This paper introduces a novel methodology for constructing technology-design trees based on text analysis to support the discovery of technology opportunities. The methodology employs KeyBERT(Keyword extraction with Bidirectional Encoder Representatio...
This paper introduces a novel methodology for constructing technology-design trees based on text analysis to support the discovery of technology opportunities. The methodology employs KeyBERT(Keyword extraction with Bidirectional Encoder Representations from Transformers) model to extract meaningful keywords from patents and design rights documents, thus facilitating the analysis of semantic similarities and categorization into similar technological and design groups. Further, it utilizes Cooperative Patent Classification (CPC) and Locarno Classification (LOC) to build technology trees and design trees, respectively, through the analysis of technical and product similarities. Additionally, the methodology employs structural similarity index mapping (SSIM) on patent and design right drawings to validate the constructed trees. The approach is distinguished by its minimal reliance on expert intervention, enhancing the efficiency and scalability of technology opportunity discovery processes.