Patent litigation mining using a large language model—Taking unmanned aerial vehicle development as the case domain

Abstract

As unmanned aerial vehicle (UAV), also called “drone”, swiftly advances with innovative functions and applications, the surge in patent applications has profoundly reshaped the intellectual property (IP) landscape in the UAV industry, leading to a growing number of litigations. This study is structured in two phases, aiming to develop an intelligent approach to analyzing the trend and evolution of patent litigations. The first phase involves macro- and micro-patent analyses of the related technology domain. Macro patent analysis elucidates the fundamental patent information in the drone industry, while micro patent analysis leverages the technology function matrix (TFM) to identify R&D hotspots and potentials. The second phase involves litigation (judgement) mining based on large language model (LLM). Beginning with the construction of a knowledge ontology, the domain infringement landscape can be detected through TFMs. A comparative analysis of the two-phase TFMs (i.e., both TFMs of patent and infringement allocations) is then conducted to pinpoint the key legal actions and the relevant technology. To drill deeper in infringement mining, dynamic topic modeling (DTM) is applied to analyze trends and dynamics in drone controller technology over time. This study aims to strengthen IP protection by developing an intelligent litigation mining approach that adopts large language model (LLM) and uses UAV/drone litigation studies as examples to show how the approach being applied in the industry.