Authors: Zhonglong Chen, Changwei Song, Yining Chen, Jianqiang Li, Guanghui Fu, Yongsheng Tong, Qing Zhao

Published on: May 07, 2024

Impact Score: 7.8

Arxiv code: Arxiv:2405.04128

Summary

• What is new: A new model for identifying negative emotions in callers to a suicide hotline, using a large-scale pre-trained model for analysis.
• Why this is important: The challenge of accurately identifying callers’ emotional states at psychological support hotlines to better assess suicide risk.
• What the research proposes: Developing an effective speech emotion recognition model that automatically detects and analyzes the emotions of callers.
• Results: The model achieved a high F1-score in recognizing negative emotions (76.96%) but demonstrated limited efficacy in fine-grained multi-label classification of emotions.

Technical Details

Technological frameworks used: nan

Models used: Large-scale pre-trained model, speech emotion recognition model, fine-grained multi-label classification model

Data used: 20,630 segments of speech data from 105 callers to the Beijing psychological support hotline

Potential Impact

Psychological support services, suicide prevention organizations, and companies developing or implementing AI for mental health services

Want to implement this idea in a business?

We have generated a startup concept here: EmoTect.