This document was generated from CDN thread
Created by: null on 29-06-2004 03:04:08 PM
Hi,
At run-time, how do elements distinguish between recognition results with low confidence versus recognition with high confidence? Is there any processing of n-best results?
Peter
Subject: RE: differentiating recognition confidence
Replied by: null on 08-07-2004 07:00:17 PM
Hi Peter,
The Audium Form element allows the developer to specify maxnbest results and also stores the returned nbest data - interpretation, confidence, etc. In addition, if the grammar returns multiple slots, the format of the stored data may differ from browser to browser. If there is some action required due to the recognition results such as low confidence, it is up to the developer to determine how to route the callflow and configure the next element accordingly. Runtime configurations can be used with the stored data with Java/XML dynamic configurations or substitution. More specific information about the Form element is available in the Audium Element Specifications document.
Note, it is always possible to create your own voice element that may include even more advanced dialog behavior.
Joe