What makes a tweet relevant for a topic

摘要: Users who rely on microblogging search (MS) engines to find relevant microposts for their queries usually follow interests and rationale when deciding whether a retrieved post is of interest them or not. While today’s MS commonly keyword-based retrieval strategies, we investigate if there exist additional micropost characteristics that are more predictive post’s relevance interestingness than its similarity with the query. In this paper, experiment corpus Twitter messages sixteen features along two dimensions: topicdependent topic-independent features. Our in-depth analysis compares importance di↵erent types reveals semantic therefore an understanding meaning tweets plays major role in determining tweet respect We evaluate our findings classification show by combining features, can achieve precision recall 35% 45% respectively.