Skip to main content Skip to main navigation


Text2Tech – Deep Learning-based Text Mining for Technology Monitoring in Automotive Production

Jan-Tilman Seipp; Felix Köhler; David Harbecke; Leonhard Hennig; Phuc Tran Truong
In: 13th Global TechMining Conference 2023 - Conference Proceedings. Global TechMining Conference, November 10, Global TechMining Conference, 2023.


The automotive industry is undergoing a transformative phase with the integration of advanced technologies and the rise of intelligent manufacturing systems. To remain competitive in this dynamic landscape, automotive production requires effective utilization of technology monitoring as a part of technology intelligence, which encompasses the acquisition, analysis, and application of relevant technological information. By harnessing NLP techniques, automotive manufacturers can extract valuable insights from vast amounts of unstructured textual data available in the form of patents, research papers, publicly funded projects, and industry news. The goal of the Text2Tech research project is to develop methods for automated extraction of technologies and its relations to other entities from unstructured text sources. We formalize this task as a combination of Named Entity Recognition (NER, Yadav et al., 2018) and Relation Extraction (RE, Bach et al., 2020). Both NER and RE are fundamental, well-researched tasks in Natural Language Processing, however, their application to novel domains such as automotive manufacturing is often hindered by the lack of training and evaluation data. Prior research has shown the promising performance of Large Language Models (LLM) in such low-resource scenarios, e.g. for approaches based on few-shot learning (Fritzler et al., 2019) and instruction-tuning (Wang et al., 2023). In this study, we present preliminary results on the performance of LLM-based few-shot and instruction-based learning for the task of low-resource Named Entity Recognition in the domain of technology monitoring.