Improvement of Extraction Accuracy of Ambiguous Expression in Requirement Specifications by Using Large Language Models

The purpose of this study is to propose a more accurate method for detecting ambiguous expressions in requirement specifications using large language models. When we fed the document files of requirement specifications as input data to a large language model and attempted to extract ambiguous expressions using only simple prompt instructions, we were able to extract some ambiguous expressions, but the accuracy was not high. Therefore, we explored a method combining this approach with existing dictionary-based techniques, using the large language model to complement the existing methods. As a result, this combined approach showed improved accuracy, suggesting the usefulness of this approach.