Towards Recovering Architectural Concepts Using Latent Semantic Indexing

Abstract

In order to address the problem of locating high-level concepts in source code we propose to use an advanced information retrieval method to exploit linguistic information found in source code, such as variable names and comments. Our technique is based on latent semantic indexing (LSI) which is also used in today's search engines. Applying LSI to source code, however, is not straightforward. Our approach therefore not only includes LSI, but also several other algorithms and methods. We discuss the algorithms and methods that turned out to be useful and provide an overview of their effects using the results obtained from a case study at Philips Healthcare.