Using UniMRCP with PureConnect
This section describes how to connect to the PureConnect platform and enabling the enhanced and phone call models in UniMRCP/
- In Intelligent Automation, set the VUI Preference Collection low confidence threshold to 0.
- Under the Defaults section of the main application:
- Set the recognition incomplete timeout to 50000ms (this is only required when using UniMRCP).
- Set the Collection Low Confidence threshold to 0.0 in the IA application defaults.
- Turn on the enhanced model in UniMRCP config.
- Turn on the phone_call model in UniMRCP config.
- Ensure both the enhanced and phone_call model configs are enabled within UniMRCP.
The Phrase Hints feature (introduced in 9.0.108.00) and the Speech Contexts feature in UniMRCP can help in enabling better contextual understanding like differentiating between Agent and Asian.
Understanding accents that are not in the language set as default can affect performance and must be considered when developing voicebots.
Specifying a voice in PureConnect Cloud
Starting from release 9.106, you can specify a voice in GIA without specifying it in UniMRCP, you need to specify the desired voices in Interaction Administrator. See Phrase hints for more information.
How to control Pitch and Speed control
There is a UniMRCP patch for gss 1.7.3 – you can now globally configure the prosody rate and pitch at the UniMRCP config file.
You can change voices only when Genesys Voice Platform MCP is configured with Google Speech resources.
Configure the Default Server Settings for your locale e.g., Personas.Google.de-de.TTSVoiceName = e-DE-Standard-A,de-DE-Standard-B,de-DE-Wavenet-A,de-DE-Wavenet-B,de-DE-Wavenet-C,de-DE-Wavenet-D. After this configuration, apply the setting in the Persona configuration.
Support for gRPC
GVP supports a gRPC integration with Google Cloud Text-to-Speech and Speech-to-Text for Voicebots purchased from Genesys.
Contact Genesys Product Management for information on configuring GVP to connect to Genesys Voicebots.
Improving Confidence Scores
If the confidence score returned from GVP is below the threshold set in Intelligent Automation, reduce the threshold value to 0. This is applicable when GVP returns a no match value.