Conceptual

LeSR: LLM-Enhanced Symbolic Reasoning for Knowledge Base Completion

A framework that fuses large language models with rule-based reasoning to complete incomplete knowledge bases. A Subgraph Extractor samples KB subgraphs; an LLM Proposer generates diverse candidate logical rules for inferring missing facts from them; and a symbolic Rule Reasoner scores and refines these rules against the KB to keep only the most significant, suppressing LLM hallucinations. The result couples the LLM's semantic richness and rule diversity with the verifiability of symbolic reasoning, generalizing across diverse KB datasets.