Ambiguity Detection and Improvement for Malay Requirements Specification: A Systematic Review

Abstract

Malaysian public sectors have invested billions in digitizing systems. Electronic government efforts created much software. Our informal interview taught us that many software projects encountered delays, and several failed. One of the main contributions of software failure is ambiguity in requirements specification (RS). Ambiguity is a familiar requirement smell that causes misinterpretation. Thus, we seek to devise a technique for detecting and improving ambiguous RS in the Malaysian public sector. One of our challenges is that the Malaysian public sector RS is developed in Malay, and most available techniques support English and other major languages. Hence, this paper investigates the automated and semi-automated techniques to detect and improve ambiguous RS. Following the standard guidelines for systematic mapping, review, snowballing, and quality assessment, we studied works from 2010 to 2022 on ambiguity detection and improvement techniques. We chose 42 articles as primary studies from 2,549. As a result, Natural Language Processing (NLP) and machine learning (ML) are the most promising techniques for automated and semi-automated ambiguous detection models. Furthermore, the ambiguous improvement technique began using deep learning (DL) in 2019. However, most proposed tools are still in the validation phase and are not widely employed, implying that tool development and validation research are progressing slowly. Apart from the generic linguistic context of RS, some research focuses on industrial domain-based RS. Our study shows that additional strategies have been developed to overcome RS-related issues.