EDA: Easy Data Augmentation Techniques for Boosting Performance on Text Classification Tasks

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, pages 6382–6388, Hong Kong, China, November 3–7, 2019. c⃝2019 Association for Computational Linguistics 6382 EDA: Easy Data Augmentation Techniques for Boosting Performance on Text Classification Tasks Jason Wei1,2 Kai Zou3 1Protago Labs Research, Tysons Corner, Virginia, USA 2Department of Computer Science, Dartmouth College 3Department of Mathematics and Statistics, Georgetown University jason.20@dartmouth.edu kz56@georgetown.edu Abstract We present EDA: easy data augmentation techniques for boosting performance on text classification tasks. EDA consists of four sim- ple but powerful operations: synonym replace- ment, random insertion, random swap, and random deletion. On five text classification tasks, we show that EDA improves perfor- mance for both convolutional and recurrent neural networks. EDA demonstrates particu- larly strong results for smaller datasets; on av- erage, across five datasets, training with EDA while using only 50% of the available train- ing set achieved the same accuracy as normal training with all available data. We also per- formed extensive ablation studies and suggest parameters for practical use. 1 Introduction Text classification is a fundamental task in natu- ral language processing (NLP). Machine learning and deep learning have achieved high accuracy on tasks ranging from sentiment analysis (Tang et al., 2015) to topic classification (Tong and Koller, 2002), but high performance often depends on the size and quality of training data, which is often te- dious to collect. Automatic data augmentation is commonly used in computer vision (Simard et al., 1998; Szegedy et al., 2014; Krizhevsky et al., 2017) and speech (Cui et al., 2015; Ko et al., 2015) and can help train more robust models, particu- larly when using smaller datasets. However, be- cause it is challenging to come up with generalized rules for language transformation, universal data augmentation techniques in NLP have not been thoroughly explored. Previous work has proposed some techniques for data augmentation in NLP. One popular study generated new data by translating sentences into French and back into English (Yu et al., 2018). Other work has used data noising as smoothing Operation Sentence None A sad, superior human comedy played out on the back roads of life. SR A lamentable, superior human comedy played out on the backward road of life. RI A sad, superior human comedy played out on funniness the back roads of life. RS A sad, superior human comedy played out on roads back the of life. RD A sad, superior human out on the roads of life. Table 1: Sentences generated using EDA. SR: synonym replacement. RI: random insertion. RS: random swap. RD: random deletion. (Xie et al., 2017) and predictive language models for synonym replacement (Kobayashi, 2018). Al- though these techniques are valid, they are not of- ten used in practice because they have a high cost of implementation relative to performance gain. In this paper, we present a simple set of univer- sal data augmentation techniques for NLP called EDA (easy data augmentation). To the best of our …

Content truncated to 500 words