Identifying Technologies in Patents Via Transformers

Timely identification of technology trends in the market is essential for continuous improvement of technology management. Technological information can be retrieved from patent applications. Existing methods mainly include text mining and search activities performed by human experts. Both show shortcomings that need to be addressed. To improve the recognition of technologies in the patent application, we introduce two methods using natural language processing. Our work shows that natural language processing can be used to improve the recognition process, although the last step has not been automated. Therefore, our solution can serve as a benchmark for further investigation