Abstract
One of the major reasons for software project failure is poor requirements, so numerous requirement smells detection solutions are proposed. Critical appraisal of the proposed requirement fault detection methods is crucial for refining knowledge of requirement smells and developing new research ideas. The objective of this paper was to systematically review studies that focused on detecting requirement discrepancies in textual requirements. After applying inclusion and exclusion criteria and forward and backward snowball sampling techniques using database-specific search queries, 19 primary studies were selected. A deep analysis of the studies shows that classical NLP-based requirement smells detection techniques are the most commonly used ones and ambiguity is the requirement smell that has the utmost attention. Further investigation depicts the scarcity of open-access datasets, and tools employed to detect requirement faults. The review has also revealed there is no comprehensive definition and classification of requirement smells.