Research map / Architectures and efficiency / Attention and transformers
Hierarchical Recurrent Attention Network for Response Generation
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 hierarchical recurrent attention over words and utterances | Multi-turn response generation on Douban Group · Douban Group test | perplexity: 41.138 vs 44.508 (S2SA), 47.467 (HRED), 45.484 (VHRED) | Helped |
| Use hierarchical recurrent attention over words and utterances | Multi-turn response generation on Douban Group · Douban Group test, HRAN vs S2SA | human preference win rate: 27.3 win, 20.6 loss | Helped |
| Use hierarchical recurrent attention over words and utterances | Multi-turn response generation on Douban Group · Douban Group test, HRAN vs HRED | human preference win rate: 27.2 win, 21.2 loss | Helped |
| Use hierarchical recurrent attention over words and utterances | Multi-turn response generation on Douban Group · Douban Group test, HRAN vs VHRED | human preference win rate: 25.2 win, 20.4 loss | Helped |
| Add utterance dependency to word-level attention (ablation) | HRAN · Douban Group test | perplexity: 41.54 -> 41.138 | Helped |
| Add word-level attention (ablation) | HRAN · Douban Group test | perplexity: 43.24 -> 41.138 | Helped |
| Add utterance-level attention (ablation) | HRAN · Douban Group test | perplexity: 47.35 -> 41.138 | Helped |
| Add utterance dependency to word-level attention (ablation) | HRAN · Douban Group test, HRAN comparison | human preference win rate: 22.3% win, 24.8% loss | Helped |
| Add word-level attention (ablation) | HRAN · Douban Group test, HRAN comparison | human preference win rate: 20.4% win, 25.0% loss | Helped |
| Add utterance-level attention (ablation) | HRAN · Douban Group test, HRAN comparison | human preference win rate: 21.1% win, 23.7% loss | Helped |
Closest papers by meaning
- Topic Aware Neural Response Generation (AAAI 2017 · similarity 0.89)
- Multiresolution Recurrent Neural Networks: An Application to Dialogue Response Generation (AAAI 2017 · similarity 0.87)
- Neural Response Generation With Dynamic Vocabularies (AAAI 2018 · similarity 0.87)
- Building End-To-End Dialogue Systems Using Generative Hierarchical Neural Network Models (AAAI 2016 · similarity 0.87)
- A Hierarchical Latent Variable Encoder-Decoder Model for Generating Dialogues (AAAI 2017 · similarity 0.86)
- From Self-Attention to Markov Models: Unveiling the Dynamics of Generative Transformers (ICML 2024 · similarity 0.84)
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.