Document Information Extraction in Engineering Reports Through Prompt Engineering with Large Language Models

This article delves into the application of Prompt Engineering to enhance the document information extraction capabilities of large language models (LLMs), with a focus on optimizing the performance of advanced models such as chatglm4 and qwen 2.5 through meticulous prompt design. A structured approach encompassing task decomposition, evaluation metric establishment, and prompt instruction design is proposed. Experiments conducted on engineering technical reports reveal that Prompt Engineering markedly improves the accuracy and robustness of information extraction, with the optimized models demonstrating superior performance in pinpointing and extracting crucial information. These findings underscore the potential of Prompt Engineering in propelling AI technology and its applications forward, particularly in key information extraction tasks pertaining to engineering technical reports.