Skip to content
conftrace_
2017 EACL EACL 2017

Latent Variable Dialogue Models and their Diversity

Abstract

AbstractWe present a dialogue generation model that directly captures the variability in possible responses to a given input, which reduces the ‘boring output’ issue of deterministic dialogue models. Experiments show that our model generates more diverse outputs than baseline models, and also generates more consistently acceptable output than sampling from a deterministic encoder-decoder model.

🌉 Interdisciplinary Bridge - Artificial Intelligence and Natural Language Processing
🧭 Keyword Pioneer - response diversity
🐣 Hot Topic Early Bird - latent variable model

Authors

Kris Cao, Stephen Clark

Topics

Deep Learning > Models > Generative Models Natural Language Processing > Generation > Dialogue Systems Machine Learning > Bayesian & Probabilistic > Probabilistic Modeling Machine Learning > Bayesian & Probabilistic > Variational Inference

Keywords

variational inference dialogue generation latent variable model latent variable encoder-decoder model response diversity
Download PDF

More from EACL 2017

All EACL 2017 papers →
Identifying beneficial task relations for multi-task learning in deep neural networks
Effective search space reduction for spell correction using character neural embeddings
Explaining and Generalizing Skip-Gram through Exponential Family Principal Component Analysis
Age Group Classification with Speech and Metadata Multimodality Fusion
Automatically augmenting an emotion dataset improves classification using audio
On-line Dialogue Policy Learning with Companion Teaching