Using Paragraph Vectors to improve our existing code review assisting tool-CRUSO

TL;DR

This paper enhances a code review tool by leveraging Paragraph Vectors trained on StackOverflow data to accurately and efficiently predict code defectiveness across multiple programming languages.

Contribution

The paper introduces SOpostsDB and CRUSO-P, novel datasets and systems that significantly improve speed, memory efficiency, and accuracy in code defectiveness prediction.

Findings

CRUSO-P reduces response time by 97.82%.

CRUSO-P decreases storage needs by 99.15%.

CRUSO-P achieves 99.6% accuracy on C code.

Abstract

Code reviews are one of the effective methods to estimate defectiveness in source code. However, the existing methods are dependent on experts or inefficient. In this paper, we improve the performance (in terms of speed and memory usage) of our existing code review assisting tool--CRUSO. The central idea of the approach is to estimate the defectiveness for an input source code by using the defectiveness score of similar code fragments present in various StackOverflow (SO) posts. The significant contributions of our paper are i) SOpostsDB: a dataset containing the PVA vectors and the SO posts information, ii) CRUSO-P: a code review assisting system based on PVA models trained on \emph{SOpostsDB}. For a given input source code, CRUSO-P labels it as {Likely to be defective, Unlikely to be defective, Unpredictable}. To develop CRUSO-P, we processed >3 million SO posts and 188200+ GitHub source files. CRUSO-P is designed to work with source code written in the popular programming languages {C, C#, Java, JavaScript, and Python}. CRUSO-P outperforms CRUSO with an improvement of 97.82% in response time and a storage reduction of 99.15%. CRUSO-P achieves the highest mean accuracy score of 99.6% when tested with the C programming language, thus achieving an improvement of 5.6% over the existing method.