An IMS-SOA Architecture for Multimodal Interaction using Distributed Keyword Dictionaries

The multimodal architecture discussed in this paper is a distributed implementation of W3C Multimodal Interaction model. Certain functions of the W3C Interaction Manager are now executed by the multimodal client, while others have been distributed to remote SOA compliant multimodal enablers. A Multimodal Interaction AS manages the interaction with the local Speech Platform and communicates with neighbor enablers to retrieve interaction rules/recognition models. Remote speech recognition is used whenever the semantic analysis of the information associated to a given command does not lead to words in the local dictionary, or when the dialect of the user determines significant recognition errors.