Addressing Limitations of Patent Research Using Machine-Learning: A Research Agenda Based on Automatic F-term Classification and Technology Spanning Vector Data

Patent classifications play a vital role in Information Systems (IS) research due to their structured but rich technological information. However, the hierarchical structure of patent classifications presents three significant limitations: restricted horizontal comparability, the creation of technological silos and inconsistencies in global classifications. In this paper, we address these limitations by introducing a machine learning (ML) classifier for automatic F-term classification of patents. Our model classifies 378,165 unique F-terms, enabling granular comparison of patents and consistent cross-national comparability. Additionally, we provide vector representations of F-terms, facilitating cross-domain technology analyses and improved technology similarity measurements. Based on this, we propose a future research agenda in five directions to refine patent classification-based metrics, enhance firm and competitor analysis, and develop analyses for cross-domain technologies. This paper sets a foundation for ongoing advancements in patent-based analyses thereby enriching IS research.