Semi-supervised emotion recognition in textual conversation via a context-augmented auxiliary training task

作者：

Highlights：

• We propose a semi-supervised ERC framework to help ERC models leveraging unlabeled data to improve performance.

• We design a Context-augmented auxiliary training task (CAUXIT) to enhance the model’s ability of learning from context.

• Our approach is compatible with different ERC models.

• The performance gain of applying CAUXIT on different ERC models verifies that our proposed semi-supervised framework is effective.

摘要

•We propose a semi-supervised ERC framework to help ERC models leveraging unlabeled data to improve performance.•We design a Context-augmented auxiliary training task (CAUXIT) to enhance the model’s ability of learning from context.•Our approach is compatible with different ERC models.•The performance gain of applying CAUXIT on different ERC models verifies that our proposed semi-supervised framework is effective.

论文关键词：Emotion recognition in textual conversation,Semi-supervised learning algorithm,Auxiliary training task,Context augmented

论文评审过程：Received 29 April 2021, Revised 4 August 2021, Accepted 5 August 2021, Available online 9 September 2021, Version of Record 9 September 2021.

论文官网地址：https://doi.org/10.1016/j.ipm.2021.102717