Abstract

This is a demonstration of the AT&T “Let’sGo” bus timetable spoken dialog system. This system was entered in the 2010 Spoken Dialog Challenge [1], where the task is to provide bus timetable information for Pittsburgh, Pennsylvania. Our primary aim in the challenge was to build a statistical spoken dialog system to commercial production standards, both in terms of user interface, and also in terms of compatibility with commercial development practices. AT&T Let’s Go illustrates how this can be done, incorporating two statistical techniques: the AT&T Statistical Dialog Toolkit (ASDT) [2], which tracks a distribution over many dialog states in real time; and regression-based confidence scores, which are trained on a corpus of in-domain recognitions [3].