Research On Microblog Recommendation And Search Based On Hashtag

With the rapid development of social network,Micro-blogging has drawn more and more attention. Most social network contain hashtag, it is convenient for uses to find specific topics and events. O n the one hand, recommending relevant hashtags to users in the cloud is challenging, because hashtags are created at tremendous speed alongside microblogs, and scattered in micro-blogging systems without a systematic organization. On the other hand, exits micro-blog real-time search index systems do not adequately consider the topics and events related to hashtags when indexing, resulting in indexing and query cost increases.To solve the above problems, a personalized hashtag recommendation approach and an hashtag based microblog search method are proposed. In this method, user's feature vector is represented by user-topics distribution found by Hashtag-LDA. Hashtag-LDA is a Latent Dirichlet Allocation(LDA)-based topic model, it jointly modeling hashtags, words and users in microblogs. The recommending method firstly finds top-k similar users according to user-topic distribution, and then computes all hashtags' frequencies appeared in these users, and finally the most relevant hashtags are recommended to user. Besides, to fully take advantages of the properties that topics and events related to hashtags and to better meet the needs of users to search the hot events, the search method just index microbloggings that contain hashtags and named entity, and then use a series of mechanisms to give ratings, and finally the results ranked by ratings.To verify the performance of recommendation and search method, the methods are tested under a Twitter dataset. The Experimental results shows that Hashtag-LDA do enhance the influence of hashtags on latent topics generation and enhance performance of the recommending method; search method not only reduces the complexity of indexing and query overhead, but also ensures the relevance of query results.