Open source software documentation mining for quality assessment

Abstract

Besides source code, the fundamental source of information about Open Source Software lies in documentation, and other non source code files, like README, INSTALL, or HowTo files, commonly available in the software ecosystem. These documents, written in natural language, provide valuable information during the software development stage, but also in future maintenance and evolution tasks. DMOSS is a toolkit designed to systematically assess the quality of non source code text found in software packages. The toolkit handles a package as an attribute tree, and performs several tree traverse algorithms through a set of plugins, specialized in retrieving specific metrics from text, gathering information about the software. These metrics are later used to infer knowledge about the software, and composed together to build reports that assess the quality of specific features of the software. This paper discusses the motivations for this work, continues with a description of the toolkit implementation and design goals. Follows an example of its usage to process a software package, and the produced report. Finally some final remarks and trends for future work are presented. © 2013 Springer-Verlag.