Pinpointing Ambiguity and Incompleteness in Requirements Engineering via Information Visualization and NLP

Abstract

[Context and motivation] Identifying requirements defects such as ambiguity and incompleteness is an important and challenging task in requirements engineering (RE). [Question/Problem] We investigate whether combining humans’ cognitive and analytical capabilities with automated reasoning is a viable method to support the identification of requirements quality defects. [Principalideas/results] We propose a tool-supported approach for pinpointing terminological ambiguities between viewpoints as well as missing requirements. To do so, we blend natural language processing (conceptual model extraction and semantic similarity) with information visualization techniques that help interpret the type of defect. [Contribution] Our approach is a step forward toward the identification of ambiguity and incompleteness in a set of requirements, still an open issue in RE. A quasi-experiment with students, aimed to assess whether our tool delivers higher accuracy than manual inspection, suggests a significantly higher recall but does not reveal significant differences in precision.