Text this: Multi-Modal Enhancement Transformer Network for Skeleton-Based Human Interaction Recognition.