Natural language processing for software requirement specifications

Natural language processing for software requirement specifications

연구 분야: Artificial Intelligence

학회: CASCON '21: Proceedings of the 31st Annual International Conference on Computer Science and Software Engineering

Software Requirement Specifications (SRS) describe the functionality and expected performance for software products. It is one of the most important documents in the software development life cycle process that affects all the subsequent phases in product development. To manage the software requirements, requirement management tools such as IBM DOORS have been used to facilitate the communication, collaboration and verification throughout the process. Such tools generate a large amount of textual data, which can be processed through various NLP techniques to automate the software development processes. In this regard, the amalgamation of NLP and requirement engineering can catalyze the process of requirement classification (e.g., classifying requirements as functional and non-functional requirements), categorization of software documents, and topic modeling over SRS documents. Moreover, NLP-based approaches can be expanded to analyze the deeper semantics of software requirements for various purposes such as information extraction, automation of UML diagrams, AI-enabled software requirement analysis, defect and error detection in SRS, and solving the natural language ambiguities present in the requirements. In this workshop, university and IBM researchers will give insights into various existing NLP methodologies applied to SRS documents, e.g., for detecting the ambiguities and defects in the SRS documents. The prospects of various NLP-based automation techniques for conflict/duplicate detection as well as other NLP applications in requirement engineering will also be presented.