Simple and Effective Way to Disambiguate and Standardize Patent Applicants Using an Attention Mechanism With Data Augmentation

Abstract

Innovation in artificial intelligence and data science has sparked evolutions across numerous industries. Some companies are focusing on developing novel technologies to seize a rapidly evolving market, while others are exploring new business models to keep pace. The former and latter are typically referred to as first movers and fast followers in the technology market and identifying them can offer insights into technology market trends. Patent analysis is a good approach to exploring first movers and fast followers. However, patent applicants are classified into different patterns based on the structure or type of a company, making it challenging to disambiguate and standardize patent applicants. Therefore, this study proposes a method to disambiguate and standardize patent applicants. We present a simple and effective data augmentation approach that can help understand patent applicant patterns. The proposed approach trains on the augmented data via the attention mechanism. Our experiments provide empirical evidence for the performance of the proposed method, which accurately classifies 96.6% of the augmented data. Moreover, statistical hypothesis testing validates that the output of the proposed method is consistent with the ground truth.