A Requirements Assessment Architecture that Combines Natural Language Parsing and Artificial Intelligence

While the need for good requirements is well understood, and there many sources that describe what constitutes a set of good requirements, poor requirements continue to plague projects. In recent years tools have become available to assist in requirements management and their rapid adoption throughout industry attests to their efficacy in data handling and display. Contemporary requirements engineering tools, however, provide few assessment capabilities and no semantic processing. This paper addresses that deficiency while building on the capability of existing tools. The paper opens by describing the philosophy we used to tackle the problem: emulate human reasoning; use several tiers of analysis; parse the requirement against a purposedesigned a context-free grammar; and employ case-based reasoning to assess both individual requirements and the total requirements set. The paper concludes with descriptions of the architecture of a tool termed REWARD, its operational concept, and its implementation.