Date of Award
Master of Science (MS)
Mocko, Gregory M
Summers , Joshua
Venhovens , Paul
In engineering design, the needs of the customer are expressed through engineering requirement statements. These requirement statements are often expressed using natural language because they are easily created and read. However, there are several problems associated with natural language requirements including but not limited to ambiguity, incompleteness, understandability, testability and over specificity. Several representation and analysis tools have been proposed to address these problems within a requirement statement. These tools include formal languages, such as UML and SysML, requirement management tools, such as IBM Telelogic Doors, and natural language processors such as QuARS. These tools assist in the systematic elicitation and creation of requirements, improve requirement visibility and traceability, and provide a central repository for shared access. However, these tools do not prescribe a formal representation of a requirement and its elements. The effectiveness of these tools can be greatly improved with a formalized syntax for expressing engineering requirements.
The research presented in this thesis examines engineering requirements from a linguistic viewpoint and leads to a formalized syntax based on parts of speech, grammatical functions, and sentence structure. Specifically, a requirement statement is decomposed into four syntactical elements: artifact, necessity, function, and condition. Further, grammar and linguistics provide the basis for requirements classification into functional or non-functional and qualitative or quantitative requirements. Finally, the deficiencies in current natural language requirements such as incompleteness, understandability, ambiguity, and specificity, are identified through the formal syntax and grammatical rules. The requirements syntax and analysis method are demonstrated on 110 requirements from the Family of Medium Tactical Vehicles (FMTV). Using the syntax and analysis method proposed, the count of incomplete requirements, percentages of function and non-functional requirements, and specificity of the requirement statements in the document were determined. Identifying such requirement measures will help to improve the expressiveness of requirement statements and help to identify if appropriate requirements are being authored for the different stages of design (i.e. conceptual, embodiment, detailed). To further improve the analysis method proposed, more quality attributes of requirement statements have to be addressed such as ambiguity and traceability. The end goal is to develop a syntax and analysis method that addresses all quality attributes of a requirement statement that is not empirically based but rule based.
Lamar, Carl, "Linguistic Analysis of Natural Language Engineering Requirements" (2009). All Theses. 671.