Information Retrieval and Language Model based Expert Search
This event took place on Wednesday 24 January 2007 at 11:30
Both research and industry communities are paying lots of attention to expert search recently. Text REtrieval Conference (TREC) has organized expert search task for 2005 and 2006. We have participated in TREC 2006 expert search task and achieved the best run judged by all information retrieval measures among 23 groups. We propose to integrate three document characteristics, namely, document authority, document internal structure, and various levels of associations between an expert and a search topic, in addition to document content, in a two-stage language model for effective expert search. We have used the TREC W3C dataset to test the effectiveness of the three document characteristics in terms of measures such as mean average precision, bpref, and Precision@10 etc.
Watch the webcast replay >>