Finding Good XML Fragments Based on k-Medoid Cluster Number Optimization and Ranking Model for Feedback

Due to low quality feedback set, traditional pseudo relevance feedback may bring into topic drift. This paper studies how to identify or find good xml documents(fragments) for feedback. We propose an effective method, in which xml element search results clustering is performed firstly by k-mediod cluster number optimization, and then those fragments with high relevant to the query are identified and ranked in the top position by ranking model. The final experimental results show that the proposed approach produces better performance and achieves high quality feedback set.