b What are the word embeddings? Why do people use this? Why do people use encoder-decoder RNNs rather than plain sequence-to- sequence RNNs for automatic translation? c In the attention model, it utilizes the key vectors K, query vectors Q and value vectors V . What is the matrix equation of outputs with the attention model in terms of Q, K, V ? Why does the output matrix need to be divided by dk , i.e., the dimension length of vectors in Q, K, V