Automatic Categorization Tool for Open Software Repositories

The world of Open Source software has demonstrated the remarkable appeal of communal software development. Large number of software projects can leverage, reuse, and coordinate their work through Internet and web-based technology. For example, SourceForge currently hosts about sixty thousand software systems. Similar strategies have been suggested for corporate software development, through notions like Corporate Source and Progressive Open Source [6, 7] When used in a corporate setting, infrastructures for project information sharing present new opportunities. For example, one would like to know all projects that have something in common, so that the project groups can collaborate and share their work. With thousands of projects, manually locating related projects can be difficult. Hence, we propose to use automatic software categorization to find clusters of related software projects, using only the source code from projects. Our experiments with a small set of C programs demonstrates potential for automatic categorization of software systems without human aid.