Discovering research areas from patents: A case study in autonomous vehicles industry

Abstract

Seven research areas introduced by the 'Autonomous Systems' research lab provide research areas required to enable the autonomous vehicle industry. For ensuring the validity of the research areas with the baseline, we apply Latent Semantic Analysis (LSA) and Latent Dirichlet Allocation (LDA) on the US patents containing 'Autonomous Vehicles' to identify keywords and research areas of relevant technologies. Keyword clustering and TF-IDF are repeatedly applied to the retrieved keywords to further filter out irrelevant words. Coherence values for LSA and LDA are evaluated to determine an adequate number of topics that need to be generated. We found that results from LSA provide a list of technologies already included in the baseline while topics from LDA provide associated keywords to support defining each technology. We conclude the numbers and topics provided by the baseline model closely represent the industry of autonomous vehicles but the identified topics from us provide a significant extension in research areas. The resulting research areas may provide overviews and guidelines on the autonomous vehicles industry for researchers and institutes