Document Structure Analysis in Associative Patent Retrieval

This paper describes our retrieval system participated in the Patent Retrieval Task at the Fourth NTCIR Workshop. The main task was an associative patent retrieval task, in which a patent application including a target claim is used to search documents that can invalidate the demand in the claim. Our system can be characterized by the structure analysis for both target claim and entire application. Based on the rhetorical structure, a claim is segmented into multiple components , each of which is used to produce an initial query. The structure of an application is used to enhance each query. The candidates of relevant documents are retrieved and ranked on a component-by-component basis. The final document list is obtained by integrating these documents lists. All passages in each document are ranked according to the relevance to the target claim. We show the effectiveness of our system by means of the formal run evaluation.