Exception Handling Recommendation Based on Self-Attention Network

Abstract

Exception handling is widely applied to prevent unexpected runtime errors in software development. However, developers, especially novices, may not be able to write proper exception handling code, causing program crashes and reducing system reliability. To assist users to handle exceptions properly, we propose EHRSAN, a novel exception handling recommendation approach. It adopts the self-attention network to predict potential runtime exceptions in the programming context and recommend relevant code to handle the exceptions.