1 paper
Mingu Lee, Jinkyu Lee, Hye Jin Jang +3
Multi-head attention mechanism is capable of learning various representations from sequential data while paying attention to different subsequences, e.g., word-pieces or syllables…