Few-shot learning through contextual data augmentation

被引:0
|
作者
Arthaud, Farid [1 ,3 ]
Bawden, Rachel [2 ]
Birch, Alexandra [3 ]
机构
[1] Ecole Normale Super, Paris, France
[2] INRIA, Le Chesnay Rocquencourt, France
[3] Univ Edinburgh, Edinburgh, Midlothian, Scotland
基金
英国工程与自然科学研究理事会;
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
Machine translation (MT) models used in industries with constantly changing topics, such as translation or news agencies, need to adapt to new data to maintain their performance over time. Our aim is to teach a pre-trained MT model to translate previously unseen words accurately, based on very few examples. We propose (i) an experimental setup allowing us to simulate novel vocabulary appearing in human-submitted translations, and (ii) corresponding evaluation metrics to compare our approaches. We extend a data augmentation approach using a pre-trained language model to create training examples with similar contexts for novel words. We compare different fine-tuning and data augmentation approaches and show that adaptation on the scale of one to five examples is possible. Combining data augmentation with randomly selected training sentences leads to the highest BLEU score and accuracy improvements. Impressively, with only 1 to 5 examples, our model reports better accuracy scores than a reference system trained with on average 313 parallel examples.
引用
收藏
页码:1049 / 1062
页数:14
相关论文
共 50 条
  • [1] Few-Shot Learning With Enhancements to Data Augmentation and Feature Extraction
    Zhang, Yourun
    Gong, Maoguo
    Li, Jianzhao
    Feng, Kaiyuan
    Zhang, Mingyang
    [J]. IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, 2024,
  • [2] FlipDA: Effective and Robust Data Augmentation for Few-Shot Learning
    Zhou, Jing
    Zheng, Yanan
    Tang, Jie
    Li, Jian
    Yang, Zhilin
    [J]. PROCEEDINGS OF THE 60TH ANNUAL MEETING OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS (ACL 2022), VOL 1: (LONG PAPERS), 2022, : 8646 - 8665
  • [3] Contextual Gradient Scaling for Few-Shot Learning
    Lee, Sanghyuk
    Lee, Seunghyun
    Song, Byung Cheol
    [J]. 2022 IEEE WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION (WACV 2022), 2022, : 3503 - 3512
  • [4] Improving Augmentation Efficiency for Few-Shot Learning
    Cho, Wonhee
    Kim, Eunwoo
    [J]. IEEE ACCESS, 2022, 10 : 17697 - 17706
  • [5] Few-shot Partial Multi-label Learning with Data Augmentation
    Sun, Yifan
    Zhao, Yunfeng
    Yu, Guoxian
    Yan, Zhongmin
    Domeniconi, Carlotta
    [J]. 2022 IEEE INTERNATIONAL CONFERENCE ON DATA MINING (ICDM), 2022, : 478 - 487
  • [6] Data Augmentation Aided Few-Shot Learning for Specific Emitter Identification
    Zhang, Xixi
    Wang, Yu
    Zhang, Yibin
    Lin, Yun
    Gui, Guan
    Tomoaki, Ohtsuki
    Sari, Hikmet
    [J]. 2022 IEEE 96TH VEHICULAR TECHNOLOGY CONFERENCE (VTC2022-FALL), 2022,
  • [7] Few-Shot Charge Prediction with Data Augmentation and Feature Augmentation
    Wang, Peipeng
    Zhang, Xiuguo
    Cao, Zhiying
    [J]. APPLIED SCIENCES-BASEL, 2021, 11 (22):
  • [8] Towards Contextual Learning in Few-shot Object Classification
    Fortin, Mathieu Page
    Chaib-draa, Brahim
    [J]. 2021 IEEE WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION WACV 2021, 2021, : 3278 - 3287
  • [9] Ortho-Shot: Low Displacement Rank Regularization with Data Augmentation for Few-Shot Learning
    Osahor, Uche
    Nasrabadi, Nasser M.
    [J]. 2022 IEEE WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION (WACV 2022), 2022, : 2040 - 2049
  • [10] Self-Supervison with data-augmentation improves few-shot learning
    Kumar, Prashant
    Toshniwal, Durga
    [J]. APPLIED INTELLIGENCE, 2024, 54 (04) : 2976 - 2997