Sequential targeting: A continual learning approach for data imbalance in text classification
作者:
Highlights:
• Handling data imbalance in text data with continual learning.
• Detecting sexual harassment and sentiment in comments of social media platforms.
• Effective trade-off between precision and recall for an overall increase in F1-score.
• Results show 56.38 %p increase on the IMDB dataset.
• Results show 16.89 %p and 34.76 %p increase on NAVER Dataset.
摘要
•Handling data imbalance in text data with continual learning.•Detecting sexual harassment and sentiment in comments of social media platforms.•Effective trade-off between precision and recall for an overall increase in F1-score.•Results show 56.38 %p increase on the IMDB dataset.•Results show 16.89 %p and 34.76 %p increase on NAVER Dataset.
论文关键词:Continual learning,Data imbalance,Deep learning,Sentiment analysis,Text classification
论文评审过程:Received 22 November 2020, Revised 6 April 2021, Accepted 16 April 2021, Available online 21 April 2021, Version of Record 6 May 2021.
论文官网地址:https://doi.org/10.1016/j.eswa.2021.115067