Enhancing Complex Linguistic Tasks Resolution Through Fine-Tuning LLMs, RAG and Knowledge Graphs (Short Paper)
摘要
Given the synergy between Large Language Models (LLMs) and Knowledge Graphs (KGs), we introduce a pipeline to tackle complex linguistic tasks, which we are experimenting in the legal domain. While LLMs offer unprecedented generative capabilities, their reliance on sub-symbolic processing can lead to fallacious outcomes. Our methodology introduces an advanced Retrieval Augmented Generation (RAG) pipeline, enriched with two KGs and optimized LLMs, promising to enhance the resolution of complex linguistic tasks. Through KG construction based on prompt engineering techniques and iterative fine-tuning, we transcend the limitations of conventional LLMs.