A non-invasive, speech-based approach for automatic suicide risk assessment can achieve high accuracy when combined with patient metadata.
