This paper gives an overview of our work (the ICTIR group) in microblog track of TREC 2011 for tweets retrieval. The basic query likelihood model with smoothing is the fundamental method in our ap proaches, we also consider other factors: the author information and the negative feedback. Firstly, we classify all queries into three cat egories, construct refined feedback in different ways to reform them; Secondly, extremely short tweets lead to poor clustering performance, the author topic models are trained for tweets expansion and smooth ing. Finally, we train negative feedback model to reduce noise im pacts in our microblog search task. Experimental results show that our methods could improve the retrieval performance greatly.
【 预 览 】
附件列表
Files
Size
Format
View
Author Model and Negative Feedback Methods on TREC 2011 Microblog Track