Research map / Label-efficient and robust learning / Semi-supervised learning
A Semi-supervised Learning Approach with Two Teachers to Improve Breakdown Identification in Dialogues
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use the two-teacher bootstrapping framework instead of MixText | DBDC5 English Track · DBDC5 English Track | Accuracy: 0.757 -> 0.779 | Helped |
| Use the two-teacher bootstrapping framework instead of XLMR+CM | DBDC5 Japanese Track · DBDC5 Japanese Track | F1(B): 0.694 -> 0.754 | Helped |
| Use [MASK]-augmented unlabeled data rather than unaugmented unlabeled data (ablation) | Ours · DBDC5 English Track | Accuracy: 0.759 -> 0.779 | Helped |
| Add the Masked Teacher (ablation) | Ours · DBDC5 English Track | Accuracy: 76.9 -> 77.9 | Helped |
| Add the Gold Teacher (ablation) | Ours · DBDC5 English Track | Accuracy: 74.0 -> 77.9 | Helped |
| Add self-training after removing the Masked Teacher (ablation) | Ours · DBDC5 English Track | Accuracy: 75.0 -> 76.9 | Helped |
| Add self-training after removing the Gold Teacher (ablation) | Ours · DBDC5 Japanese Track | Accuracy: 73.1 -> 73.4 | Helped |
| Add both teachers and bootstrapping | Ours · LIF Test-I | Macro F1: 81.4 -> 86.5 | Helped |
| Increase labeled training data from 50% to 100% | Ours · LIF Test-I | Macro F1: 86.4 -> 86.5 | No clear effect |
| Use RoBERTa-large with SSMBA augmentation | DBDC5 English Track · DBDC5 English Track | Accuracy: 0.750 -> 0.752 | No clear effect |
| Use [MASK]-augmented unlabeled data rather than unaugmented unlabeled data (ablation) | Ours · LIF Test-I | Macro F1: 85.6 -> 86.5 | Helped |
Closest papers by meaning
- Opening the Black Box: Unraveling the Classroom Dialogue Analysis (Student Abstract) (AAAI 2024 · similarity 0.84)
- Response Enhanced Semi-supervised Dialogue Query Generation (AAAI 2024 · similarity 0.83)
- An Efficient Self-Learning Framework For Interactive Spoken Dialog Systems (ICML 2024 · similarity 0.83)
- Utterance-level Emotion Recognition in Conversation with Conversation-level Supervision (AAAI 2025 · similarity 0.83)
- SD-Eval: A Benchmark Dataset for Spoken Dialogue Understanding Beyond Words (NeurIPS 2024 · similarity 0.83)
- Friends-MMC: A Dataset for Multi-modal Multi-party Conversation Understanding (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.