Feasibility Analysis of Entity Recognition as a Means to Create an Autonomous Technology Radar

Abstract

Keeping up to date with the latest technology trends is crucial task for manufacturing companies to remain successful on a globally competitive market. Designing a technology radar is an established, yet mostly manual, process for visualizing recent technology trends. The challenge of identifying and visualizing technologies is addressed by the project TechRad which uses machine learning to realize an autonomous technology scouting radar. One of its core functionalities is the identification of technologies in text documents. This is implemented via Natural Language Processing (NLP). This paper aims to summarize the challenges and possible solutions for using entity recognition to identify relevant technologies in text documents. The authors present an early stage of implementation of the entity recognition model. This contains the selection of Transfer Learning as a suitable method, the creation of a dataset consisting of different data sources, as well as the applied model training process. Finally, the performance of the chosen method is benchmarked and evaluated in a series of tests.