Utilizing Keywords in Source Code to Improve Code Summarization

The source code summary is a short description in natural language that describes its functionality. The purpose of code summaries is to help developers understand the codes in software projects, and the automatic generation of source code summaries plays an important role in many software engineering activities. Inspired by observations that keyword information is often used to guide text summarization, we propose a novel code summarization model called CosKeywords, which can focus on the key information in the source code during the encoding process. To evaluate CosKeywords, we conducted extensive experiments on two existing datasets. Experiment results show that CosKeywords can outperform the state-of-the-art models on all datasets in terms of BLEU4 and METEOR.