問答模型 `keras.layers.add`、`keras.layers.concatenate`  **用API實現：** * 文本： ~~~ embedded_text = layers.Embedding(text_vocabulary_size, 64)(text_input) ↑將輸入嵌入長度為64的向量 encoded_text = layers.LSTM(32)(embedded_text) ↑利用LSTM將向量編碼為單個向量 question_input = Input(shape=(None,),dtype='int32',name='question') ~~~ * 問題： ~~~ embedded_question = layers.Embedding( question_vocabulary_size, 32)(question_input) encoded_question = layers.LSTM(16)(embedded_question) ~~~ * concatenate： ~~~ concatenated = layers.concatenate([encoded_text, encoded_question],axis=-1) ~~~ * 回答： ~~~ answer = layers.Dense(answer_vocabulary_size, activation='softmax')(concatenated) ~~~