| Tasks that involve processing of natural language texts are key parts of software engineering processes, and computer-aided software engineering (CASE) should include natural-language-based techniques. My thesis shows that currently-available natural language processing technology can be used to automate language-oriented software engineering tasks.;Software testing is a particularly good area for building automated natural-language-based tools. A software tester compares the way a software system performs to an expectation, usually in the form of a text document, of how the system should perform. I show that this part of the testing process can be automated.;Previously, I built a proof-of-concept software testing tool, SIFT, that can extract testing-related information from software documentation and then use that information to generate tests for the software system. It uses the approach of extracting specific testable facts from semi-formatted documents. SIFT was developed as a proof-of-concept prototype and, while its results were promising, more functionality was required to demonstrate the feasibility of the approach. Most of the modules of the system could have been upgraded to improve its capabilities, and this thesis enhances SIFT by building an improved domain model for it.;Domain modeling is a cornerstone of knowledge-based and object-oriented software development, yet consensus on appropriate methods, formalisms, and techniques for domain models has not been reached. This thesis proposes a new domain modeling format that is directly based on the linguistic constructs of the sublanguage of the domain. In addition to analyzing the effectiveness of the domain model in SIFT, I test whether the domain model is also effective for non-language-based tools by enhancing a non-language- based software testing tool, TESTGEN, to use the domain model. Both SIFT and TESTGEN are improved by inclusion of the new domain model, although neither uses all of the information of the model. |
