DAGA: Data Augmentation with a Generation Approach for Low-resource Tagging Tasks

TL;DR

This paper introduces DAGA, a novel data augmentation method using language models trained on linearized labeled sentences, improving low-resource tagging tasks in both supervised and semi-supervised settings.

Contribution

The paper proposes a new augmentation approach with language models for high-quality synthetic data generation in low-resource tagging tasks, applicable to various settings.

Findings

Consistently outperforms baseline methods in low-resource scenarios.

Effective in supervised and semi-supervised settings for NER, POS, and sentiment analysis.

Improves model performance especially with limited gold training data.

Abstract

Data augmentation techniques have been widely used to improve machine learning performance as they enhance the generalization capability of models. In this work, to generate high quality synthetic data for low-resource tagging tasks, we propose a novel augmentation method with language models trained on the linearized labeled sentences. Our method is applicable to both supervised and semi-supervised settings. For the supervised settings, we conduct extensive experiments on named entity recognition (NER), part of speech (POS) tagging and end-to-end target based sentiment analysis (E2E-TBSA) tasks. For the semi-supervised settings, we evaluate our method on the NER task under the conditions of given unlabeled data only and unlabeled data plus a knowledge base. The results show that our method can consistently outperform the baselines, particularly when the given gold training data are less.