Automatically Learning Data Augmentation Policies for Dialogue Tasks

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, pages 1317–1323, Hong Kong, China, November 3–7, 2019. c⃝2019 Association for Computational Linguistics 1317 Automatically Learning Data Augmentation Policies for Dialogue Tasks Tong Niu Mohit Bansal UNC Chapel Hill {tongn, mbansal}@cs.unc.edu Abstract Automatic data augmentation (AutoAug- ment) (Cubuk et al., 2019) searches for optimal perturbation policies via a controller trained using performance rewards of a sam- pled policy on the target task, hence reducing data-level model bias. While being a powerful algorithm, their work has focused on computer vision tasks, where it is comparatively easy to apply imperceptible perturbations without changing an image’s semantic meaning. In our work, we adapt AutoAugment to automatically discover effective perturbation policies for natural language processing (NLP) tasks such as dialogue generation. We start with a pool of atomic operations that ap- ply subtle semantic-preserving perturbations to the source inputs of a dialogue task (e.g., different POS-tag types of stopword dropout, grammatical errors, and paraphrasing). Next, we allow the controller to learn more complex augmentation policies by searching over the space of the various combinations of these atomic operations. Moreover, we also explore conditioning the controller on the source in- puts of the target task, since certain strategies may not apply to inputs that do not contain that strategy’s required linguistic features. Empirically, we demonstrate that both our input-agnostic and input-aware controllers discover useful data augmentation policies, and achieve significant improvements over the previous state-of-the-art, including trained on manually-designed policies. 1 Introduction Data augmentation aims at teaching invariances to a model so that it generalizes better outside the training set distribution. Recently, there has been substantial progress in Automatic Data Augmen- tation (AutoAugment) for computer vision (Cubuk et al., 2019). This algorithm searches for optimal perturbation policies via a controller trained with reinforcement learning, where its reward comes from training the target model with data perturbed by the sampled augmentation policy. Each pol- icy consists of 5 sub-policies sampled randomly during training, and each sub-policy consists of 2 operations applied in sequence. These oper- ations are semantic-preserving image processing functions such as translation and rotation. We adapt AutoAugment to NLP tasks, where the operations are subtle, semantic-preserving text perturbations. To collect a pool of such opera- tions, the first challenge we face is that the dis- crete nature of text makes it less straightforward to come up with semantic-preserving perturbations. We thus employ as a starting point Should-Not- Change strategies (equivalent to operations in this paper) proposed by Niu and Bansal (2018) which are shown to improve their dialogue task perfor- mance when trained on data perturbed by these strategies. Importantly, we next divide their op- erations into several smaller ones (such as Gram- mar Errors divided into Singular/Plural Errors and Verb Inflection Errors) and also add a new opera- tion Stammer (word repetition). This modification provides a much larger space of operation com- binations for the model to explore from, so that …

Content truncated to 500 words