Innovation Hierarchy Based Patent Representation Model

Patent representation is a critical task for patent data mining. However, the patent content has depth and layers, which is hard to understand. There are two challenges in patent representation, including: how to identify the patent innovations and demonstrate the innovation hierarchy. To solve above issues, we propose the Innovation Hierarchy Based Patent Representation Model (IHPRM). In this model, we use the attention mechanism to identify innovations and the nested hierarchical Dirichlet process to represent the innovation hierarchy. In this paper, 1,000 patents are used to test the effectiveness of IHPRM. The result shows that IHPRM has better performance than the comparative model.

I. Introduction

With the development of innovation-driven economy, intellectual property has received extensive attention from countries around the world. As an important part of intellectual property, the patent brings a multitude of benefits to enterprises, which causes an increasing number of patent applications. When patent examiners manually examine the patents, they expect a patent representation that can outstand the innovations and represent the innovation hierarchy, since the representation will be convenient for the patent examination.