Dialog state tracking is a key component of many modern dialog systems, most of which are designed with a single, well-defined domain in mind. This paper shows that dialog data drawn from different dialog domains can be used to train a general belief tracking model which can operate across all of these domains, exhibiting superior performance to each of the domain-specific models. We propose a training procedure which uses out-of-domain data to initialise belief tracking models for entirely new domains. This procedure leads to improvements in belief tracking performance regardless of the amount of in-domain data available for training the model.
@article{arxiv.1506.07190,
title = {Multi-domain Dialog State Tracking using Recurrent Neural Networks},
author = {Nikola Mrkšić and Diarmuid Ó Séaghdha and Blaise Thomson and Milica Gašić and Pei-Hao Su and David Vandyke and Tsung-Hsien Wen and Steve Young},
journal= {arXiv preprint arXiv:1506.07190},
year = {2015}
}
Comments
Accepted as a short paper in the 53rd Annual Meeting of the Association for Computational Linguistics (ACL 2015)