Syntax-Tree Similarity for Test-Case Derivability in Software Requirements

Abstract

Software testing has been important for software engineering to contribute to developing high-quality software. Decision table testing is a general technique to derive test cases with information on conditions and actions from software requirements. Deriving conditions and actions from requirements is key for efficient decision table testing. This paper proposes and evaluates a syntax-tree similarity method for test-case derivability in software requirements. We define the syntax-tree similarity technique used in our method as selecting test-case-derivable sentences from requirements at pre-processing. The syntax tree is defined as divided into sub-trees that consist of a root to each leaf. The syntax-tree similarity technique calculates the similarity between each sentence in the requirements and test-case-derivable sentence. The method involves natural language processing to select test-case-derivable sentences from the requirements on the basis of syntax-tree similarity then determines conditions and actions through dependency and case analyses. After selecting requirements by syntax-tree similarity, our method derives conditions and actions from the requirements by the deriving rules we define. Experiments revealed that the F-measure of the accuracy of the derived conditions and actions increased 16% from that reported in prior work. The results from case studies further indicate the effectiveness of our method.