SDA: Improving Text Generation with Self Data Augmentation
This work addresses the challenge of data augmentation for text generation, a problem for researchers and practitioners working with deep neural networks for natural language processing, offering an incremental improvement to existing MLE-based training procedures.
This paper proposes Self Data Augmentation (SDA) to enhance text generation by integrating a self-imitation-learning phase into the standard maximum likelihood estimation (MLE) paradigm. The method is general and can be adapted to any MLE-based training, showing significant improvements over baselines on synthetic and real datasets.
Data augmentation has been widely used to improve deep neural networks in many research fields, such as computer vision. However, less work has been done in the context of text, partially due to its discrete nature and the complexity of natural languages. In this paper, we propose to improve the standard maximum likelihood estimation (MLE) paradigm by incorporating a self-imitation-learning phase for automatic data augmentation. Unlike most existing sentence-level augmentation strategies, which are only applied to specific models, our method is more general and could be easily adapted to any MLE-based training procedure. In addition, our framework allows task-specific evaluation metrics to be designed to flexibly control the generated sentences, for example, in terms of controlling vocabulary usage and avoiding nontrivial repetitions. Extensive experimental results demonstrate the superiority of our method on two synthetic and several standard real datasets, significantly improving related baselines.