Embedding API dependency graph for neural code generation

The problem of code generation from textual program descriptions has long been viewed as a grand challenge in software engineering. In recent years, many deep learning based approaches have proposed, which can generate sequence description. However, the existing ignore global relationships among API methods, are important for understanding usage APIs. this paper, we propose to model dependencies methods an dependency graph (ADG) and incorporate embedding into sequence-to-sequence (Seq2Seq) model. addition encoder-decoder structure, new module named “embedder” is introduced. way, decoder utilize both structural description predict target code. We conduct extensive experiments on three public datasets two programming languages (Python Java). Our proposed approach, called ADG-Seq2Seq, yields significant improvements over state-of-the-art maintains its performance length increases. Extensive ablation tests show that ADG effective outperforms baselines.