LLM-KBQA: A Knowledge Base Question Answering Framework Based on Large Language Models

This chapter delves into the challenges of filtering and analyzing vast amounts of data in the current era, highlighting the role of knowledge graphs in providing structured data representation. It introduces the LLM-KBQA framework, which leverages large language models to enhance natural language understanding and integrates with knowledge bases to transform queries into graph statements while preserving semantics. The framework consists of a semantic parsing module fine-tuned on large models and an information retrieval module based on coarse and fine granularity. The chapter also discusses related work in the field, including traditional KBQA methods and advancements in deep learning. Experiments on the KgCLUE dataset validate the framework's accuracy and efficiency, with contributions including a semantic parsing module, a combined KBQA framework, and verification of its effectiveness and scalability. The chapter concludes with a summary of the framework's performance and potential future improvements.