Is this specifically for audio or does it cover traditional text sentiment analysis too?