A semantic query expansion-based patent retrieval approach

Abstract

Since patent documents are important technical resources, effective patent retrieval has become more and more crucial. Unlike common information retrieval, patent retrieval is a recall-oriented retrieval, and patent query inputs are usually long. However, current patent retrieval approaches cannot effectively capture user query intents and obtain good expansion terms, which lead to low retrieval effectiveness. To address this issue, this paper proposes a novel semantic query expansion-based patent retrieval approach according to patent-specific characteristics. Firstly, patent domain features are extracted by using a domain-dependent term frequency scheme. Based on domain features, query inputs are analyzed to determine query domains. Furthermore, query domain matching is employed to generate candidate expansion terms, and semantic-based similarity computation is adopted to select expansion terms. Experiment results show that our approach achieves better retrieval performance than other state-of-art approaches.