How to Use a Neural Architecture to Map User Music Tastes to Voice Queries

Not all research goes to the final stages. It's either incorporated into other things or it just flat out doesn't go anywhere. Can you walk us through some examples or even just watch your approach to when things don't quite go the way that you'd hope they would? I have a really good example which was being very sad and then being very happy and so on. This is how research works. We're all working always together.