We have been troubleshooting some voice recognition calls. In the past the speech entries were able to be recovered and saved as a wav file so we can hear exactly what the caller said and compare that to the results in the IP logs. I have looked on our ASR and media servers and I do not see any directories that appear to contain any speech capture results from our Nuance server. Has anyone had experience finding and retrieving these types of files? If so, where were are they stored? Thanks,

One option is to log into the ASR web configuration interface and set the property "reco:ASRDebugWaveLogging" to true. When reco:ASRDebugWaveLogging property is enabled, the audio that is passed from Interactive Intelligence's software to Nuance's software is also saved to WAV files in the "Debug WAV logging directory." You can log into the speech server's web interface and go to Config > Properties to set the reco:ASRDebugWaveLogging property. Then go to Config > Parameters to view or edit the "Debug WAV logging directory" parameter. Alternatively, and perhaps it's already configured to do this, Nuance's software can also save the audio that it receives. This is part of Nuance's call logging. For Nuance Recognizer 9 (NR9), by default the call logs are saved to Program Files\Nuance\Recognizer\data\callLogs\. In Nuance Recognizer 10 (NR10), ProgramData\Nuance\Enterprise\Nuance\callLogs\ is used instead by default. The actual call logs will be several directories below, depending on other settings and the date and time of the call. Nuance call logging is configured in the Baseline.xml file located at Program Files\Nuance\Recognizer\config. In particular, the swirec_suppress_event_logging and swirec_suppress_waveform_logging settings should be set to 0 for NR9 or false for NR10. (False convolutedly means don't suppress, which means enabled.) When configuring Nuance call logs this way, only the endpointed audio is saved. That means the initial silence and background noise is not saved, for example while the prompt is playing and before the caller starts to speak. The reco:ASRDebugWaveLogging audio always contains all the audio that is passed to the ASR engine.