Research map / Label-efficient and robust learning / Recognition and long-tail
Topic-Aware Multi-turn Dialogue Modeling
Yi Xu, Hai Zhao, Zhuosheng Zhang
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 |
|---|---|---|---|
| Use topic-aware segmentation and dual-attention matching | BERT response selection · E-commerce | R10@1: 62.7 -> 66.0; 3.3% | Improved |
| Use topic-aware segmentation and dual-attention matching | BERT response selection · Ubuntu | R10@1: 81.9 -> 82.1 | Improved |
| Use topic-aware segmentation and dual-attention matching | BERT response selection · Douban | R10@1: 29.5 -> 28.2 | Hurt |
| Add word-level segment weights (ablation) | TADAM · E-commerce | R10@1: 62.0 -> 66.0 | Improved |
| Add segment-level weights (ablation) | TADAM · E-commerce | R10@1: 63.1 -> 66.0 | Improved |
| Add both word-level and segment-level weights (ablation) | TADAM · E-commerce | R10@1: 62.2 -> 66.0 | Improved |
| Add the last-segment match (ablation) | TADAM · E-commerce | R10@1: 64.0 -> 66.0 | Improved |
| Add multi-turn matching (ablation) | TADAM · E-commerce | R10@1: 63.8 -> 66.0 | Improved |
| Use dual cross-attention instead of segment-query single matching (ablation) | TADAM · E-commerce | R10@1: 62.7 -> 66.0 | Improved |
| Use dual cross-attention instead of response-query single matching (ablation) | TADAM · E-commerce | R10@1: 62.4 -> 66.0 | Improved |
| Encode segments and response together rather than separately (ablation) | TADAM · E-commerce | R10@1: 36.0 -> 66.0 | Improved |
| Use topic segments instead of concatenated utterances (ablation) | TADAM · E-commerce | R10@1: 62.9 -> 66.0 | Improved |
Closest papers by meaning
- Filling the Gap of Utterance-aware and Speaker-aware Representation for Multi-turn Dialogue (AAAI 2021 · similarity 0.87)
- NaturalConv: A Chinese Dialogue Dataset Towards Multi-turn Topic-driven Conversation (AAAI 2021 · similarity 0.86)
- ConvNTM: Conversational Neural Topic Model (AAAI 2023 · similarity 0.86)
- F2RVLM: Boosting Fine-grained Fragment Retrieval for Multi-Modal Long-form Dialogue with Vision Language Model (AAAI 2026 · similarity 0.86)
- Aligning Spoken Dialogue Models from User Interactions (ICML 2025 · similarity 0.84)
- SAMA: Towards Multi-Turn Referential Grounded Video Chat with Large Language Models (NeurIPS 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.
