Towards Software Component Procurement Automation with Latent Semantic Analysis

One of the first steps of component procurement is the identification of required component features in large repositories of existing components. On the highest level of abstraction, component requirements as well as component descriptions are usually written in natural language. Therefore, we can reformulate component procurement as a text analysis problem and apply latent semantic analysis for automatically identifying suitable existing components in large repositories, based on the descriptions of required component features. In this article, we motivate our choice of this technique for feature identification, describe how it can be applied to feature tracing problems, and discuss the results that we achieved with the application of this technique in a number of case studies.