Identification of emerging technology topics (ETTs) using BERT-based model and sematic analysis: a perspective of multiple-field characteristics of patented inventions (MFCOPIs)

The proliferation of large language models (LLMs) has significantly expanded the landscape of research on technology opportunity identification. However, there remains a crucial need to enhance the accuracy and interpretability of results obtained through emerging technology topic identification. In this paper, we present a novel approach that leverages a BERT-based model and semantic analysis to identify emerging technology topics (ETTs) from the perspective of multiple-field characteristics of patented inventions (MFCOPIs). By utilizing a unique dataset encompassing MFCOPI, our methodology emphasizes an increased proportion of novel technical processes in the analysis content while mitigating the interference of redundant technical information. To enhance the interpretability of recognition results, our proposed model employs the BERT model for detecting potential content similarities in inventive characteristics and incorporates semantic structure analysis to expand the technical process content. We empirically validate our model by employing nanotechnology as a case study, demonstrating its effectiveness and accuracy. Through our research, we extend the existing methodologies for recognizing emerging technology, ultimately elevating the quality of recognition results.