Detecting Code Comment Inconsistency using Siamese Recurrent Network

Comments are the internal documentation of corresponding code blocks, which are essential to understand and maintain a software. In large scale software development, developers need to analyze existing codes, where comments assist better readability. In practice, developers commonly ignore comments' updating with respect to changing codes, which leads the code comment inconsistency. Traditionally researchers detect these inconsistencies based on code-comment tokens. However, sequence ordering in codecomments is ignored in existing solution, as a result inconsistencies for invalid sequences of codes and comments are neglected. This paper solves these inconsistencies using siamese recurrent network which uses word tokens in codes and comments as well as their sequences in corresponding codes or comments. Proposed approach has been evaluated with a benchmark dataset,…