DUTIR at TREC 2008 Relevance Feedback Track

This paper details our experiments carried out at TREC 2008 Relevance Feedback Track. We focused on the analysis of feedback documents, both relevant and non-relevant, to explore more useful information to improve retrieval performance. In our experiments, local co-occurrence model and a Rocchio formula were used to select good expansion terms. Five runs were submitted. These runs used different amount of relevance info for analysis.

• Type: Conference
• Year: 2008
• Where: TREC
• Authors: Xiaoling Sun, Yuan Lin, Hongfei Lin, Zhihao Yang
• Keywords: Feedback Track, Information Technology, Local Co-occurrence Model, TREC 2008, Trec 2008 Relevance