Session Audio
-
sexta-feira, 10 de agosto de 2007 17:13
In the MSS trace logging properties, we have it set to record All Audio for 100% of calls.
However, when I play or export the audio, I only hear one side of the call (i.e., I don't hear the prompts that MSS is playing). Is this how it's supposed to be? I had assumed that "All Audio" meant both sides of the call. Is there any way to record both sides of the entire call?
Todas as Respostas
-
sexta-feira, 10 de agosto de 2007 21:13
Search for "How to: Enable Event Tracing of Call Activity" in the Speech Server documentation. -
sexta-feira, 10 de agosto de 2007 21:54
Are you playing or exporting the audio from the Analytics and Tuning studio?
When you import the ETL into the SQL database did you use the command line option /audio: Session (without the space)?
/audio:<level> This flag indicates how much audio to import.
Options are "None", "Recognizer", "Turn"
and "Session". (Default: "None" if /filter is Reports,
otherwise defaults to "Recognizer") -
segunda-feira, 13 de agosto de 2007 18:15
I reimported the log files using /audio
ession this time. Now when I try to listen to the session audio I get a dialog that says, "The system cannot find the file specified."
When I try to save the session audio, after I select the folder to save to, nothing gets saved.
-
segunda-feira, 13 de agosto de 2007 19:41
Did you do import with a clean database? Or try to import again on the same database? I don't think it would work with an re-import with the same set of sessions.
Would it be possible to try to import to a clean database?
-
segunda-feira, 13 de agosto de 2007 20:04I created a new database and re-imported everything. I didn't get any audio as I described above.
-
segunda-feira, 13 de agosto de 2007 20:24
I see.
Can you try MssContentExtract on your ETL?
MssContentExtract.exe <log file>
[/out:<path>]
[/nogrammar]
[/verbose]And see if there are full session audio extracted?
If not, then it might be an issue with your logging settings.
Can you confirm that you have checked:
Analytics and tuning events
All audio for: 100 percent of calls
To be safe, please have the Recognizer UN-checked.
Please make sure you have apply these settings before making the calls. Also, if you want the grammars to be logged, make sure you restart the OCS 2007 Speech Server service before making the calls.
-
segunda-feira, 13 de agosto de 2007 20:52
I ran MssContentExtract on the etl files and it did output wav files. But these wav files are only one side of the call. They do not include audio from MSS.
I have check the logging settings again and they are set as you described.
-
segunda-feira, 13 de agosto de 2007 23:10Can you tell me if you are using a TIM?
-
terça-feira, 14 de agosto de 2007 13:16No.
-
terça-feira, 14 de agosto de 2007 13:49
After listening to some of the wav files that are created with MssContentExtract, it sounds like what is actually being recorded is the recognizer audio. Not the session audio.
I have confirmed several times that we do not have recognizer audio checked in the settings and that we have it set to record all audio for 100% of calls. MSS has been restarted with this setting also.
-
terça-feira, 14 de agosto de 2007 17:00
Okay. I think that the log files I was originally trying this with must have been created without session audio logging turned on. I reimported some log files from today and I can get the session audio in the
Analytics and Tuning Studio. For some reason, though, I don't get the session audio using MssContentExtract, but that's fine. -
terça-feira, 14 de agosto de 2007 19:26
Yes, you are correct. MssContentExtract.exe only extract recognizer audio and grammar, not the session audio. My mistake.
