| Concept location, the problem of finding the implementation of human oriented concepts in the source code, is a fundamental problem that lies at the heart of software comprehension. Among the most popular concept location techniques is grep, a well known example of a string pattern matching UNIX utility. In previous work, grep was used along with ontology fragments (grepOF) in order to leverage the power of partial domain comprehension in large open source software. In this thesis, we explore the use of ontology fragments that record partial program comprehension. We perform concept location using these ontology fragments with information retrieval based queries and we call this technique IROF. We study the effectiveness of ontology fragments in capturing programmers' knowledge during concept location and quantify the learned knowledge to infer any performance traits about IROF. We performed a case study on Eclipse, locating the concepts described in defect reports, and compared this new technique to grepOF. It was found that, using IROF, programmers' knowledge varied significantly during the concept location process. Moreover, the increase in programmers' knowledge represented by the ontology fragments was significantly different while using the IROF and grepOF technique; and programmers needed to acquire less extra knowledge using IROF. |
