protocol group notes 7/28/2011

In today's protocol call we discussed how re-reco might work, and settled on the following decisions:

1. Use the existing Save-Waveform, Waveform-URI, and Input-Waveform-URI headers. The Waveform-URI header will refer to the audio captured so far.

2. In continuous recognition, send START/END-OF-SPEECH events throughout the session. These are useful for UI indicators to the user, and also contain the source-time header, which the client MAY use to calculate a re-reco interval.

3. In the RECOGNITION-COMPLETE event, explicitly include both the start and end time, which may be used as a re-reco interval (presumably this would also apply to the INTERMEDIATE-RESULT event, although we didn't discuss this).

4. Consider using the media fragment URI spec (http://www.w3.org/TR/media-frags/) or some subset, for the client to specify the time interval for a re-recognition. We need to review this and discuss in mail.



We also discussed that retained audio is more than just a URI - the client should be able to retrieve the audio if it has appropriate credentials.



We discussed that language selection is not explicitly discussed in the draft and should be.



For next week, we have some open topics in mail, that we will discuss if they are unresolved in mail during the week.