Semantic Graph based Pseudo Relevance Feedback for Biomedical Information Retrieval
Yuanyuan Zhang, James Z. Wang, Pradip K. Srimani · 2016
This paper proposed a novel pseudo relevance feedback strategy to facilitate the retrieval of more relevant biomedical documents by improving the quality of both feedback documents and expansion terms. Firstly, an ontology-graph based query expansion technique is applied to retrieve more relevant feedback documents. Secondly, useful expansion terms are extracted from the feedback documents based on a semantic graph based ranking approach. We add the expansion terms to the user query to retrieve more relevant documents. We use 10-fold cross validation technique to evaluate the performance of the proposed pseudo relevance feedback strategy over OHSUMED test collection. The experimental results demonstrate that the proposed strategy improves the retrieval performance by 33.8% over free-text based query in 11-point average precision. The proposed strategy also achieves better retrieval performance than two representative pseudo relevance feedback approaches. We have integrated this new strategy into G-Bean, a graph-based biomedical search engine. G-Bean is available at: http://bioinformatics.clemson.edu:8080/G-Bean/index.jsp