Extending belief update models for the robust dialog state trackers

Abstract

Dialog trackers are one of the essential components of dialog systems which are used to infer the true user goal from the speech processing results. We engineer a previous statistical dialog tracker by applying various algorithms to models of the tracker: the observation model and the belief refinement model. And we propose new models which overcome limitations of the general belief update model of the tracker: the belief transformation model and the parameterized belief update model. We describe our experience to build robust dialog state trackers for the first Dialog State Tracking Challenge (DSTC). We also report experimental results on a number of approaches to the models, and compare the overall performance of our tracker to other submitted trackers.