Exploring the Naturalness of Buggy Code with Recurrent Neural Networks (1803.08793v1)

Abstract: Statistical LLMs are powerful tools which have been used for many tasks within natural language processing. Recently, they have been used for other sequential data such as source code.(Ray et al., 2015) showed that it is possible train an n-gram source code language mode, and use it to predict buggy lines in code by determining "unnatural" lines via entropy with respect to the LLM. In this work, we propose using a more advanced language modeling technique, Long Short-term Memory recurrent neural networks, to model source code and classify buggy lines based on entropy. We show that our method slightly outperforms an n-gram model in the buggy line classification task using AUC.