Proposal of Extracting State Variables and Values from Requirement Specifications in Japanese by using Dependency Analysis
Masanosuke Ohto, Hiroya Ii, Kozo Okano, Shinpei Ogata · Procedia Computer Science · 2021
A requirement specification for software is usually described in a natural language and thus may include sentences containing ambiguity and contradiction. Design errors due to ambiguous expressions and contradictions are often found later in the development process. These ambiguous sentences would force the developer to go back to the design process again. In order to prevent this kind of rework, a method of automatically converting a required specification written in Japanese to a state transition model is desired to help detect ambiguity and contradiction points of the specification. In previous studies, we have improved the extraction accuracy by refining the parsing rules. However, in order to parse a variety of sentences, it is necessary to set up new rules or refine the rules. In this paper, we propose a method of using a dependency analyzer in syntactic parsing. Using dependency analyzers eliminates the need for both of rule setting and refinement. Thus, the disadvantage would be eliminated. In addition, the effectiveness of the proposed method is reported by extracting and showing the elements necessary to create a state transition diagram.