基于多任务学习的正逆向情绪分值回归方法

Translated title of the contribution: Emotion Regression Approach with Both Forward and Reverse Values Based on Multi-task Learning

Xiaoya Gao, Yat Mei Lee, Lu Zhang, Shoushan Li

Research output: Journal article publicationJournal articleAcademic researchpeer-review

2 Citations (Scopus)

Abstract

提出一种基于多任务学习的情绪分值回归方法。首先,针对每一种情绪分值设计了正向打分和逆向打分;其次,将每一种分值的回归任务分为正向打分回归子任务和逆向打分回归子任务;最后,提出一种多任务学习方法用于主任务(正向打分回归子任务)和辅助任务(逆向打分回归子任务)的共同学习。该方法通过3种不同的共享机制实现中间特征信息共享,从而提升主任务的性能。结果表明,所提出的多任务学习方法能比基准方法获得更好的回归性能。
Translated title of the contributionEmotion Regression Approach with Both Forward and Reverse Values Based on Multi-task Learning
Original languageChinese (Simplified)
Pages (from-to)60-65
JournalZhengzhou Daxue Xuebao/Journal of Zhengzhou University
Volume52
Issue number1
DOIs
Publication statusPublished - Mar 2020

Keywords

  • emotion regression
  • multi-task learning
  • forward and reverse value
  • LSTM

Fingerprint

Dive into the research topics of 'Emotion Regression Approach with Both Forward and Reverse Values Based on Multi-task Learning'. Together they form a unique fingerprint.

Cite this