Taming Continuous Posteriors for Latent Variational Dialogue Policies
Not on the map. Why?
Abstract (arXiv)
Utilizing amortized variational inference for latent-action reinforcement learning (RL) has been shown to be an effective approach in Task-oriented Dialogue (ToD) systems for optimizing dialogue success. Until now, categorical posteriors have been argued to be one of the main drivers of performance. In this work we revisit Gaussian variational posteriors for latent-action RL and show that they can yield even better performance than categoricals. We achieve this by simplifying the training procedure and propose ways to regularize the latent dialogue policy to retain good response coherence. Using continuous latent representations our model achieves state of the art dialogue success rate on the MultiWOZ benchmark, and also compares well to categorical latent methods in response coherence.
Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 30 of its 42 references that are in the map. First posted on arXiv 16 May 2022.
Builds on
30 of 42 references.
References · 30
- 2014Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translationcited by 4,057
- 2013Auto-Encoding Variational Bayescited by 4,053
- 2013Playing Atari with Deep Reinforcement Learningcited by 3,274
- 2018UMAP: Uniform Manifold Approximation and Projection for Dimension Reductioncited by 2,898
- 2018MultiWOZ -- A Large-Scale Multi-Domain Wizard-of-Oz Dataset for Task-Oriented Dialogue Modellingcited by 766
- 2017Learning Discourse-level Diversity for Neural Dialog Models using Conditional Variational Autoencoderscited by 222
- 2017Deal or No Deal? End-to-End Learning for Negotiation Dialoguescited by 209
- 2016A Hierarchical Latent Variable Encoder-Decoder Model for Generating Dialoguescited by 174
- 2018Neural Approaches to Conversational AIcited by 165
- 2020Unsupervised Evaluation of Interactive Dialog with DialoGPTcited by 139
- 2019Task-Oriented Dialog Systems that Consider Multiple Appropriate Responses under the Same Contextcited by 99
- 2017Natural Language Does Not Emerge 'Naturally' in Multi-Agent Dialogcited by 86
18 more
- 2019Rethinking Action Spaces for Reinforcement Learning in End-to-end Dialog Agents with Latent Variable Modelscited by 53
- 2019Training Neural Response Selection for Task-Oriented Dialogue Systemscited by 49
- 2018DialogWAE: Multimodal Response Generation with Conditional Wasserstein Auto-Encodercited by 44
- 2019Semantically Conditioned Dialog Response Generation via Hierarchical Disentangled Self-Attentioncited by 44
- 2020CoCo: Controllable Counterfactuals for Evaluating Dialogue State Trackerscited by 43
- 2019Structured Fusion Networks for Dialogcited by 40
- 2017Hybrid Code Networks: practical and efficient end-to-end dialog control with supervised and reinforcement learningcited by 32
- 2020LAVA: Latent Action Spaces via Variational Auto-encoding for Dialogue Policy Optimizationcited by 30
- 2019Hierarchical Reinforcement Learning for Open-Domain Dialogcited by 27
- 2020Multi-Domain Dialogue Acts and Response Co-Generationcited by 26
- 2020UniConv: A Unified Conversational Neural Architecture for Multi-domain Task-oriented Dialoguescited by 19
- 2017Sample-efficient Actor-Critic Reinforcement Learning with Supervised Data for Dialogue Managementcited by 18
- 2018A Hierarchical Latent Structure for Variational Conversation Modelingcited by 17
- 2021Towards Automatic Evaluation of Dialog Systems: A Model-Free Off-Policy Evaluation Approachcited by 16
- 2016Policy Networks with Two-Stage Training for Dialogue Systemscited by 14
- 2018Improving Variational Encoder-Decoders in Dialogue Generationcited by 13
- 2018Towards Explainable and Controllable Open Domain Dialogue Generation with Dialogue Actscited by 8
- 2020SUMBT+LaRL: Effective Multi-domain End-to-end Neural Task-oriented Dialog Systemcited by 3
Built on by
Citing papers per year: 2024: 1.
1 paper.