Get spoken input as string

Is i开发者_如何学JAVAt possible to get a string representation of what a user said in an VoiceXML application? For example to use it in a database query.

Most examples i've seen so far just try to match against a (xml-)defined number of choices. This is unsuitable if one wants to use the input to match against a database with let's say >100k entries.

Thanks

You can get the spoken input as a string with utterance attribute of the last result (VoiceXML specification)

However, that's not your problem. You want an open ended grammar or a system that can handle 100K entries. That is an artifact of the recognition engine, not the VoiceXML language. In theory, you could build a 100K SRGS grammar. In practice, that's probably far too large, if they are all unique inputs without a cleaner way to define. You may be able to do this with native grammars with engines like Nuance with either a static list or entries or a statistical language model made up of item components (assuming they share common components).

You can look at dictation engines. I don't think I've seen any attached to VoiceXML platforms as they usually aren't well aligned to the normal VoiceXML use cases and most dictation engines require training and regular sentence structures to fully identify input.