Patent infringement analysis using a text mining technique based on SAO structure

As can be seen in the emergence of non-practicing entities, patent infringement lawsuits are very significant events for companies, both financially and technologically. Thus, the importance of patent infringement analysis has been emphasized to support a decision-making process of potential stakeholders. Since identifying patent infringement needs to consider various factors, the most appropriate method is to review the expert analysis in each case. However, as the size of valuable data continues to grow in recent years, the need for automated quantitative analysis that enables to perform such processes without experts has increased. Thus, this research aims to develop an automated approach for patent infringement using Subject-Action-Object structure-based text mining technique and SAO2Vec, which focus on the functions of technology in patent documents and product documents. The proposed framework consists of three modules. In the first module, the types of companies in which patent infringement can occur are defined, and then lists of companies selected by various databases are identified. In the second module, vectors of the SAO structures are derived from the patent documents of the selected company using the Doc2Vec-based SAO2Vec. In the last module, the results of the first and second modules are used to calculate the patent infringement indicators. To validate the suggested approach, we applied it to the case of Nintendo, which had recently become an issue in patent infringement lawsuits. We found that the proposed indicators were a statistically good indicator to judge the patent infringement, identifying the pairs of patents that have a high possibility of patent infringement. (C) 2020 Elsevier B.V. All rights reserved.

Automated summary

The research aims to develop an automated approach for patent infringement using Subject-Action-Object structure-based text mining technique and SAO2Vec. The proposed framework consists of three modules and aims to help identify potential patent infringers.