Harnessing Large Language Models for Precision Topic Extraction and Technology Patent Nomination: A GPT-centric Methodology
Franck Tshibanda Nkolongo, Saïd Echchakoui, Mehdi Adda · Procedia Computer Science · 2025
Topic modeling is widely used for analyzing large textual datasets, particularly in technology patent nomination. Traditional methods, such as Latent Dirichlet Allocation (LDA) and Non-Negative Matrix Factorization (NMF), often struggle with semantic comprehension, scalability, and adaptability. This study investigates Large Language Models (LLMs) to overcome these limitations. Leveraging their advanced language understanding, GPT models generate coherent and contextually relevant topics, outperforming conventional methods. They also excel in identifying emerging trends, enhancing patent nomination processes. Our findings demonstrate the potential of GPT-based approaches to streamline patent analysis and accelerate technological discovery.