Code summarization, aiming to generate succinctnatural language description of source code, is extremely useful for code search and code comprehension. It has played an important role in softwaremaintenance and evolution. Previous approachesgenerate summaries by retrieving summaries fromsimilar code snippets. However, these approachesheavily rely on whether similar code snippets canbe retrieved, how similar the snippets are, and failto capture the API knowledge in the source code,which carries vital information about the functionality of the source code. In this paper, we propose anovel approach, named TL-CodeSum, which successfully uses API knowledge learned in a different but related task to code summarization. Experiments on large-scale real-world industry Javaprojects indicate that our approach is effective andoutperforms the state-of-the-art in code summarization.
Available at: http://works.bepress.com/david_lo/331/