IPC Selection Using Collection Selection Algorithms

Abstract

In this paper we view the automated selection of patent classification codes as a collection selection problem that can be addressed using existing methods which we extend and adapt for the patent domain. Our work exploits the manually assigned International Patent Classification (IPC) codes of patent documents to cluster, distribute and index patents through hundreds or thousands of sub-collections. We examine different collection selection methods (CORI, Bordafuse, ReciRank and multilayer) and compare their effectiveness in selecting relevant IPCs. The multilayer method, in addition to utilizing the topical relevance of IPCs at a specific level (e.g. sub-class), exploits the topical relevance of their ancestors in the IPC hierarchy and aggregates those multiple estimations of relevance to a single estimation. The results show that multilayer outperforms CORI and fusion-based methods in the task of IPC suggestion.