@TOC
# 代码 ```python # -*-coding:utf-8-*- import time from keras_preprocessing.sequence import pad_sequences start = time.time() import tensorflow as tf from tensorflow import keras from tensorflow.keras.preprocessing.text import Tokenizer import numpy as np import tensorflow.keras.layers as layers; tokenizer = Tokenizer() # 要学习的语句 data = "Please forgive me for falling in love with you.\n" \ "Forgive me for loving you with all my heart.\n" \ "Forgive me for never wanting to be apart.\n" \ "Two star-crossed lovers in perfect harmony Just give me a chance and you will agree.\n" \ "I was meant for you.\n" \ "And you were meant for me." print(data) corpus = data.split("\n") print("---------------------------------------------------") print(corpus) tokenizer.fit_on_texts(corpus) total_words = len(tokenizer.word_index)+1 input_seq = [] for line in corpus: token_list = tokenizer.texts_to_sequences([line])[0] for i in range(1,len(token_list)): n_gram_seq = token_list[:i+1] input_seq.append(n_gram_seq) max_seq_len = max([len(x) for x in input_seq]) # 前置0填充 input_seq = np.array(pad_sequences(input_seq,maxlen=max_seq_len,padding='pre')) print("---------------------------------------------------") print(input_seq) # 特征和标签的划分 xs = input_seq[:,:-1] label = input_seq[:,-1] # 将标签转化为one-hot独热编码 ys = tf.keras.utils.to_categorical(label,num_classes=total_words) # 模型 model = keras.Sequential() model.add(layers.Embedding(total_words,240,input_length=max_seq_len-1)) model.add(layers.Bidirectional(layers.LSTM(150))) model.add(layers.Dense(total_words,activation='softmax')) adam = keras.optimizers.Adam(lr=0.01) model.compile(loss="categorical_crossentropy",optimizer=adam,metrics=['accuracy']) # 开始训练 history = model.fit(xs,ys,epochs=50,verbose=1) # 测试 test_text = "I want to" next_words = 20 for _ in range(next_words): token_list = tokenizer.texts_to_sequences(([test_text]))[0] token_list = pad_sequences([token_list],maxlen=max_seq_len-1,padding='pre') predicted = model.predict_classes(token_list,verbose=0) # print(tokenizer.word_index.items()) output_word = "" for word,index in tokenizer.word_index.items(): if index == predicted: output_word = word test_text = (test_text + " " + output_word) break print(test_text) print("共用时:",(time.time()-start),"秒") ```
# 效果展示 - I want to … I want to meant for you with all my heart 我想全心全意为你着想 - I hate you … I hate you were meant for me for never wanting to be apart 我恨你是我命中注定永远不想分开的 - I want to eat … I want to eat star crossed lovers in perfect harmony just give me a chance 我想要完美和谐吃着满天星的恋人给我一个机会