Pega Voice AI is customizable to specific use cases, industries, and dialects. Voice AI models can learn from the training data that customers provide to improve the overall accuracy of Voice AI’s real-time analysis and suggestions.
Voice AI training data requirements
Provide training data to Pega Voice AI to receive more customized suggestions.
You can provide any one of the following training data sets:
- 100 hours of uncompressed stereo call recordings in any audio file format, such as .wav.
- 100 hours of human-generated call transcripts and associated call recordings.
By using the VAD Record Only Mode property in the Voice AI configuration set, you can enable the Voice AI desktop app to automatically record calls for use as training data. For more information, see Configuring the Voice AI configuration set.
After Voice AI receives your training data, it uses the data to train the Voice AI speech and language models. The training process takes up to three weeks, after which the models are available for use.
If you experience any difficulties with data collection, data transfer, or usage of Voice AI models, contact your Pega representative.
Voice AI model configuration settings
To use a trained model that best suits your business needs, update the speech and language models in the Voice AI configuration set to the latest values that Pega provides to you.
For more information, see Configuring the Voice AI configuration set.