Automatic source code summarization with graph attention networks

Source code summarization aims to generate concise descriptions for snippets in a natural language, thereby facilitates program comprehension and software maintenance. In this paper, we propose novel approach– GSCS –to automatically summaries Java methods, which leverages both semantic structural information of the snippets. To end, utilizes Graph Attention Networks process tokenized abstract syntax tree program, employ multi-head attention mechanism learn node features diverse representation sub-spaces, aggregate by assigning different weights its neighbor nodes. further harnesses an additional RNN-based sequence model obtain optimizes structure combining output with transformed embedding layer. We evaluate our approach on two widely-adopted datasets; experiment results confirm that outperforms state-of-the-art baselines. • Highlight importance Design neural network architecture task. New data wrangling methods summarization. Provide open-source implementation datasets