Automatic Documentation Generation via Source Code Summarization

Programmers need software documentation. However, documentation is expensive to produce and maintain, and often becomes outdated over time. Programmers often lack the time and resources to write documentation. Therefore, automated solutions are desirable. Designers of automatic documentation tools are limited because there is not yet a clear understanding of what characteristics are important to generating high quality summaries. I propose three specific research objectives to improving automatic documentation generation. I propose to study the similarity between source code and summary. Second, I propose studying whether or not including contextual information about source code improves summary quality. Finally, I propose to study the problem of similarity in source code structure and source code documentation. This paper discusses my work on these three objectives towards my Ph.D. dissertation, including my preliminary and proposed work.