Exploring Data Augmentation for Code Generation Tasks

TL;DR

This paper investigates data augmentation techniques to improve code generation tasks, demonstrating consistent performance gains in code translation and summarization, and analyzing their effects on output quality and data imperfections.

Contribution

It introduces and adapts data augmentation methods specifically for downstream code generation tasks, achieving notable improvements.

Findings

Up to 6.9% improvement in code translation

Up to 7.5% improvement in code summarization

Enhancement in output code style and numeric consistency

Abstract

Advances in natural language processing, such as transfer learning from pre-trained language models, have impacted how models are trained for programming language tasks too. Previous research primarily explored code pre-training and expanded it through multi-modality and multi-tasking, yet the data for downstream tasks remain modest in size. Focusing on data utilization for downstream tasks, we propose and adapt augmentation methods that yield consistent improvements in code translation and summarization by up to 6.9% and 7.5% respectively. Further analysis suggests that our methods work orthogonally and show benefits in output code style and numeric consistency. We also discuss test data imperfections.