Model based development of speech recognition grammar for VoiceXML

Abstract

Speech interaction is a natural form of interaction between human and devices. This interaction technology is currently in high demand but often has limitations. A limited form of interaction is thus in use to achieve best possible efficiency. The limited form of speech interaction uses direct commands instead of complete natural language. The VoiceXML is a W3C (World wide consortium) recommended web based speech interaction application development language that performs the dialogue management. It has been used as a base language in this thesis for the case study.

The VoiceXML uses a grammatical base to recognise the user utterance for the commands. This thesis applies the model based development approach to create a hierarchical data model for the speech grammar. Further, the grammar has been separated from the interaction code. MetaEdit+ tool has been used for developing the grammar model and for generating the grammar file.

The approach is further compared with other grammar models. In conclusion, the applied approach is found suitable for the grammar modelling in VoiceXML application development.

Descriptors: Data Model, Grammar Model, MetaEdit+ , VoiceXML, Grammar generation, Hierarchical Model, Speech interaction