An experience in using a tool for evaluating a large set of natural language requirements

Requirements analysis is an important phase in a software project. It is often performed in an informal way by specialists who review documents looking for ambiguities, technical inconsistencies and incompleteness. Automatic evaluation of Natural Language (NL) requirements documents has been proposed as a means to improve the quality of the system under development. We show how the tool QuARS Express, introduced in a quality analysis process, is able to manage complex and structured requirement documents containing metadata, and to produce an analysis report rich of categorized information that points out linguistic defects and indications about the writing style of NL requirements. In this paper we report our experience using this tool in the automatic analysis of a large collection of natural language requirements, produced inside the MODCONTROL project.