End-to-End Spoken Language Understanding Using Transformer Networks and Self-Supervised Pre-Trained Features.