Elena Barzizza, Nicolò Biasetton, Giorgio Caligiuri, Daniele Pennisi, Luigi Salmaso

Abstract: In today's competitive market, staying abreast of concurrent product characteristics is imperative for companies to maintain a competitive edge. Leveraging open-source Large Language Models (LLMs) presents a promising avenue for efficient and comprehensive analysis. This paper delves into the current landscape of commercial and open-source LLMs, assessing their potential for analyzing product characteristics. Additionally, it explores the feasibility of fine-tuning these models, including the utilization of Retrieval Augmented Generation (RAG), to enhance response accuracy and depth. Through evaluation, Mistral 7B emerges as a suitable open-source model for implementation, balancing performance with computational constraints. Furthermore, it outlines the process of refining LLMs using proprietary data, market intelligence, patent insights, and data gathered from web scraping to develop a comprehensive analytical tool for R&D purposes. This tool enables efficient extraction, analysis, and visualization of pertinent information, empowering decision-makers to steer innovation effectively.

Keywords: Large Language Models, R&D, Product Characteristics, Smart data

Date Published: March 4, 2025
DOI: 10.11159/jmids.2025.003

View Article