RNTI

MODULAD
Vers des LLM moins gourmands pour interroger des graphes de connaissances en langue naturelle
In EGC 2024, vol. RNTI-E-40, pp.473-480
Abstract
Knowledge graphs are widely used to represent a domain. In particular, they enable data from different sources to be brought together and stored in a structured way. However, accessing these resources can be complex, as we need to master a query language that is compatible with the formalism used to represent the knowledge graph. In this article, we propose a question-answering system using Large Language Models to query knowledge graphs in natural language. We compare different techniques in order to best combine frugality and performance with the aim of controlling the resources used. Our work is based on a knowledge graph describing the Sustainable Development Goals defined by the United Nations.