Is this true? On Google I only see mentions of using voice data to improve their speech transcribing.