Using back-and-forth translation to create artificial augmented textual data for sentiment analysis models

作者：

Highlights：

• A novel data augmentation method, namely back-and-forth translation, is proposed.

• The proposed method decreases a sentiment classification model’s total error rate.

• The proposed total error decrease scales non-linearly with sample size.

• The method is more effective at smaller sample sizes compared to larger sample sizes.

摘要

•A novel data augmentation method, namely back-and-forth translation, is proposed.•The proposed method decreases a sentiment classification model’s total error rate.•The proposed total error decrease scales non-linearly with sample size.•The method is more effective at smaller sample sizes compared to larger sample sizes.

论文关键词：Natural language processing,Translation,Sentiment analysis,Data augmentation

论文评审过程：Received 19 May 2020, Revised 26 November 2020, Accepted 9 April 2021, Available online 16 April 2021, Version of Record 29 April 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2021.115033