Mixup-Transformer: Dynamic Data Augmentation for NLP Tasks

TL;DR

This paper introduces Mixup-Transformer, a novel data augmentation method for NLP that applies mixup to transformer models like BERT, improving performance across various tasks and data scenarios.

Contribution

It adapts mixup for NLP by integrating it into transformer architectures, demonstrating its effectiveness in improving model performance on the GLUE benchmark.

Findings

Mixup-Transformer significantly boosts NLP model accuracy.

Effective in low-resource training scenarios.

Proven domain-independent data augmentation technique.

Abstract

Mixup is the latest data augmentation technique that linearly interpolates input examples and the corresponding labels. It has shown strong effectiveness in image classification by interpolating images at the pixel level. Inspired by this line of research, in this paper, we explore i) how to apply mixup to natural language processing tasks since text data can hardly be mixed in the raw format; ii) if mixup is still effective in transformer-based learning models, e.g., BERT. To achieve the goal, we incorporate mixup to transformer-based pre-trained architecture, named "mixup-transformer", for a wide range of NLP tasks while keeping the whole end-to-end training system. We evaluate the proposed framework by running extensive experiments on the GLUE benchmark. Furthermore, we also examine the performance of mixup-transformer in low-resource scenarios by reducing the training data with a certain ratio. Our studies show that mixup is a domain-independent data augmentation technique to pre-trained language models, resulting in significant performance improvement for transformer-based models.