Generative AI approach for inventive process visualisation – enhancing human-AI hybrid understanding and comparing of patents
Amy J.C. Trappey, Chun-Yi Wu, Y.-C. Lee, L.P. Hung
Abstract
Generative AI (GenAI) and Large Language Models (LLMs) have significantly advanced the generation of text documents (e.g. summary, report, essay, etc.) for various purposes. This study aims to apply state-of-the-art GenAI and LLM technologies, together with BERT–based extraction, for reading and interpreting patent semantics, extracting their inventive processes for human visualisation and system’s intelligent comparison among patents, which are considered critical and time-consuming works during patent-related activities, e.g. writing, applying, rebuttal, and/or litigating. This research leverages a human-intelligence-driven AI system’s design, integrating human expertise to fine-tune the language model for better contextual understanding and logical reasoning in process diagram generation. Moreover, this paper demonstrates the gen-AI-based inventive process visualisation and similarity analysis of 6G related telecommunication patents. The approach enables the accurate visualisation of logical steps described in complex sequences of patent texts. The similarity analysis sub-system also incorporates an algorithmic framework for detecting similar textual nodes among patent processes, which supports investigations of patentability, validity, and potential infringement of target patents. By combining generative AI and language modelling with human intelligence for inventive process visualisation and analysis, this study lays the foundation for advancing systematic innovation and patentability and strengthening intellectual property (IP) protection.