Toward Summary Extraction Method for Functional Topic

Understanding the function of software code is the basis for software reuse. Topic modeling technologies can mine functional topics from source code and help developers comprehend the functional concerns about a software system and the corresponding implementations in source code. However, lacking clear explanations makes these functional topics hard to be understood by the developers. Furthermore, giving proper interpretations to these topics manually is time consuming. In this paper, we firstly use self-parameter-optimizing Latent Dirichlet Allocation (LDA) technology to mine the functional topic for the software. Then we put forward an approach for calculating the relevancy between functional topics and software documents. The LexRank technology is used to automatically generate summary for those topics from software documents such as user manuals, pairs of question and answer, mailing lists and so on. The experiment results indicate that these summaries improve developers understanding the function of software code.