Research map / Large language models / Language, safety and interpretability
DialogBERT: Discourse-Aware Response Generation via Learning to Recover and Rank Utterances
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 hierarchical DialogBERT architecture and NUG+MUR+DUOR objectives instead of the strongest reported baseline | DialogBERT · Weibo | PPL: 34.15 -> 27.30 | Helped |
| add MUR to the NUG-only model (ablation) | DialogBERT · Weibo | PPL: 31.73 -> 28.08 | Helped |
| add DUOR to the NUG-only model (ablation) | DialogBERT · Weibo | PPL: 31.73 -> 27.29 | Helped |
| combine MUR and DUOR with NUG (ablation) | DialogBERT · Weibo | BLEU: 8.25 -> 8.54 | Helped |
| add MUR to the NUG-only model (ablation) | DialogBERT · DailyDialog | BLEU: 14.41 -> 13.81 | Hurt |
| add DUOR to the NUG-only model (ablation) | DialogBERT · DailyDialog | BLEU: 14.41 -> 14.42 | No clear effect |
| combine MUR and DUOR with NUG (ablation) | DialogBERT · DailyDialog | PPL: 22.16 -> 20.49 | Helped |
| combine MUR and DUOR with NUG (ablation) | DialogBERT · MultiWOZ | NIST: 49.83 -> 55.03 | Helped |
| use context encoder depth L=3 instead of L=6 (ablation) | DialogBERT · MultiWOZ | PPL: 4.92 -> 4.94 | Hurt |
| use context encoder depth L=3 instead of L=6 (ablation) | DialogBERT · DailyDialog | PPL: 20.49 -> 20.85 | Hurt |
| use masked utterance regression rather than masked utterance retrieval (ablation) | DialogBERT · MultiWOZ | NIST: 51.25 -> 52.74 | Helped |
| use distributed utterance order ranking rather than utterance order classification (ablation) | DialogBERT · MultiWOZ | NIST: 48.07 -> 53.29 | Helped |
Closest papers by meaning
- Mechanism-Aware Neural Machine for Dialogue Response Generation (AAAI 2017 · similarity 0.89)
- Building End-To-End Dialogue Systems Using Generative Hierarchical Neural Network Models (AAAI 2016 · similarity 0.89)
- DialectLLM: A Dialect-Aware Dialog[ue] Generation Framework Beyond Standard American English (NeurIPS 2026 · similarity 0.88)
- Improving Open-Domain Dialogue Response Generation with Multi-Source Multilingual Commonsense Knowledge (AAAI 2024 · similarity 0.88)
- Paralinguistics-Aware Speech-Empowered Large Language Models for Natural Conversation (NeurIPS 2024 · similarity 0.88)
- From Retrieval to Generation: A Simple and Unified Generative Model for End-to-End Task-Oriented Dialogue (AAAI 2024 · similarity 0.87)
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.