Program comprehension is a current and a practical software engineering problem. Software engineers might try to comprehend the code by reading all of the source code--usually a substantial task. This paper highlights the idea of code summation as an alternative to time-consuming and subjective comprehension of the entire program.

The focus of the paper is an automated code summation tool applied during the summary building process. An interesting aspect of the authors’ contribution is the inclusion of a grounded theory approach (via eye-tracking studies) that provides empirical evidence regarding the kind of keywords the programmers would use to build the summaries. Through eye movement and heat paths, the authors provide new evidence (for example, a method’s signature is most critical), contradict previous beliefs regarding programmer behavior (for example, control flow is not as important as it has been suggested in the past), and use the empirical evidence to develop an automated tool for building program summaries the way professional programmers would develop. The automated tool developed by the authors extracts the keywords and the number of times a certain keyword occurs in the code. Then, these keywords are weighed based upon the place in the code (for example, signature or control flow) where the keyword occurs. These weights mirror the way the programmers would have read these keywords and are validated through the eye-tracking study.

This work would interest researchers trying to further improve program comprehension and industry professionals who can use the tool (and provide feedback) to improve the source code summarization in their organizations. This work is relevant and generalizable due to the inclusion of behavioral research methods to understand the program behavior during the source code summarization, and then empirically validating the tool (that was subsequently developed) through experimentation with professional software developers.