Program comment generation with improved distributed representation by Seq2seq model using parse tree information

Comments in a program's source code are important for understanding the program. Understanding the logical flow and overall procedure of the programs is important as the next step especially for beginners learning programming language, and it is inferred that appropriate comments on the source code can support it. In this study, we generate comments for source code using a distributed representation of line dependencies constructed with Word2Vec and using parse tree information obtained from the source code as input. Also, we generate comments not only for each line of source code but also for blocks, which are the logical units of processing.