Biterm-LDA: A Recommendation Model for Latent Friends on Weibo

Dongsheng Liu, Yuwei Chen


Topic models are a popular and common clustering approach for textual data, which provides promising applications in topic discoveries. An extended topic model is resented in this paper for latent friends’ recommendation in social networks, named as Biterm-LDA. For the extension of the traditional topic models, the generation of topic LDA (Latent Dirichlet Allocation) is changed on the basic of BTM (Biterm Topic Model). Direct at the short texts, co-occurred words take place the single word when modeling. To evaluate the results, we generate a data collection from Weibo. It contains 10,000 blogs from 548 users. Experimental results reveal that compared with traditional and classical topic models, biterm-LDA has better performance in the topic extraction, and achieves more accurate results when recommending friends furtherly.

Full Text:



  • There are currently no refbacks.