By Craig Kirkwood, Planet Publish Editor
July 2, 2002

The World Wide Web Consortium (W3C) has issued the Speech Recognition Grammar Specification as a W3C Candidate Recommendation. Speech grammars allow voice-based application authors to create rules describing what users are expected to say after listening to each application prompt.

Advancement of this document to Candidate Recommendation is a statement that the specification is stable, and an invitation to the Web development community at large to make further implementations and provide technical feedback.

Talking and Listening to the Web Requires a Framework

The W3C Voice Browser Working Group is defining a suite of markup languages covering dialog (VoiceXML 2.0); speech synthesis (Speech Synthesis Markup Language); speech recognition (Speech Grammar, Stochastic Language Models, Semantic Interpretation for Speech Recognition, Natural Language Semantics); call control (Voice Browser Call Control: CCXML) and other aspects of interactive voice response applications. All of these contribute to the W3C Speech Interface Framework.

"W3C is working on expanding the Web to include access from the one billion plus telephones worldwide," explained Dave Raggett, W3C Voice Browser Activity lead and W3C Fellow from Openwave Systems. "People will be able to interact via spoken commands and listening to recorded speech, synthetic speech and music. This will also benefit people with visual impairments or needing Web access while keeping theirs hands & eyes free for other things."

The Speech Recognition Grammar specification is the first of the W3C Speech Interface Framework suite to be advanced to Candidate Recommendation status.

Speech Grammars Provide XML Rules for Understanding Speech

Speech Grammars allow authors to specify rules covering the sequences of words that users are expected to say in particular contexts. The W3C Speech Recognition Grammar specification defines an XML language for context-free speech grammars.

In more practical terms, Speech grammars make it easier to allow for variations in the way people answer questions. For example, people may say "Yes", "Sure", or "Fine", or may say dates as "Tomorrow", "July 8th" or "8 July". The Speech Recognition Grammar Specification provides an XML language for application authors to define rules covering all the expected combinations of words that users are likely to say in a given context.

Because the set of rules are independent of any individual markup language, they may be used with versions of VoiceXML, or with other dialog markup languages.

Implementers Encouraged to Further Build on Speech Interface Framework

W3C welcomes implementation reports for the Speech Grammar specification. The Implementation Report Plan provides an extensive set of test cases; there is already significant implementation experience amongst the companies involved in the W3C Voice Browser Working Group that developed the specification. W3C expects at least two interoperable implementations of each required feature in the specification as a condition on advancing it to Proposed Recommendation status.

Leaders in Voice Industry Produce Speech Grammar Specification through W3C Consensus-Based Process

The W3C Voice Browser Working Group consists of W3C Members and invited experts who hold leadership roles in the development of Voice Interaction technologies. Those active in the Speech Grammars specification include BeVocal, Cisco Systems, Comverse, IBM, Locus Dialogue, Lucent, Microsoft, Nuance Communications, Openwave, Philips, PipeBeach, Scansoft, SpeechWorks International, Tellme Networks and Unisys. Many have provided testimonials for the Speech Grammar specification, indicating current and planned product support.

For further information visit: http://www.w3.org/TR/2002/CR-speech-grammar-20020626/