Extracting software requirements-related information from online news using DomText-WMDS

Abstract

Currently, there are not many studies that assess software requirements extraction from non-software artifacts. Most of the research in these related areas are focuses on software artifacts such as project descriptions or user reviews as a source of requirements extraction. This research aims to identify relevant information to the software requirements from online news using the vector space model. This software requirements-related information can assist systems analysts in discovering the problem domain based on the lesson learned presented by stakeholders in online news. This research proposes DomText-WMDS to extract requirements-related information from online news. We used online news and public software requirements specification dataset to develop software-specific vocabulary using domain specificity technique. Then we expanded the specific vocabulary software to obtain more comprehensive results by building vector space model from online news documents. This updated version of software-specific vocabulary can be used for basic filtering of software requirements-related information that previously extracted using the part-of-speech (POS) chunking. This study improved the performance for extracting software requirements-related information, with precision and recall 61.09% and 60.66% compared to domain specificity approach that only manages to obtain 43.34% and 40.78%.