Tree Kernels to Support Formal Methods-Based Testing of Evolving Specifications

Abstract

Tree Kernels (TKs) are a family of functions measuring the similarity between two tree-structured objects. TKs have been successfully employed in several fields of AI, including Natural Language Processing and Software Engineering (e.g.: software testing and code clone detection). A recent research line has proved that the information about source code changes captured by TKs can fruitfully be applied to select and prioritize test cases in Regression Testing, a crucial activity in modern software development processes. We suggest that a similar approach can be adopted also in the field of formal specification of safety-critical systems, whenever a structured language (e.g., a variant of Statecharts, hierarchical automata or the Promela description language) is adopted for the specification task and test cases are (semi-)automatically generated from the specifications. In evolutionary approaches to specification development, the information on changes between two consecutive versions of the specification can aid in focusing the activity of test generation and their execution on the specification modules that were mainly affected by the specification changes.