A re-entrant reusable dialog component framework for context sharing and mixed initiative

Abstract

This paper details a proposal for context sharing objects and subdialogs in order to support their parallel activation and mixed initiative across objects, subdialogs and documents. This is achieved by requiring that objects and subdialogs be re-entrant: they can be initiated, interrupted, inspected and resumed with partially filled result/state object. When used in the context of VoiceXML, this proposal requires extensions of the VoiceXML specification, the form interpretation algorithm and the VoiceXML execution model. Such an extension is anyway expected in order to provide support for advanced mixed initiative, i.e. beyond the fields of a same form or within a same document as currently specified in VoiceXML 1.0 [1 J. We also introduce the notion of service objects in addition to interaction objects as currently addressed by the W3C Voice Activity.