Enabling voice analytics
Important
As a condition of using this feature, you acknowledge that the collection, use, storage, and retention of your caller’s biometric identifiers and biometric information (“biometric data”) in the form of a digital voice profile requires the caller’s informed consent via a written release. Such consent is required under various state laws, including biometrics laws in Illinois, Texas, Washington and other state privacy laws.
You must provide a written release to each caller through a process that clearly reflects each caller’s informed consent before using Amazon Chime SDK voice analytics service, as required under the terms of your agreement with AWS governing your use of the service.
Note
To enable voice analytics, you must have at least one Amazon Chime SDK Voice Connector and at least one Amazon Chime SDK call analytics configuration. For more information about creating Voice Connectors, see Creating an Amazon Chime SDK Voice Connector. For information about creating a call analytics configuration, see Creating call analytics configurations. For information about updating a configuration, see
The topics in this section explain how to enable Amazon Chime SDK voice analytics for Amazon Chime SDK Voice Connectors. Voice analytics uses machine learning to enable some or all of the following:
Speaker search – Converts a caller's voice into a vector embedding. It then compares the embedding to a database of known voice embeddings. If it finds a match or matches, it returns a ranked list of high-likelihood voice profile ID matches, along with a corresponding set of confidence scores.
Note
Speaker search is not designed for authentication or identity verification use cases, such as verifying the identity of a speaker with extremely high accuracy.
Voice tone analysis – Predicts the sentiment expressed in a speech signal based on a combined analysis of linguistic and tonal information.
Note
As a reminder, you must comply with all legal requirements when using voice tone analysis. This includes obtaining consent from the speaker as required by law, and not using the feature to make decisions about the speaker that would produce legal or similarly significant impacts, such as employment, housing, credit worthiness, or financial offers.
To enable voice analytics, administrators use the Amazon Chime SDK console to do the following:
Configure Voice Connectors to use one or more of the features listed above.
Create notification targets. Notification targets asynchronously receive voice analysis events, and you must have at least one target.
Create voice profile domains. Voice profile domains contain sets of voice profiles. In turn, a voice profile consists of a vector embedding of a caller's voice, plus a unique ID. By default, you can create 3 voice profile domains, and each domain can house 20,000 voice profiles. You can request an increase for both limits as needed.
Developers can use a set of APIs to do those same tasks. For more information, see Using the Amazon Chime SDK PSTN voice analytics service, in the Amazon Chime SDK Developer guide.
