Text data augmentations: Permutation, antonyms and negation

作者：

Highlights：

• Three novel text data augmentation techniques.

• Accuracy improvement across eight datasets, 4.1%.

• TF-IDF filtering

• Max accuracy reduction 2.08%.

• Minimum training time reduction 212%.

摘要

•Three novel text data augmentation techniques.•Accuracy improvement across eight datasets, 4.1%.•TF-IDF filtering•Max accuracy reduction 2.08%.•Minimum training time reduction 212%.

论文关键词：Text,Augmentation,Multilabel,Multiclass,LSTM

论文评审过程：Received 8 July 2020, Revised 21 December 2020, Accepted 19 February 2021, Available online 11 March 2021, Version of Record 9 April 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2021.114769