REFINING NLP SEMANTIC MATCHES THROUGH DIALOGUE WITH LARGE LANGUAGE MODELS

Abstract

Advancements in digital technologies are driving innovation in Industry 4.0 (I4.0) ecosystems. The vision for Industry 4.0 by 2030 emphasizes sovereignty, sustainability, and interoperability, aiming to turn traditional value chains into dynamic networks. While we have successfully tackled many issues related to physical and syntactical interoperability, semantic interoperability remains a challenge. Using NLP-based methods for semantic matching provides a straightforward way to handle data with different meanings, but these methods often lack clarity, especially when important matches appear lower in the rankings. In this paper, we explore how we can enhance these semantic matches by using Large Language Models(LLMs), specifically the Retrieval-Augmented Generation(RAG) approach. Since clear explanations are important for users to select the best matches, we use LLMs not only to improve the rankings but also to provide understandable explanations and support further questions. Our experiments show that this makes semantic matching more user-friendly, helping users navigate complex semantic information more effectively.