Research map / Architectures and efficiency / RNNs and normalization
Building End-To-End Dialogue Systems Using Generative Hierarchical Neural Network Models
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 |
|---|---|---|---|
| Initialize HRED word embeddings with pretrained Word2Vec (ablation) | HRED · MovieTriples test | perplexity: 36.59 ± 0.19 -> 33.95 ± 0.16 | Helped |
| Pretrain HRED on the SubTle Q-A corpus (ablation) | HRED · MovieTriples test | perplexity: 36.59 ± 0.19 -> 27.14 ± 0.12 | Helped |
| Use a bidirectional utterance encoder (ablation) | HRED + SubTle · MovieTriples test | perplexity: 27.14 ± 0.12 -> 26.81 ± 0.11 | Helped |
| Use the hierarchical HRED architecture instead of a concatenated-utterance RNN | RNN · MovieTriples test | perplexity: 35.63 ± 0.16 -> 36.59 ± 0.19 | Hurt |
| Pretrain on SubTle rather than use no pretraining (ablation) | RNN · MovieTriples test | perplexity: 35.63 ± 0.16 -> 27.09 ± 0.13 | Helped |
| Pretrain HRED on SubTle rather than use no pretraining (ablation) | HRED · MovieTriples test U3 | perplexity: 36.26 ± 0.29 -> 26.60 ± 0.19 | Helped |
| Use a bidirectional encoder instead of the standard HRED encoder (ablation) | HRED + SubTle · MovieTriples test U3 | word error-rate: 64.03% ± 0.10 -> 63.91% ± 0.09 | Helped |
| Replace Word2Vec initialization with learned embeddings through SubTle pretraining (ablation) | HRED · MovieTriples test | perplexity: 33.95 ± 0.16 -> 27.14 ± 0.12 | Helped |
Closest papers by meaning
- A Hierarchical Latent Variable Encoder-Decoder Model for Generating Dialogues (AAAI 2017 · similarity 0.93)
- Mechanism-Aware Neural Machine for Dialogue Response Generation (AAAI 2017 · similarity 0.90)
- DialogBERT: Discourse-Aware Response Generation via Learning to Recover and Rank Utterances (AAAI 2021 · similarity 0.89)
- From Retrieval to Generation: A Simple and Unified Generative Model for End-to-End Task-Oriented Dialogue (AAAI 2024 · similarity 0.89)
- Best of Both Worlds: Transferring Knowledge from Discriminative Learning to a Generative Visual Dialog Model (NeurIPS 2017 · similarity 0.89)
- Multiresolution Recurrent Neural Networks: An Application to Dialogue Response Generation (AAAI 2017 · similarity 0.89)
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.