A patent quality evaluation model based on ALBERT and BiGRU

Abstract

Accurately and efficiently evaluating patent quality is an important issue in the field of patents, demanding high-quality methods for patent assessment. This paper investigates quality evaluation indicators and patent quantification and proposes a patent quality assessment model based on the ALBERT-BiGRU network. Firstly, a patent quality evaluation indicator system is constructed, and the numerical indicators of patents are quantified. By integrating numerical indicators and text summaries, the patent evaluation model, ALBERT-BiGRU, is developed and trained and tested on a patent dataset with patent grade labels. The experimental results achieve an accuracy of 80.8%. This approach addresses the existing research gap where many methods overlook the textual features of patents. The proposed solution significantly improves the effectiveness of patent quality evaluation, providing a new perspective and approach for patent protection and intellectual property management.