Catch-phrase based document representation for improved prior art search

Abstract

With rising number of patent infringement cases, it is vital to both individual researchers and patent lawyers to search and assess existing patents similar to a patent before its filing. This paper proposes a query reduction based methodology that furnishes a set of ranked patent documents similar to a given input document. It utilizes catch phrase based document representation to measure patent similarity. A new measure for quantifying the similarity between overlapping catch phrases is proposed. Different document similarity measures utilizing the catch phrase similarity score are also presented. Proposed measures are evaluated using Mean Average Precision (MAP) and Normalized Discounted Cumulative Gain (nDCG).