Domain lexicon-based query expansion for patent retrieval

Patent retrieval is important for technology survey and knowledge protection. Its aim is to search as many patent documents relevant to the patent document query as possible, which is considered as a recall-oriented task. However, existing methods suffer from the term mismatch problem caused by the frequent use of many non-standard technical terminologies in patents. To address the issue, we present a novel patent retrieval approach by utilizing domain lexicon-based query expansion. In particular, we present the domain lexicon construction scheme and the domain lexicon-based query expansion algorithm to augment the query with suitable expansion concepts. Experimental results on the CLEF-IP patent data set demonstrate that the proposed approach can achieve significant improvement in retrieval performance compared to several typical methods.