Automatic Generation of Test Cases Using Document Analysis Techniques

Satoshi Masuda, Tohru Matsuodani, Kazuhiko Tsuda

In software maintenance, software testing consumes 55% of the total software maintenance work. The problem is how to reduce the software testing work while still insuring high quality software.  Some solutions involve software execution automation tools, outsourcing the testing tasks at lower labor rates. Such solutions still depend upon individual skills in generation of the test cases. In contrast, we focused on generation of test cases rather than the skills and developed a method for the automatic generation of test cases by using our natural language document analysis techniques which use text parsers for extracting and complementing parameter values from documents. We applied the method to Internet banking system maintenance projects and insurance system maintenance projects.  In this paper, we discuss our method and techniques for automatic generation of test cases and their use in these industry case studies.  Our document analysis tool helped automatically generate 95% of the required test cases from the design documents. The work of creating test cases was reduced by 48% in our case studies.