the topic relevance of hashtags to posts based on content models. On the other hand, we predict an active user's hashtag usage preference in a collaborative filtering manner. Overall, we recommend hashtags by relevant scores for a specific microblog posted by a specific user. Experimental results show that our model is an effective solution for hashtag suggestion (MRR is around 96%) which outperforms the state-of-the-art methods. © 2014 Springer International Publishing Switzerland.