Data Augmentation for Text Generation Without Any Augmented Data

被引:0
作者
Bi, Wei [1 ]
Li, Huayang [1 ]
Huang, Jiacheng [1 ]
机构
[1] Tencent AI Lab, Shenzhen, Peoples R China
来源
59TH ANNUAL MEETING OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS AND THE 11TH INTERNATIONAL JOINT CONFERENCE ON NATURAL LANGUAGE PROCESSING, VOL 1 (ACL-IJCNLP 2021) | 2021年
关键词
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Data augmentation is an effective way to improve the performance of many neural text generation models. However, current data augmentation methods need to define or choose proper data mapping functions that map the original samples into the augmented samples. In this work, we derive an objective to formulate the problem of data augmentation on text generation tasks without any use of augmented data constructed by specific mapping functions. Our proposed objective can be efficiently optimized and applied to popular loss functions on text generation tasks with a convergence rate guarantee. Experiments on five datasets of two text generation tasks show that our approach can approximate or even surpass popular data augmentation methods.
引用
收藏
页码:2223 / 2237
页数:15
相关论文
共 38 条
  • [1] [Anonymous], 2015, Character-level convolutional networks for text classification
  • [2] Bahdanau D, 2016, Arxiv, DOI arXiv:1409.0473
  • [3] Bengio Y., 2009, P 26 ANN INT C MACHI, P41, DOI DOI 10.1145/1553374.1553380
  • [4] Cai H., 2020, P 58 ANN M ASS COMP, P6334, DOI DOI 10.18653/V1/2020.ACL-MAIN.564
  • [5] Chen Liqun, P INT C LEARN REPR I
  • [6] Chen X., ARXIV PREPRINT ARXIV
  • [7] Cheng Y, 2019, 57TH ANNUAL MEETING OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS (ACL 2019), P4324
  • [8] Devlin J., 2019, PROC NAACL, V1, P4171
  • [9] A decision-theoretic generalization of on-line learning and an application to boosting
    Freund, Y
    Schapire, RE
    [J]. JOURNAL OF COMPUTER AND SYSTEM SCIENCES, 1997, 55 (01) : 119 - 139
  • [10] Guo Hongyu, 2019, ARXIV190508941