SeqMix: Augmenting Active Sequence Labeling via Sequence Mixup

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, pages 8566–8579, November 16–20, 2020. c⃝2020 Association for Computational Linguistics 8566 SeqMix: Augmenting Active Sequence Labeling via Sequence Mixup Rongzhi Zhang Georgia Tech rongzhi.zhang@gatech.edu Yue Yu Georgia Tech yueyu@gatech.edu Chao Zhang Georgia Tech chaozhang@gatech.edu Abstract Active learning is an important technique for low-resource sequence labeling tasks. How- ever, current active sequence labeling methods use the queried samples alone in each itera- tion, which is an inefficient way of leverag- ing human annotations. We propose a simple but effective data augmentation method to im- prove label efficiency of active sequence label- ing. Our method, SeqMix, simply augments the queried samples by generating extra la- beled sequences in each iteration. The key dif- ficulty is to generate plausible sequences along with token-level labels. In SeqMix, we address this challenge by performing mixup for both sequences and token-level labels of the queried samples. Furthermore, we design a discrim- inator during sequence mixup, which judges whether the generated sequences are plausi- ble or not. Our experiments on Named Entity Recognition and Event Detection tasks show that SeqMix can improve the standard active sequence labeling method by 2.27%–3.75% in terms of F1 scores. The code and data for SeqMix can be found at https://github. com/rz-zhang/SeqMix. 1 Introduction Many NLP tasks can be formulated as sequence labeling problems, such as part-of-speech (POS) tagging (Zheng et al., 2013), named entity recogni- tion (NER) (Lample et al., 2016), and event extrac- tion (Yang et al., 2019). Recently, neural sequential models (Lample et al., 2016; Akbik et al., 2018; Vaswani et al., 2017) have shown strong perfor- mance for various sequence labeling task. How- ever, these deep neural models are label hungry— they require large amounts of annotated sequences to achieve strong performance. Obtaining large amounts of annotated data can be too expensive for practical sequence labeling tasks, due to token- level annotation efforts. Active learning is an important technique for se- quence labeling in low-resource settings. Active sequence labeling is an iterative process. In each it- eration, a fixed number of unlabeled sequences are selected by a query policy for annotation and then model updating, in hope of maximally improving model performance. For example, Tomanek et al. (2007); Shen et al. (2017) select query samples based on data uncertainties; Hazra et al. (2019) compute model-aware similarity to eliminate redun- dant examples and improve the diversity of query samples; and Fang et al. (2017); Liu et al. (2018) use reinforcement learning to learn query policies. However, existing methods for active sequence la- beling all use the queried samples alone in each iteration. We argue that the queried samples pro- vide limited data diversity, and using them alone for model updating is inefficient in terms of leveraging human annotation efforts. We study the problem of enhancing active se- quence labeling via data augmentation. We aim to generate augmented labeled sequences for the queried samples in each iteration, thereby introduc- ing more data diversity and improve model gener- alization. …

Content truncated to 500 words