Personalized Tweet Ranking based on AHP A case study of micro-blogging message ranking in T.Sina

摘要

Micro-blog's handiness is besieging users with overloaded short snippets of tweets surging into their page. How to evaluate quality of tweets with limited content and rank them to direct user attention according to user's intangible preference is a new significant topic. In this paper, we study the problem of user-specific tweet evaluation and ranking. We propose a comprehensive, personalized tweet ranking mechanism (TweetRank) with a technique of AHP (Analytic Hierarchy Process) in operational research. Based on mathematics and psychology, the AHP can quantify the weight of each impact factor and model user blur preference precisely. Case study in Chinese micro-blog platform of T.sina showed that TweetRank greatly outperformed time-based ranking currently used in T.Sina, improving percentage of interesting content in Top10 to 60% from 20%. The work can be very helpful in directing user attention and can generalize to other recommendation contexts.