Detecting Ambiguity in Requirements Analysis Using Mamdani Fuzzy Inference

Abstract

Natural language is the most common way to specify requirements during elicitation of requirements as stakeholders can better specify the services they want from a particular system. However, it is arguable that requirements gathered in natural language is free from error especially ambiguity. Ambiguity in requirements can cause requirement engineers or system analysts to perceive the requirements according to their understanding instead of stakeholders understanding. This study attempts to detect ambiguity mainly vagueness as early as possible using Mamdani fuzzy inference when analyzing requirements. Dataset used in this study comprises raw requirements that are still in natural language form. In order to create fuzzy rules, the analysis of the requirements in natural language involves the process of capturing the text patterns of the requirements. The results show that it is possible to use Mamdani fuzzy inference that can detect ambiguity in requirements analysis phase.