Why humans review “Hey Portal” voice interactions
When storage of your voice interactions is turned on, we use both machine learning and trained reviewers to process information to improve, troubleshoot, and train our voice services and to help make our voice services get smarter, more accurate and better for everyone.
Making sure that our voice services can understand and respond to a wide range of speech patterns, phrases, local dialects and accents is important to us. Trained reviewers are particularly helpful in identifying voice interactions that machines are likely to misinterpret. For example, someone might say "Newark,” but with their accent, Meta Portal could interpret what they said as "New York.” Trained reviewers are also useful in evaluating how well our voice services execute voice commands. For example, we can evaluate if Meta Portal actually called the person you meant to call. This helps improve our voice services to respond more accurately over time.
Your voice interactions will not be reviewed by trained reviewers if you turn off the storage of your Meta Portal voice interactions.