Research Article
Knowledge Extraction for Commercial Aircraft Process Specification Based on LLM Generation Rules
@INPROCEEDINGS{10.1007/978-3-031-96944-7_2, author={Wenhao Xu and You Song and Ruiqiang Lyu and Bencheng Cui and Qiao Zheng and Baoguo Chen}, title={Knowledge Extraction for Commercial Aircraft Process Specification Based on LLM Generation Rules}, proceedings={Computer and Information Security. First World Conference, WCCIS 2024, Kuala Lumpur, Malaysia, September 20--22, 2024, Proceedings}, proceedings_a={WCCIS}, year={2026}, month={11}, keywords={Commercial Aircraft Process Specification Knowledge Extraction LLM (Large Language Model) Rule Generation}, doi={10.1007/978-3-031-96944-7_2} }- Wenhao Xu
You Song
Ruiqiang Lyu
Bencheng Cui
Qiao Zheng
Baoguo Chen
Year: 2026
Knowledge Extraction for Commercial Aircraft Process Specification Based on LLM Generation Rules
WCCIS
Springer
DOI: 10.1007/978-3-031-96944-7_2
Abstract
In the current scenario, process specifications for the design of commercial aircraft parts primarily exist in the form of unstructured PDF documents. As a result, the cost of manual data structuring becomes excessively high. Existing technologies fail to meet the demands for efficiently building structured knowledge bases in aviation manufacturing. This paper proposes and implements a knowledge extraction approach for commercial aircraft process specifications. The approach can effectively extract knowledge from PDF format process specifications, converting it into structured data that is easier to manage and retrieve. It uses LLM to generate knowledge extraction rules and combines the rules to achieve knowledge recognition, parsing and extraction. LLM learns the rules through the provided examples, and subsequently applies them to generate new extraction rules from unseen process specifications effectively. These new extraction rules are used as the input the extraction algorithm, thus constraining knowledge extraction. The approach was performed on 26 aviation manufacturing process specifications. After expert review, the knowledge extraction accuracy of the approach was found to be 94.4%. Our study not only provides strong support for the construction of aviation knowledge base, but also has certain application potential for similar data structuring problems in other fields.