Leveraging natural language analysis of software: Achievements, challenges, and opportunities

Summary form only given. Studies continue to report that more time is spent reading, locating, and comprehending code than actually writing code. The increasing size and complexity of software systems makes it significantly more challenging for humans to perform maintenance tasks on software without automated and semi-automated tools to support them, especially in the error-prone tasks. Thus, software engineers increasingly rely on software engineering tools to automate maintenance tasks as much as possible. The program analyses that drive today's software engineering tools have historically focused on analyzing the program's data and control flow, dependencies, and other structural information about the program to uncover and prove program properties. Yet, a software system is more than just the source code and its structure. To build effective software tools, the underlying automated analyses need to use all the information available to make the tools as intelligent and useful as possible. By adapting natural language processing (NLP) to source code analysis, and integrating information retrieval (IR), NLP, and traditional program analyses, we can expect significant improvement in automated and semi-automated software engineering tools for many different software engineering tasks. In this talk, I will overview research in text analysis of software and discuss our achievements to date, the challenges faced in text analysis, and the opportunities for text analysis of software in the future.