A Study to Identify and Classify Ambiguities in User Stories Using Machine Learning
Ambiguity in natural language; User Stories; Machine Learning
Ambiguity in requirements writing is one of the most common defects found in requirements documents. There are a variety of concepts about what is ambiguity in requirements and to identify ambiguity one must better understand each concept. Ambiguity can compromise the quality of User Stories and can be present in requirements written in natural language. In the literature, there are few studies that investigate the potential of Machine Learning algorithms to classify ambiguity in User Stories. This dissertation aims to propose an approach to identify and classify ambiguity in User Stories through the use of Machine Learning algorithms. Thus, a checklist was developed to help in the identification of ambiguities in User Stories and a Machine Learning approach will be used using two algorithms: (i) Support Vector Machine; (ii) Random Forest. Each model generated by the algorithm will be evaluated and compared.