Chaos to Clarity with Semantic Inferencing for Python Source Code Snippets.

Our work investigates using semantic indicators in source code to generate descriptions of its functionality. We applied natural language processing (NLP) techniques to improve the performance of a neural machine translation (NMT) model on source code. We trained the model on 20,000 code-comment pairs from the Source Code Analysis Dataset (SCAD) using a custom Python source code tokenizer and a learning model with attention, RNNs, LSTMs, and auto-encoders. We evaluated the model’s performance on a hold-out dataset using several established translation metrics, and found that our methods outperform prior work with a mean Bilingual Evaluation Understudy (BLEU) score of 0.47.