A Rapid Development Framework for Multilingual Spoken Dialogue Systems

We propose a new framework for rapid development of multilingual spoken dialogue systems based on a data modeling driven method. The development process begins with the data model definition in the Groovy programming language, with an extension for implementing spoken dialogue systems, such as the annotation of the target task type and the initiative type of the realized dialogue. Since the data model definition can be inherited from the existing semantic Web schema, the developer can choose the necessary properties from the schema via the GUI interface. By applying the data model definition, our framework generates all the necessary components for a spoken dialogue system based on the model-view-controller (MVC) model. In addition, template prompt entries, which the developer uses to specify the contents of the system utterance, are automatically generated. The fulfilled contents are translated to various languages using a translation Web service. The multilingual speech interaction is realized by an HTML5 speech API specifying only the "lang" attribute of the HTML element. Therefore, to realize a first prototype of a foreign language spoken dialogue system, the developer does not necessarily need knowledge of the target language.