Automatic generation of information state update dialogue systems that dynamically create voice XML, as demonstrated on the iPhone

Abstract

We demonstrate DUDE1 (Dialogue and Understanding Development Environment), a prototype development environment that automatically generates dialogue systems from business-user resources and databases. These generated spoken dialogue systems (SDS) are then deployed on an industry standard Voice XML platform. Specifically, the deployed system works by dynamically generating context-sensitive Voice XML pages. The dialogue move of each page is determined in real time by the dialogue manager, which is an Information State Update engine. Firstly, we will demonstrate the development environment which includes automatic generation of speech recognition grammars for robust interpretation of spontaneous speech, and uses the application database to generate lexical entries and grammar rules. A simple graphical interface allows users (i.e. developers) to easily and quickly create and the modify SDS without the need for expensive service providers. Secondly, we will demonstrate the deployed system which enables participants to call up and speak to the SDS recently created. We will also show a pre-built application running on the iPhone and Google Android phone for searching for places such as restaurants, hotels and museums. © 2009 Association for Computational Linguistics.