SDF-FuzzIA: a DSS based on LLMs and Fuzzy Ontologies

Abstract

The proliferation of AI algorithms based on LLMs has highlighted, alongside considerable potential, questions regarding their adaptability to the generality of solvable problems. Among these are explainability, introspectiveness, and precision when applied to information that is not strictly linguistic in nature. This work contains a description of SDF-FuzzIA, the AI module of the SDF project for consolidation and exploration of geothematic data. The purpose of SDF-FuzzIA is to integrate a GPLMM with a Fuzzy ontology, ensuring that the suggestions which constitute the core of its DSS are grounded in concrete facts. As SDF is a work in progress, results will not be presented; rather, design and implementation choices will be discussed.