Remodularization analysis using semantic clustering

Abstract

As software evolves, its structure inevitably gets harder to understand and maintain, unless explicit effort is done to improve it. To tackle this problem, the common recommendation consists in organizing the software structure into modules. This process is often performed by optimizing the value of cohesion and coupling metrics, for example.However, recent work question the usefulness of structural metrics. More specifically, structural metrics do not seem to fully express the architectural improvement resulted from software maintenance. In this master dissertation, we follow an existing methodology to assess software metrics regarding real remodularization cases, i.e., in whichthe maintenance was performed by the softwares architects. We use a set of recently proposed metrics, called conceptual metrics, which consider textual similarity between software artifacts. To accomplish this task, we report an experiment on using Semantic Clustering to evaluate software remodularizations. Semantic Clustering is an approach that relies on information retrieval and clustering techniques to extract sets of similar classes in a system according to their vocabularies. In fact, we adapted Semantic Clustering to support remodularization analysis. We then evaluate our adaptation using six real-world remodularizations of four software systems. As a result, we conclude that Semantic Clustering and conceptual metrics can be used to express and explain the intention of architects when performing common modularization operations, suchas module decomposition.