4.3.3 An Architecture for an Intelligent Requirements Elicitation and Assessment Assistant

Abstract

Requirements engineering is crucial to the success of the system engineering process. Unfortunately, while current generation tools provide good support to requirements capture, they offer only modest assistance for cognitive tasks such as requirements assessment. This paper opens by introducing the attributes needed to define a requirement and the qualities of good requirements and requirements sets. An argument is then presented for the inclusion of artificial intelligence to perform requirements analysis and this is followed by a review of intelligent requirement analysis tools identifying the need for enhancements. From this basis, the paper describes the architecture of an intelligent requirements elicitation and assessment assistant that is currently under construction.

6.1.3 A Requirements Assessment Architecture that Combines Natural Language Parsing And Artificial Intelligence

“…Educated humans are innately able to convey complex concepts through natural language but it remains a challenging medium for computer science because words and phrases can be ambiguous and meaning is notoriously difficult to convey concisely in the absence of contextual information and real-world knowledge. Our philosophy to ease this significant problem is to constrain requirements language use via what is known as a "contextfree" grammar (Scott and Cook, 2003).…”

“…Several strategies have been previously pursued to perform computer-aided requirements assessment. Scott and Cook (2003) reviewed available techniques and identified that Case-Based Reasoning (CBR), while untried in this problem domain, showed the most promise. The reason for selecting CBR is that it mimics the pattern matching process that human requirements engineers employ when assessing requirements (Cook, 1991: p45).…”

“…Thus their use significantly aids shared meaning. We have used these standards as the template from which to generate a context-free grammar that reduces processing to a manageable level (Scott and Cook, 2003).…”

“…The first tier examines the individual requirement statements as they undergo parsing to identify the grammatical structure. During the examination, aspects of requirement that do not correspond with the grammar or the word usage defined in Scott & Cook, 2003, which derives from recognized best practice, are brought to the attention of the user. (An example of the latter is the avoidance of the use of pronouns and adverbs when writing requirements, as recommended in the INCOSE SE Handbook, (2000)).…”

“…This observation was the impetus to develop a wizard that elicits specific information and uses it to write requirement statements. This tool is described in Scott and Cook (2004).…”