Extraction of non-functional requirement using semantic similarity distance

Abstract

Functional and non-functional requirements are important equally in software development. Usually, the requirements are expressed in natural languages. The functional and non-functional requirements are written inter-mixed in software requirement document. The extraction of requirement from the software requirement document is a challenging task. Most of the recent studies adopted a supervised learning approach for the extraction of non-functional requirements. However, there is a drawback of supervised learning such as training of model and retrain if the domain changed. The proposed approach manipulates the textual semantic of functional requirements to identify the non-functional requirements. The semantic similarity is calculated based on co-occurrence of patterns in large human knowledge repositories of Wikipedia. This study finds the similarity distance between the popular indicator keywords and requirement statements to identify the type of non-functional requirement. The proposed approach is applied to PROMISE “NFR dataset.” The performance of the proposed approach is measured in terms of precision, recall and F-measure. Furthermore, the research applies three pre-processing approaches (traditional, part of speech tagging and word augmentation) to increase the performance of NFR extraction. The proposed approach outperforms the results of existing studies.