A role for chunking and fuzzy reasoning in a program comprehension and debugging tool

Abstract

We are applying artificial intelligence techniques to develop a tool called BUG-DOCTOR that assists software engineers with program comprehension and debugging. In this paper we describe two of BUG-DOCTOR's knowledge sources, the Chunker and the Plan Processor. The Chunker identifies candidate chunks in the target code using program analysis techniques and a set of heuristics. Candidate chunks map to higher level concepts, and have a signature which captures their major identifying characteristics. The Plan Processor uses a signature to retrieve a set of program plans from a Plan Library with features that are similar to those of the candidate chunk. Its fuzzy reasoner then ranks the retrieved plans. The plan chosen as most similar to the candidate chunk is used for program comprehension and debugging tasks that follow. We believe that this approach could lead to more scalable tools for program comprehension and debugging.