Research map / Label-efficient and robust learning / Semi-supervised learning
Inferring Emotion from Conversational Voice Data: A Semi-Supervised Multi-Path Generative Neural Network Approach
Suping Zhou, Jia Jia, Qi Wang, Yufei Dong, Yufeng Yin, Kehua Lei
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| train local and global classifiers jointly in one objective (multi-path) instead of global only (ablation) | MDNN · IEMOCAP acoustic+text | UA (%): 70.0 -> 76.7 | Improved |
| add joint training of global classifier on top of local classifiers (ablation) | MDNN · SVAD13 acoustic | F1 (%): 52.3 -> 52.8 | Improved |
| group features into local classifiers instead of one DNN (ablation) | DNN · SVAD13 acoustic | F1 (%): 48.3 -> 52.3 | Improved |
| global classifier on grouped features instead of one DNN (ablation) | DNN · IEMOCAP acoustic+text | UA (%): 72.8 -> 70.0 | Hurt |
| global classifier on grouped features instead of one DNN (ablation) | DNN · SVAD13 acoustic+text | UA (%): 57.5 -> 59.0 | Improved |
| add 24k unlabeled data via semi-VAE multi-path training (ablation) | MGNN · SVAD13 acoustic | accuracy/F1 (%): 52.8 -> 54.3 | Improved |
| VAE instead of stacked autoencoder pretraining | single-path classifier · SVAD13 acoustic, 24k unlabeled | accuracy/F1 (%): 50.0 -> 50.4 | Improved |
| multi-path with semi-VAE blocks instead of single semi-VAE | semi-VAE · SVAD13 acoustic, 24k unlabeled | accuracy/F1 (%): 50.4 -> 54.3 | Improved |
| add text features to acoustic features (ablation) | MDNN · IEMOCAP | UA (%): 62.7 -> 76.7 | Improved |
| MDNN vs LSTM on text-only features | MDNN · IEMOCAP text only | UA (%): 73.1 -> 66.9 | Hurt |
Closest papers by meaning
- Emotional Face-to-Speech (ICML 2025 · similarity 0.84)
- EmoReg: Directional Latent Vector Modeling for Emotional Intensity Regularization in Diffusion-based Voice Conversion (AAAI 2025 · similarity 0.83)
- Adaptive Graph Learning for Multimodal Conversational Emotion Detection (AAAI 2024 · similarity 0.83)
- Dynamic Interactive Bimodal Hypergraph Networks for Emotion Recognition in Conversations (AAAI 2025 · similarity 0.83)
- Semi-IIN: Semi-Supervised Intra-Inter Modal Interaction Learning Network for Multimodal Sentiment Analysis (AAAI 2025 · similarity 0.83)
- Dual-View Learning for Conversational Emotion Recognition Through Context and Emotion-Shift Modeling (AAAI 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
