A scoring method of XML fragments considering query-oriented statistics

In this paper, we propose a scoring method for searching XML fragments related with user's information need. In conventional XML fragment retrieval systems, the term-weighting schemes based on traditional information retrieval techniques are usually used directly for scoring retrieved XML fragments. However, we believe that retrieved XML fragments should be scored considering not only traditional retrieved-document-oriented statistics like the tfidf scoring but also query-oriented ones such as constituent rate of query keywords and statistics of the query results, so that it remains possible that such techniques will help the conventional XML fragment retrieval systems to improve their relevances. Our experimental results show that our method considering query conditions is more effective for searching XML fragments correctly than conventional approaches.