3 ms·Embed, encode, attend, predict: The new deep learning formula for NLP models5 points by syllogism 10y ago