An ensemble framework for patent classification

Abstract

An important task when a patent application arrives at a patent office is to assign one or more classification codes. This manual, intellectually demanding task needs to be supported or even fully automated by classification systems that will classify patent applications, hopefully with an accuracy close to patent professionals. Like in many other text analysis problems, in the last years, this task has been studied using deep learning techniques. However, these techniques did not manage to reach a classification accuracy high enough to totally depend on. An ensemble system that combines multiple classifiers obtaining better results could address this patent classification problem. However, this technique has not been explored in the domain of patent classification, and even in general, there are few studies focusing on the design of such systems. Our study investigates the design aspects of ensemble systems for patent classification and introduces an ensemble framework, which although is targeting the patent classification problem can be transferred to any other research domain.