A Neural Architecture for Generating Natural Language Descriptions from Source Code Changes

TL;DR

This paper introduces a neural model that automatically generates natural language descriptions of code changes from commit data, effectively capturing semantics across multiple programming languages and project contexts.

Contribution

The paper presents a novel encoder-decoder architecture trained on commit data to generate accurate descriptions of code modifications, including cross-project generalization.

Findings

Effective in generating feasible descriptions

Works across multiple programming languages

Performs well in cross-project scenarios

Abstract

We propose a model to automatically describe changes introduced in the source code of a program using natural language. Our method receives as input a set of code commits, which contains both the modifications and message introduced by an user. These two modalities are used to train an encoder-decoder architecture. We evaluated our approach on twelve real world open source projects from four different programming languages. Quantitative and qualitative results showed that the proposed approach can generate feasible and semantically sound descriptions not only in standard in-project settings, but also in a cross-project setting.