Using Event Identification Algorithm (EIA) to Improve Microblog Retrieval Effectiveness

摘要

Different from traditional Web page or database document, microblog has its unique features. Considering its sensitivity to the time, we introduce a new factor to improve the tweet retrieval effectiveness. The ranking score of a retrieved tweet is adjusted by the event identification algorithm (EIA). We evaluated our approach using 49 search topics and a total of 9,096,198 tweets collected provided by the trec 2011 microblog track. Our initial results indicated that EIA helped to improve the performance for topics with worse original average precision (AP) score and for topics that are represented with relatively short queries (one or two terms).