关键词: 信息抽取, 检索增强, 大语言模型, 提示工程, 少样本学习

Abstract: In serious task scenarios such as aeronautical data link command parsing in the military domain, information extraction systems must achieve both high accuracy and structural consistency to ensure the safety and reliability of mission execution. However, traditional extraction methods based on rules or few-shot prompting often suffer from severe hallucinations when dealing with complex semantics, flexible sentence structures, and highly constrained formats. To address this issue, a retrieval enhanced information extraction algorithm based on a large language model was developed. This algorithm combines retrieval enhancement technology with information extraction technology. First, a meta-language library oriented to the target information types was constructed to provide standardized extraction templates. Then, an efficient template retrieval process was achieved through an inverted index combined with keyword matching and semantic expansion strategies. Finally, prompt instructions were designed to guide the large language model to convert natural language into structured information, enabling highly controllable and accurate extraction of natural language commands in the aviation data-link scenario. Experimental results show that this algorithm is superior to existing information extraction algorithms, greatly reducing the illusions that exist in the extraction of information from large models. Meanwhile, the proposed framework exhibits strong domain independence and can be extended to other high-reliability extraction scenarios by reconstructing the meta-language library, providing a generalizable technical pathway for building intelligent and verifiable information processing systems.

Key words: information extraction, retrieval augmentation, Large Language Model (LLM), prompt engineering, few-shot learning