chatbot_inference_non-Attention.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pickle\n",
    "import operator\n",
    "\n",
    "question = np.load('pad_question.npy')\n",
    "answer = np.load('pad_answer.npy')\n",
    "answer_o = np.load('answer_o.npy')\n",
    "with open('vocab_bag.pkl', 'rb') as f:\n",
    "    words = pickle.load(f)\n",
    "with open('pad_word_to_index.pkl', 'rb') as f:\n",
    "    word_to_index = pickle.load(f)\n",
    "with open('pad_index_to_word.pkl', 'rb') as f:\n",
    "    index_to_word = pickle.load(f)\n",
    "vocab_size = len(word_to_index) + 1\n",
    "maxLen=20"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from keras.preprocessing import sequence\n",
    "def generate_train(batch_size):\n",
    "    print('\\n*********************************generate_train()*********************************')\n",
    "    steps=0\n",
    "    question_ = question\n",
    "    answer_ = answer\n",
    "    while True:\n",
    "        batch_answer_o = answer_o[steps:steps+batch_size]\n",
    "        batch_question = question_[steps:steps+batch_size]\n",
    "        batch_answer = answer_[steps:steps+batch_size]\n",
    "        outs = np.zeros([batch_size, maxLen, vocab_size], dtype='float32')\n",
    "        for pos, i in enumerate(batch_answer_o):\n",
    "            for pos_, j in enumerate(i):\n",
    "                if pos_ > 20:\n",
    "                    print(i)\n",
    "                outs[pos, pos_, j] = 1 # one-hot\n",
    "        yield [batch_question, batch_answer], outs\n",
    "        steps += batch_size\n",
    "        if steps == 100000:\n",
    "            steps = 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.layers import Embedding\n",
    "from keras.layers import Input, Dense, LSTM, TimeDistributed, Bidirectional, Dropout, Concatenate, RepeatVector, Activation, Dot\n",
    "from keras.layers import concatenate, dot                    \n",
    "from keras.models import Model\n",
    "from keras.utils import plot_model\n",
    "from keras.callbacks import ModelCheckpoint, TensorBoard\n",
    "from keras.initializers import TruncatedNormal\n",
    "import pydot\n",
    "import os, re"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From c:\\users\\王嘉意\\appdata\\local\\programs\\python\\python36\\lib\\site-packages\\tensorflow\\python\\framework\\op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Colocations handled automatically by placer.\n",
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_answer (InputLayer)       (None, None)         0                                            \n",
      "__________________________________________________________________________________________________\n",
      "input_question (InputLayer)     (None, None)         0                                            \n",
      "__________________________________________________________________________________________________\n",
      "embedding_1 (Embedding)         (None, None, 100)    4335000     input_question[0][0]             \n",
      "                                                                 input_answer[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "encoder_lstm (LSTM)             [(None, None, 512),  1255424     embedding_1[0][0]                \n",
      "__________________________________________________________________________________________________\n",
      "decoder_lstm (LSTM)             [(None, None, 512),  1255424     embedding_1[1][0]                \n",
      "                                                                 encoder_lstm[0][1]               \n",
      "                                                                 encoder_lstm[0][2]               \n",
      "__________________________________________________________________________________________________\n",
      "time_distributed_1 (TimeDistrib (None, None, 256)    131328      decoder_lstm[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "time_distributed_2 (TimeDistrib (None, None, 43350)  11140950    time_distributed_1[0][0]         \n",
      "==================================================================================================\n",
      "Total params: 18,118,126\n",
      "Trainable params: 18,118,126\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "truncatednormal = TruncatedNormal(mean=0.0, stddev=0.05)\n",
    "embed_layer = Embedding(input_dim=vocab_size, \n",
    "                        output_dim=100, \n",
    "                        mask_zero=True,\n",
    "                        input_length=None,\n",
    "                        embeddings_initializer= truncatednormal)\n",
    "LSTM_encoder = LSTM(512,\n",
    "                      return_sequences=True,\n",
    "                      return_state=True,\n",
    "                      kernel_initializer= 'lecun_uniform',\n",
    "                      name='encoder_lstm'\n",
    "                        )\n",
    "LSTM_decoder = LSTM(512, \n",
    "                    return_sequences=True, \n",
    "                    return_state=True, \n",
    "                    kernel_initializer= 'lecun_uniform',\n",
    "                    name='decoder_lstm'\n",
    "                   )\n",
    "\n",
    "#encoder输入 与 decoder输入\n",
    "input_question = Input(shape=(None, ), dtype='int32', name='input_question')\n",
    "input_answer = Input(shape=(None, ), dtype='int32', name='input_answer')\n",
    "\n",
    "input_question_embed = embed_layer(input_question)\n",
    "input_answer_embed = embed_layer(input_answer)\n",
    "\n",
    "\n",
    "encoder_lstm, question_h, question_c = LSTM_encoder(input_question_embed)\n",
    "\n",
    "decoder_lstm, _, _ = LSTM_decoder(input_answer_embed, \n",
    "                                  initial_state=[question_h, question_c])\n",
    "\n",
    "# attention = dot([decoder_lstm, encoder_lstm], axes=[2, 2])\n",
    "# attention = Activation('softmax')(attention)\n",
    "# context = dot([attention, encoder_lstm], axes=[2,1])\n",
    "# decoder_combined_context = concatenate([context, decoder_lstm])\n",
    "decoder_combined_context = decoder_lstm\n",
    "\n",
    "# Has another weight + tanh layer as described in equation (5) of the paper\n",
    "decoder_dense1 = TimeDistributed(Dense(256,activation=\"tanh\"))\n",
    "decoder_dense2 = TimeDistributed(Dense(vocab_size,activation=\"softmax\"))\n",
    "output = decoder_dense1(decoder_combined_context) # equation (5) of the paper\n",
    "output = decoder_dense2(output) # equation (6) of the paper\n",
    "\n",
    "model = Model([input_question, input_answer], output)\n",
    "\n",
    "model.compile(optimizer='rmsprop', loss='categorical_crossentropy')\n",
    "\n",
    "model.load_weights('models/W--157-0.6163-.h5')\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_question (InputLayer)  (None, None)              0         \n",
      "_________________________________________________________________\n",
      "embedding_1 (Embedding)      (None, None, 100)         4335000   \n",
      "_________________________________________________________________\n",
      "encoder_lstm (LSTM)          [(None, None, 512), (None 1255424   \n",
      "=================================================================\n",
      "Total params: 5,590,424\n",
      "Trainable params: 5,590,424\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_answer (InputLayer)       (None, None)         0                                            \n",
      "__________________________________________________________________________________________________\n",
      "embedding_1 (Embedding)         (None, None, 100)    4335000     input_answer[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "input_1 (InputLayer)            (None, 512)          0                                            \n",
      "__________________________________________________________________________________________________\n",
      "input_2 (InputLayer)            (None, 512)          0                                            \n",
      "__________________________________________________________________________________________________\n",
      "decoder_lstm (LSTM)             [(None, None, 512),  1255424     embedding_1[1][0]                \n",
      "                                                                 input_1[0][0]                    \n",
      "                                                                 input_2[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "time_distributed_1 (TimeDistrib (None, None, 256)    131328      decoder_lstm[1][0]               \n",
      "__________________________________________________________________________________________________\n",
      "time_distributed_2 (TimeDistrib (None, None, 43350)  11140950    time_distributed_1[1][0]         \n",
      "==================================================================================================\n",
      "Total params: 16,862,702\n",
      "Trainable params: 16,862,702\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "question_model = Model(input_question, [encoder_lstm, question_h, question_c])\n",
    "question_model.summary()\n",
    "answer_h = Input(shape=(512,))\n",
    "answer_c = Input(shape=(512,))\n",
    "encoder_lstm = Input(shape=(maxLen,512))\n",
    "target, h, c = LSTM_decoder(input_answer_embed, initial_state=[answer_h, answer_c])\n",
    "# attention = dot([target, encoder_lstm], axes=[2, 2])\n",
    "# attention_ = Activation('softmax')(attention)\n",
    "# context = dot([attention_, encoder_lstm], axes=[2,1])\n",
    "# decoder_combined_context = concatenate([context, target])\n",
    "decoder_combined_context = target\n",
    "output = decoder_dense1(decoder_combined_context) # equation (5) of the paper\n",
    "output = decoder_dense2(output) # equation (6) of the paper\n",
    "answer_model = Model([input_answer, answer_h, answer_c, encoder_lstm], [output, h, c])\n",
    "answer_model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.preprocessing import sequence\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib\n",
    "import jieba\n",
    "import requests\n",
    "def act_weather(city):\n",
    "    #TODO: Get weather by api\n",
    "    url = 'http://wthrcdn.etouch.cn/weather_mini?city=' + city\n",
    "    page = requests.get(url)\n",
    "    data = page.json()\n",
    "    temperature = data['data']['wendu']\n",
    "    notice = data['data']['ganmao']\n",
    "    outstrs = \"地点： %s\\n气温： %s\\n注意： %s\" % (city, temperature, notice)\n",
    "    return outstrs + ' EOS'\n",
    "def input_question(seq):\n",
    "    seq = jieba.lcut(seq.strip(), cut_all=False)\n",
    "    sentence = seq\n",
    "    try:\n",
    "        seq = np.array([word_to_index[w] for w in seq])\n",
    "    except KeyError:\n",
    "        seq = np.array([36874, 165, 14625])\n",
    "    seq = sequence.pad_sequences([seq], maxlen=maxLen,\n",
    "                                          padding='post', truncating='post')\n",
    "    print(seq)\n",
    "    return seq, sentence\n",
    "def decode_greedy(seq, sentence):\n",
    "    question = seq\n",
    "    for index in question[0]:\n",
    "        if int(index) == 5900:\n",
    "            for index_ in question[0]:\n",
    "                if index_ in [7851, 11842,2406, 3485, 823, 12773, 8078]:\n",
    "                    return act_weather(index_to_word[index_])\n",
    "    answer = np.zeros((1, 1))\n",
    "#     attention_plot = np.zeros((20, 20))\n",
    "    answer[0, 0] = word_to_index['BOS']\n",
    "    i=1\n",
    "    answer_ = []\n",
    "    flag = 0\n",
    "    encoder_lstm_, question_h, question_c = question_model.predict(x=question, verbose=1)\n",
    "#     print(question_h, '\\n')\n",
    "    while flag != 1:\n",
    "        prediction, prediction_h, prediction_c = answer_model.predict([\n",
    "            answer, question_h, question_c, encoder_lstm_\n",
    "        ])\n",
    "#         attention_weights = attention.reshape(-1, )\n",
    "#         attention_plot[i] = attention_weights\n",
    "        word_arg = np.argmax(prediction[0, -1, :])#\n",
    "        answer_.append(index_to_word[word_arg])\n",
    "        if word_arg == word_to_index['EOS']  or i > 20:\n",
    "            flag = 1\n",
    "        answer = np.zeros((1, 1))\n",
    "        answer[0, 0] = word_arg\n",
    "        question_h = prediction_h\n",
    "        question_c = prediction_c\n",
    "        i += 1\n",
    "    result = ' '.join(answer_)\n",
    "#     attention_plot = attention_plot[:len(result.split(' ')), :len(sentence)]\n",
    "#     plot_attention(attention_plot, sentence, result.split(' '))\n",
    "    return ' '.join(answer_)\n",
    "def decode_beamsearch(seq, beam_size):\n",
    "    question = seq\n",
    "    encoder_lstm_, question_h, question_c = question_model.predict(x=question, verbose=1)\n",
    "    sequences = [[[word_to_index['BOS']], 1.0, question_h, question_c]]\n",
    "    answer = np.zeros((1, 1))\n",
    "    answer[0, 0] = word_to_index['BOS']\n",
    "    answer_ = ''\n",
    "    flag = 0\n",
    "    last_words = [word_to_index['BOS']]\n",
    "    for i in range(maxLen):\n",
    "        all_candidates = []\n",
    "        for j in range(len(sequences)):\n",
    "            s, score, h, c = sequences[j]\n",
    "            last_word = s[-1]\n",
    "            if not isinstance(last_word, int):\n",
    "                last_word=last_word[-1]\n",
    "            answer[0, 0] = last_word\n",
    "            output, h, c, _ = answer_model.predict([answer, h, c, encoder_lstm_])\n",
    "            output = output[0, -1]\n",
    "            for k in range(len(output)):\n",
    "                candidate = [seq+[k], score*-np.log(output[k]), h, c]\n",
    "            all_candidates.append(candidate)\n",
    "        ordered = sorted(all_candidates, key=lambda tup:tup[1])\n",
    "        sequences = ordered[:beam_size]\n",
    "    answer_ = sequences[0][0]\n",
    "    print(answer_[0])\n",
    "    answer_ = [index_to_word[x] for x in answer_[0] if (x!=0)]\n",
    "    answer_ = ' '.join(answer_)\n",
    "    return answer_\n",
    "def plot_attention(attention, sentence, predicted_sentence):\n",
    "    zhfont = matplotlib.font_manager.FontProperties(fname='simkai.ttf')\n",
    "    fig = plt.figure(figsize=(10,10))\n",
    "    ax = fig.add_subplot(1, 1, 1)\n",
    "    attention = [x[::-1] for x in attention]\n",
    "    ax.matshow(attention, cmap='viridis')\n",
    "    fontdict = {'fontsize': 20}\n",
    "    ax.set_xticklabels([''] + sentence, fontdict=fontdict,fontproperties=zhfont)\n",
    "    ax.set_yticklabels([''] + predicted_sentence, fontdict=fontdict, fontproperties=zhfont)\n",
    "#     ax.yaxis.set_ticks_position('right') #y轴刻度位置靠右\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "你饿不饿\n",
      "[[ 10 729  36 729   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "    0   0]]\n",
      "['你', '饿', '不', '饿']\n",
      "1/1 [==============================] - 0s 7ms/step\n",
      "ANSWER:  饿 啊 你 都 不 给 我 不 我 自己 想 EOS\n",
      "你身高多少\n",
      "[[  10 8113  123    0    0    0    0    0    0    0    0    0    0    0\n",
      "     0    0    0    0    0    0]]\n",
      "['你', '身高', '多少']\n",
      "1/1 [==============================] - 0s 7ms/step\n",
      "ANSWER:  有 多少 的 多少 EOS\n",
      "你体重多少\n",
      "[[   10 15708   123     0     0     0     0     0     0     0     0     0\n",
      "      0     0     0     0     0     0     0     0]]\n",
      "['你', '体重', '多少']\n",
      "1/1 [==============================] - 0s 15ms/step\n",
      "ANSWER:  你 说 呢 能 不 在 吗 EOS\n",
      "你吃饭了没\n",
      "[[ 10 230  57  61   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "    0   0]]\n",
      "['你', '吃饭', '了', '没']\n",
      "1/1 [==============================] - 0s 7ms/step\n",
      "ANSWER:  我 吃 电 的 哦 EOS\n",
      "你男的女的\n",
      "[[ 10 243  15  98  15   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "    0   0]]\n",
      "['你', '男', '的', '女', '的']\n",
      "1/1 [==============================] - 0s 8ms/step\n",
      "ANSWER:  女生 啦 EOS\n",
      "我生气了，哄哄我\n",
      "[[   56  2074    57 36966 11692    56     0     0     0     0     0     0\n",
      "      0     0     0     0     0     0     0     0]]\n",
      "['我', '生气', '了', '，', '哄哄', '我']\n",
      "1/1 [==============================] - 0s 12ms/step\n",
      "ANSWER:  发怒 1 分钟 便 失去 60 秒 的 幸福 EOS\n",
      "干嘛呢\n",
      "[[229 171   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "    0   0]]\n",
      "['干嘛', '呢']\n",
      "1/1 [==============================] - 0s 11ms/step\n",
      "ANSWER:  谁 爱 你 啊 EOS\n",
      "天气太热了，真难受\n",
      "[[36874   165 14625     0     0     0     0     0     0     0     0     0\n",
      "      0     0     0     0     0     0     0     0]]\n",
      "['天气', '太热', '了', '，', '真', '难受']\n",
      "1/1 [==============================] - 0s 8ms/step\n",
      "ANSWER:  当然 都 吗 你 不 认识 他 EOS\n",
      "你多大年龄\n",
      "[[  10  598  982 4593    0    0    0    0    0    0    0    0    0    0\n",
      "     0    0    0    0    0    0]]\n",
      "['你', '多', '大', '年龄']\n",
      "1/1 [==============================] - 0s 8ms/step\n",
      "ANSWER:  有 在 那么 小心 大 是 我 还是 我 这么 我要 他 和 牛 他 EOS\n",
      "青岛天气太热了\n",
      "[[36874   165 14625     0     0     0     0     0     0     0     0     0\n",
      "      0     0     0     0     0     0     0     0]]\n",
      "['青岛', '天气', '太热', '了']\n",
      "1/1 [==============================] - 0s 8ms/step\n",
      "ANSWER:  当然 都 吗 你 不 认识 他 EOS\n",
      "你叫什么名字\n",
      "[[  10  300   30 1348    0    0    0    0    0    0    0    0    0    0\n",
      "     0    0    0    0    0    0]]\n",
      "['你', '叫', '什么', '名字']\n",
      "1/1 [==============================] - 0s 15ms/step\n",
      "ANSWER:  我 不 叫 EOS\n",
      "你的名字是什么\n",
      "[[  10   15 1348   29   30    0    0    0    0    0    0    0    0    0\n",
      "     0    0    0    0    0    0]]\n",
      "['你', '的', '名字', '是', '什么']\n",
      "1/1 [==============================] - 0s 14ms/step\n",
      "ANSWER:  哦 EOS\n",
      "我生气了\n",
      "[[  56 2074   57    0    0    0    0    0    0    0    0    0    0    0\n",
      "     0    0    0    0    0    0]]\n",
      "['我', '生气', '了']\n",
      "1/1 [==============================] - 0s 8ms/step\n",
      "ANSWER:  发怒 1 分钟 便 失去 60 秒 的 幸福 EOS\n",
      "你是机器人么\n",
      "[[  10   29 3197  126    0    0    0    0    0    0    0    0    0    0\n",
      "     0    0    0    0    0    0]]\n",
      "['你', '是', '机器人', '么']\n",
      "1/1 [==============================] - 0s 9ms/step\n",
      "ANSWER:  我 是 三毛 的 情人 EOS\n",
      "你是机器么\n",
      "[[ 10  29 489 126   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "    0   0]]\n",
      "['你', '是', '机器', '么']\n",
      "1/1 [==============================] - 0s 7ms/step\n",
      "ANSWER:  你 丫 的 才 是 狗 还是 条 疯狗 EOS\n",
      "你是人类么\n",
      "[[  10   29 2659  126    0    0    0    0    0    0    0    0    0    0\n",
      "     0    0    0    0    0    0]]\n",
      "['你', '是', '人类', '么']\n",
      "1/1 [==============================] - 0s 8ms/step\n",
      "ANSWER:  我 是 女 的 你 想 试试 吗 EOS\n",
      "机器人会比人类聪明么\n",
      "[[3197   22  940 2659  302  126    0    0    0    0    0    0    0    0\n",
      "     0    0    0    0    0    0]]\n",
      "['机器人', '会', '比', '人类', '聪明', '么']\n",
      "1/1 [==============================] - 0s 7ms/step\n",
      "ANSWER:  哈哈 努力 学 多点 自然 就 聪明 了 EOS\n",
      "机器人比人类聪明么\n",
      "[[3197  940 2659  302  126    0    0    0    0    0    0    0    0    0\n",
      "     0    0    0    0    0    0]]\n",
      "['机器人', '比', '人类', '聪明', '么']\n",
      "1/1 [==============================] - 0s 6ms/step\n",
      "ANSWER:  的 人 是 我 主人 的 什么 EOS\n",
      "青岛天气很热\n",
      "[[7826 5873   78 1545    0    0    0    0    0    0    0    0    0    0\n",
      "     0    0    0    0    0    0]]\n",
      "['青岛', '天气', '很', '热']\n",
      "1/1 [==============================] - 0s 5ms/step\n",
      "ANSWER:  他 对 你 说 了 抱歉 好好 的 别 难受 那 就 喜欢 下去 吧 EOS\n",
      "你有男朋友么\n",
      "[[  10   23 1173  126    0    0    0    0    0    0    0    0    0    0\n",
      "     0    0    0    0    0    0]]\n",
      "['你', '有', '男朋友', '么']\n",
      "1/1 [==============================] - 0s 6ms/step\n",
      "ANSWER:  没有 啦 EOS\n",
      "我恨你\n",
      "[[  56 4127   10    0    0    0    0    0    0    0    0    0    0    0\n",
      "     0    0    0    0    0    0]]\n",
      "['我', '恨', '你']\n",
      "1/1 [==============================] - 0s 7ms/step\n",
      "ANSWER:  你 不 觉得 这个 话题 没什么 意思 么 EOS\n",
      "我爱你\n",
      "[[604   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "    0   0]]\n",
      "['我爱你']\n",
      "1/1 [==============================] - 0s 6ms/step\n",
      "ANSWER:  你 打算 向 我 求婚 吗 EOS\n",
      "我喜欢你\n",
      "[[56 38 10  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0]]\n",
      "['我', '喜欢', '你']\n",
      "1/1 [==============================] - 0s 6ms/step\n",
      "ANSWER:  他们 会 幸福 的 EOS\n",
      "你在干嘛呢\n",
      "[[ 10   2 229 171   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "    0   0]]\n",
      "['你', '在', '干嘛', '呢']\n",
      "1/1 [==============================] - 0s 7ms/step\n",
      "ANSWER:  能 干嘛 呢 聊天 呗 上网 呗 EOS\n",
      "x\n"
     ]
    }
   ],
   "source": [
    "while True:\n",
    "    seq = input()\n",
    "    if seq == 'x':\n",
    "        break\n",
    "    seq, sentence = input_question(seq)\n",
    "    print(sentence)\n",
    "    answer = decode_greedy(seq, sentence)\n",
    "#     answer=decode_beamsearch(seq, 3)\n",
    "    print('ANSWER: ', answer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Building prefix dict from the default dictionary ...\n",
      "Loading model from cache E:\\Temp\\jieba.cache\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "你饿不饿啊\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Loading model cost 0.892 seconds.\n",
      "Prefix dict has been built succesfully.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[ 10 730  36 730 128   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "    0   0]]\n",
      "\r",
      "1/1 [==============================] - 0s 156ms/step\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "那快 去 吃 点 东西 吧 别老 玩 QQ 了 EOS\n",
      "你饿不饿\n",
      "\n",
      "[[ 10 730  36 730   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "    0   0]]\n",
      "1/1 [==============================] - 0s 3ms/step\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "饿 EOS\n",
      "我生气了\n",
      "\n",
      "[[  56 2088   57    0    0    0    0    0    0    0    0    0    0    0\n",
      "     0    0    0    0    0    0]]\n",
      "1/1 [==============================] - 0s 3ms/step\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "淡定 淡定 ~ 消消气 EOS\n",
      "哄哄我\n",
      "\n",
      "[[11716    56     0     0     0     0     0     0     0     0     0     0\n",
      "      0     0     0     0     0     0     0     0]]\n",
      "1/1 [==============================] - 0s 2ms/step\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWQAAAJKCAYAAAAbVT/7AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvOIA7rQAAGVpJREFUeJzt3XmUpmV9p/HrSzdNE1YVFJAtSMC4sMkMGoxRCTJKIhomEk0w6kmI2yi4REeNjmZ0Ro0zkChgO3pc0MEVBlEy4K4YCD0CIhgMIquA7EJDQwO/+eN5myl6qrurq7vq/VVxfc6pU1XP+7x33wVVV9/cz/MWqSokSeO30bgnIEkaGGRJasIgS1ITBlmSmjDIktSEQZakJgyyVivJrh3HUm9J/mOSZ417HnORQdaafCTJ/0myU7Ox1NuuwKfGPYm5yCDrQUnenuQVSVZ+X5wD7A1knGNpzjkTuBEgyd5JTkvymiT+u18Lg6yJLgU+Duw5+vw+YGlVXTXmsTS3nLPyg6q6EDgf+DCw/9hmNEcY5IeRJAsmfPyEJJ9I8r4km40Onzd6/8vR+62Bc2d6LM0PSU5M8rfADcCmK49X1buAQ4ALxjW3uWLhuCegWfX6JMcA3wKuBZ4N7AL8DPgkww/SCuDXo4twTwIeSLIEuAV4e1XdPwNjaY5L8mbgBcAHRoeuSLJrVV0BUFVnjmtuc4lBfng5BfgQ8Iqquj/JzsDVwClJ/hOwB8P3xLLR26bAD4HbgMuABcD9MzCW5qgki4G3AIcDB1XVxaPj5wPPSLIbsA/w5aq6cnwznRsM8sNIVf0iyXJgEXA3sDnw4aq6Pcl3GPb6DgD2qao7kvwzcHRVXTKTY2nuqqrlwLuTXAmckOQw4HnAM4DXAqcD3wRuHd8s5w73kB9+bgAWJVkE7Mjww0JVfaeq/hdwW1XdMTp3R4bVLAAT7piYibE0RyV5NMNfzAcA/wQ8mWHVfCNwZFV9AnhRkh3GN8u5wRXyw88NwBeA5cC3gS2Bm+DBF29sOdoP3BnYBjg3yWNGHy9Icjrwghp+kfaGHEtzTJKPM1w7uA74OvCLqnr8hMd/ADwd+B7wVuCJwDFjmOqcEX8WHl6SfBl4bVVdl+TpDD8oOwCbAD8HLgeuBK4CrmcI7J0Me76LgOVVdc2GHktzT5JNga2q6vrRPcbXM1zUO7mqrk3yDOAo4LPAicB+VXXz+Gbcn0F+mEmyV1X9eMLnjwbuGe39/j3wo6r65OixrGn1uiHH0tyV5EDg1cBzgN0mbFOtvLi3DfDiqvrBmKY4ZxhkAQ+udq5k2H544uhC3JEM+4HvqKp7xzGW+hnt/+8F/CHwJ8AlwBLgY1W164TztmbYU94IeIK3Oa6dF1a00jEM2wlPm7DC+QLwp8DXk6zL9YYNOZb62Rr4BPBHwMuq6o8ZtqhWrDwhyW8BZwDHAbcDb5s4QJKtZm22c4hBFkn2A/4COKSqrl15vKruAd4HHMSwEprVsdRTVd0C7F9V+wI/Gh3ek+FFQAuS/CXwGuAlVXUiw7/v1yV5MUCSRwL/bgxTb8+VysNckt2B9wIHV9XPJzllCXA08JuzOZZ6q6oHRh8uHt1t8TzgS6NtiY+tcu7lSf49w38d7Q7cwfBiIa3CIGtv4PlVtWKyB6tqRZKXMbzCbjbH0hxQVcuSvBw4Czh2Ded9N8mhwFeARwDvmqUpzile1JM0a5LsAhwPXFNVfzXu+XRjkCWpCS/qSVITBlmSmjDIktSEQda0JDlq3HNQX35/TI9B1nT5A6c18ftjGgyyJDXhbW9TtCib1GI2W/uJDxMruIeN2WTc01BTfn881B3celNVbbu283yl3hQtZjMOyEHjnoakOegb9aUp/f8E3bKQpCYMsiQ1YZAlqQmDLElNGGRJasIgS1ITBlmSmjDIktSEQZakJgyyJDVhkCWpCYMsSU0YZElqwiBLUhMGWZKaMMiS1IRBlqQmDLIkNWGQJakJgyxJTRhkSWrCIEtSEwZZkpowyJLUhEGWpCYMsiQ1YZAlqQmDLElNGGRJasIgS1ITBlmSmjDIktSEQZakJgyyJDVhkCWpCYMsSU0YZElqwiBLUhMGWZKaMMiS1IRBlqQmDLIkNWGQJakJgyxJTRhkSWrCIEtSEwZZkpowyJLUhEGWpCYMsiQ1YZAlqQmDLElNGGRJasIgS1ITBlmSmjDIktSEQZakJgyyJDVhkCWpCYMsSU3MuyAn2SjJ9UnOS7JwdGzjJL9O8rZxz0+SVmfhuCcwAw4GHgk8q6ruGx07DLgR+NDYZiVJazHvVsjAW4Bjq+qnE469AXhjVd0DkOSpST6T5O+SPHIss5SkVcyrFXKSQ4A9gedPOHYE8ChgiyQfBrYDfgn8NvBnwAHA787+bCXpoeZNkJNsDpwALAD2AH40Wv2+HngPsBnw7qq6cXT+T4A3AR8Zz4wl6aHm05bF/wA+CpwN7DM69mbgpVX1WeAC4INJdkuyK/CGqvpQVS0fx2QlaVXzIshJ3glcUFXvB+4CrkqyMfBfq+qy0Wn/AryU4WLfFVV15hTGPSrJ0iRLV3DPjM1fkmAebFkk+V3g+1X17dGhjYG7q2oFcHuSxVW1vKp+neRm4PKpjl1VS4AlAFvmkbWh5y5JE835FXJVTYwxwG8A9wEk2Qv4kwmP3QncnWSXJIclWTSLU5WkNZrzQZ7E5sCK0ccfAG6Z8NgDDF/z9cDbgMtHd2ZI0tjNxyBvz7AKfgrwe8APJjx2P7Dx6H7k5zOspE9KMh//OUiaY+ZViJJsAewO3AHsCJxQVRNXyGFYQVNVNwDPA26d7XlK0mTm/EW9VezO8JfMr4DTgNOSLGCI8L0Mq+cH/xKqqksY7lmWpLGbVytk4FLgnKq6t0YY9o1fC9zG8OKQy9Y0gCSNy7xaIVfVXUkOWuVYAe9NEmCfVX7HhSS1Ma+CDLC6V95V1X+e7blI0rqYb1sWkjRnGWRJasIgS1ITBlmSmjDIktSEQZakJgyyJDVhkCWpCYMsSU0YZElqwiBLUhMGWZKaMMiS1IRBlqQmDLIkNWGQJakJgyxJTRhkSWrCIEtSEwZZkpowyJLUhEGWpCYMsiQ1YZAlqQmDLElNGGRJasIgS1ITBlmSmjDIktSEQZakJgyyJDVhkCWpCYMsSU0YZElqwiBLUhMGWZKaMMiS1IRBlqQmDLIkNWGQJakJgyxJTRhkSWrCIEtSEwZZkpowyJLUhEGWpCYMsiQ1YZAlqQmDLElNGGRJasIgS1ITBlmSmjDIktSEQZakJgyyJDVhkCWpCYMsSU0sHPcE5opssogFu+w27mmoqQeuunbcU1Bny6d2mitkSWrCIEtSEwZZkpowyJLUhEGWpCYMsiQ1YZAlqQmDLElNGGRJasIgS1ITBlmSmjDIktSEQZakJgyyJDVhkCWpCYMsSU0YZElqwiBLUhMGWZKaMMiS1IRBlqQmDLIkNWGQJakJgyxJTRhkSWrCIEtSEwZZkpowyJLUhEGWpCYMsiQ1YZAlqQmDLElNGGRJasIgS1ITBlmSmjDIktSEQZakJgyyJDVhkCWpCYMsSU0YZElqwiBLUhMGWZKaMMiS1IRBlqQmDLIkNWGQJakJgyxJTRhkSWrCIEtSEwZZkpowyJLUhEGWpCYMsiQ1YZAlqQmDLElNGGRJasIgS1ITBlmSmjDIktSEQZakJuZlkDO4NMmLkixKsm+SnZJ8JcnJ456fJE1mXga5qgo4DfgPwMnA0cBjgR2Az41xapK0WgvHPYEZ9C3gEVX1FysPJNkZ+Ob4piRJqzefg3w28J4kbwR+a3TsnKpaNsY5SdJqzdsgV9Wvk3yKYevi6qqqJE9K8lJgN2BpVZ0+3llK0v8zb4OcZHPgQuAI4KlJ9gbuAq4GbgU2TXJGVd0/xmlK0oPmRZCTPBp4BbANsAfDFsXtwD8BPwdeCGxXVb8a2yQlaS3mxV0Wo9BeCPwaOAE4A3h5VR0DnAgEWJbkcUmekSRTGTfJUUmWJll67/13z9T0JQmYJytkgKo6AzgjyeMY9o1vBt5bVfcluQ/438D2wAUMq+ZrpzDmEmAJwFaLt6uZmrskwTwKMgwvCAGOB84H9k+yZ1VdCtxWVU8f7+wkac3mVZCBtwG7AAcCi4BPJTkTWD7WWUnSFMyLPWSAJC8GXg4cUlU3V9V1wPOA7YDNk/xBknnz9Uqaf+bFCjnJy4CXAs+oql+uPF5V9wFvSnI2cBJwU5LPAmcB51XVPeOYryRNZs6vGJM8h+H+4oMmxniiqjoF2J/h/uN3At9nuOviZ0lOT/Le0X3LkjQ2c36FXFVnTvG8y5I8DXgl8ExgGXAJw73K57paljRucz7I62K0hfHh0ZsktTLntywkab4wyJLUhEGWpCYMsiQ1YZAlqQmDLElNGGRJasIgS1ITBlmSmjDIktSEQZakJgyyJDVhkCWpCYMsSU0YZElqwiBLUhMGWZKaMMiS1IRBlqQmDLIkNWGQJakJgyxJTRhkSWrCIEtSEwZZkpowyJLUhEGWpCYMsiQ1YZAlqQmDLElNGGRJasIgS1ITBlmSmjDIktSEQZakJgyyJDVhkCWpCYMsSU0YZElqwiBLUhMGWZKaMMiS1IRBlqQmDLIkNWGQJakJgyxJTRhkSWrCIEtSEwZZkpowyJLUhEGWpCYMsiQ1YZAlqQmDLElNGGRJasIgS1ITBlmSmjDIktTEwnFPYM4oyIr7xj0LNXXri/Yb9xTU2ac/O6XTXCFLUhMGWZKaMMiS1IRBlqQmDLIkNWGQJakJgyxJTRhkSWrCIEtSEwZZkpowyJLUhEGWpCYMsiQ1YZAlqQmDLElNGGRJasIgS1ITBlmSmjDIktSEQZakJgyyJDVhkCWpCYMsSU0YZElqwiBLUhMGWZKaMMiS1IRBlqQmDLIkNWGQJakJgyxJTRhkSWrCIEtSEwZZkpowyJLUhEGWpCYMsiQ1YZAlqQmDLElNGGRJasIgS1ITBlmSmjDIktSEQZakJgyyJDVhkCWpCYMsSU0YZElqwiBLUhMGWZKaMMiS1IRBlqQmDLIkNWGQJakJgyxJTRhkSWrCIEtSEwZZkpowyJLURNsgJ2k7N0maCesdvSRPTbLr+k/l//O+JBcledF0B0hyaJI/3JCTkqSZsiFWoXsDlyd52XQHSLJPknOSHDDh8H7AJsCt6zG3bYHTkrx8PcaQpFmxIYL8QyDAL6c7QFVdAJwGvAogye+Nxvy3VXXWesztztH7a9ZjDEmaFQs3wBg3jd7/bH0Gqar3Jdli9OlWwFuBlyS5uao+P81htwKuB765PnOTpNmwIYJ8/+j9zdN5cpIDgWcDjwB2SbLb6OOrgH8BLkqyUVU9MI3hHwd8bZrPlaRZtSGCnNH7Ox9yMElV1RSefz7wWIawnwJ8B3gi8CtgRVXdsR5z+23gM6P5HMiwL31CVd23HmNK0ozYEEH+DYaYHprkCcCTgacAuye5HDiwqla7eq6qu5LcBewM7AvcA3wDuAP4FvCa9ZjbvwFWXtBbBPw9cH2S+4BdGP4iWAF8pqp+uh5/jiStt3UKcpI9GfZ2NwcewxDRrYG7gWcCvwC+DBzPsHd7TVWtmMLQl47G+y7wCuDJVfWQuyuSPApYVlXLVzn+eIawAmwBbMNwd8WODFsfJyVZDOwE3Aa8ZHTu5qO5bwUcnuQpVbXqKv8o4CiAxQu2QJJm0joFuaouTfIRhlvdfjp6OxB4V1W9aT3mcSXDaniX0Zxek2QrhtDuDOwBbAlcmuSgqrppwnMXAO8AHgX8M/AThm2QjYG3V9WxAEneC2xbVUetw9e7BFgCsNUm201l+0WSpm2dtyyqaimwdOXno1Xz7kl2rqqrpjJGkqcCH2CI6ELgBoYLeNsD9wJnM9y9cTvD1sWdq1tpV9XFwLMm+TNeC7x6wqHnAO+fyvwkaRw2xB7y7wOfBE5IcthULphV1TlJXghsUlUP3r+c5GiG1fcFDNsPK1fIy5N8dR3vlrgceEeSTzFccNwD+No6PF+SZtV6BXm0r/ts4I3AHzGset8wxac/Cnh5ko1HH+/EcEFwK+AshoguZ9ifvovhNrjzpzq3qnrd6CXdxzKsjs9huIAnSS2t7wr5TcA5VXVJkisY9ngvraqPTuG5/wqcy7AK/i5wNUOI/xH4dlUdt55zo6quSPIzhtvongT8OMkrq+p76zu2JG1o037pdJInMdyS9tcw3L7GsDo+Pskxa3t+DU6tqo9U1VeBPYFbGFba+yZ553TnNmGObwcOZ9hjfiJwLfDtJFNdxUvSrJlWkJPsAHweeGNVnbvyeFV9keF3Uvy3JCcn2X6K420JfBB4y2if+CjgmUk+l2TzacxvmyRfYHghyP5VdU1V3Qg8F/gi8KEk+63ruJI0k9Y5yEmeBpwEvK6qPjbJKa9m+GU+RwD/muS4JLuvYbzHMGxVfK6qTgWoqnuBFzKsai9O8orRvcRrm9umSV4F/BfgH6rq8In3M48uOB4JnAnsOqUvWJJmybq+MGRvYAfg4Kq6f7Jzquq6JL8D/OXo3C2BtyX5QlX944SxFjL8drfDgXdX1ddXGef2JAcDZwAfBz6Y5GSGlfn3V/Oy7BXAR9d0N0ZVrUjyXMD7iiW1sq4vDLkQuHAK510NrG0PeAvg1Kr6hzWM86vR70j+A2A7hnuTi+Fl0PdMcv6UfkeFv2xIUkcb4j7kaRltJaz1l8+PInvqzM9IksbL/2+dJDVhkCWpCYMsSU0YZElqwiBLUhMGWZKaMMiS1IRBlqQmDLIkNWGQJakJgyxJTRhkSWrCIEtSEwZZkpowyJLUhEGWpCYMsiQ1YZAlqQmDLElNGGRJasIgS1ITBlmSmjDIktSEQZakJgyyJDVhkCWpCYMsSU0YZElqwiBLUhMGWZKaMMiS1IRBlqQmDLIkNWGQJakJgyxJTRhkSWrCIEtSEwZZkpowyJLUhEGWpCYMsiQ1YZAlqQmDLElNGGRJasIgS1ITBlmSmjDIktSEQZakJgyyJDVhkCWpCYMsSU0YZElqwiBLUhMGWZKaMMiS1IRBlqQmFo57AnPGA/dTdy4b9yzU1LnvP2HcU1BjCz49tfNcIUtSEwZZkpowyJLUhEGWpCYMsiQ1YZAlqQmDLElNGGRJasIgS1ITBlmSmjDIktSEQZakJgyyJDVhkCWpCYMsSU0YZElqwiBLUhMGWZKaMMiS1IRBlqQmDLIkNWGQJakJgyxJTRhkSWrCIEtSEwZZkpowyJLUhEGWpCYMsiQ1YZAlqQmDLElNGGRJasIgS1ITBlmSmjDIktSEQZakJgyyJDVhkCWpCYMsSU0YZElqwiBLUhMGWZKaMMiS1IRBlqQmDLIkNWGQJakJgyxJTRhkSWrCIEtSEwZZkpowyJLUhEGWpCYMsiQ1YZAlqQmDLElNGGRJasIgS1ITBlmSmjDIktSEQZakJgyyJDXRNshJrkhSq3k7ecJ5hyT5YZK7ktyS5H8m2WmS8Z6Z5DtJbh+9fTXJnrP7VUnS6i0c9wTW4ibgmEmOXwGQ5PnAKcA1wLuAxwCvB34nyb5VdcvovD8GPg98D3gr8FjgDcA3kjy+qpbN8NchSWvVPcjLquqkyR5IEuBYYBlwYFVdMzp+GXACQ8j/ZnT6e4DrgOdU1b2j8y4HPg78OXD8TH4RkjQVbbcspmAf4DeB01fGeOTTwArgBROO7QrctDLGI18HjgOun+F5StKUzOUg7zV6/5OJB6vqLuBKYM8kK/8LYCmwV5LXJdlodN71VXV0VX1l1mYsSWvQPcgbJdlmkrcFwKNH59w8yfNuBjYGth59/kqGlfBxwGVJ3pxk8xmfvSStg+5B3gm4cZK3JwOLR+fcO8nzVh7bFKCqLgaeALwP2Az4AHBFksPX9IcnOSrJ0iRL731g+Xp+KZK0Zt2DfANw8CRvlwErC7lokuetPHb3ygNVdWtVvR3YGXgVwwXNL43uwJhUVS2pqv2rav9FGy1e3WmStEF0v8tieVV9Y7IHktww+nDbSR7eluHC3m1JtgZ2BK6pqtuq6h7gxCRnA+cx3IHxxQ0/dUlaN91XyGty4ej9XhMPjvaGdwZ+WlX3AYcCFwF/OvG8qroIuJjhDgxJGrs5G+SquhD4OXBokh0nPPTnDCv/U0efXzR6f9jE5yfZHtiTIcqSNHbdtyw2S/Jnq3nsZIZX5Z0GfDfJsQyv1HsTw21v/x2gqn6c5PPAEUnOAr4CbAn8FcNFv7+ZZGxJmnXdg7wN8JnVPHZqVX0tyXMZXjb9foYLfacAf11Vt00490iGlfCRDLe+3QScDxxRVefN1OQlaV20DXJV7TrF884EzlzLOSuAvx29SVJLc3YPWZLmG4MsSU0YZElqwiBLUhMGWZKaMMiS1IRBlqQmDLIkNWGQJakJgyxJTRhkSWrCIEtSEwZZkpowyJLUhEGWpCYMsiQ1YZAlqQmDLElNGGRJasIgS1ITBlmSmjDIktSEQZakJgyyJDVhkCWpCYMsSU0YZElqwiBLUhMGWZKaMMiS1IRBlqQmDLIkNWGQJakJgyxJTRhkSWrCIEtSEwZZkpowyJLUhEGWpCYMsiQ1YZAlqQmDLElNGGRJasIgS1ITBlmSmjDIktSEQZakJgyyJDVhkCWpCYMsSU0YZElqwiBLUhMGWZKaMMiS1IRBlqQmDLIkNWGQJakJgyxJTRhkSWrCIEtSE6mqcc9hTkhyI3DluOfRyDbATeOehNry++Ohdqmqbdd2kkHWtCRZWlX7j3se6snvj+lxy0KSmjDIktSEQdZ0LRn3BNSa3x/T4B6yJDXhClmSmjDIktSEQZakJgyyJDVhkCWpif8LAJwlfk0HGTYAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "你 是 大美女 EOS\n"
     ]
    }
   ],
   "source": [
    "from tkinter import Text, Entry, Button, StringVar, Scrollbar, END\n",
    "from tkinter.messagebox import showinfo\n",
    "import tkinter as tk\n",
    "root = tk.Tk()\n",
    "root.title(\"聊天机器人-小意\")\n",
    "root.geometry(\"425x385\")\n",
    "root.resizable(0,0)\n",
    "root.attributes('-toolwindow', True,\n",
    "                '-alpha', 1,\n",
    "                '-fullscreen', False, \n",
    "                '-topmost', True)\n",
    "showinfo('Hello', '[小意]准备好了！')\n",
    "root.overrideredirect(True)\n",
    "def center_window(w, h):\n",
    "    ws = root.winfo_screenwidth()\n",
    "    hs = root.winfo_screenheight()\n",
    "    x = (ws/2) - (w/2)\n",
    "    y = (hs/2) - (h/2)\n",
    "    root.geometry('%dx%d+%d+%d' % (w, h, x, y))\n",
    "center_window(425, 356)\n",
    "text_recv = Text(root, width=61, height=20, bd=0, selectbackground='#00BFFF')\n",
    "text_recv.place(x=0, y=0)\n",
    "text_recv.tag_config('me_msg', \n",
    "                     foreground='#000000', \n",
    "#                      background='#00BFFF',\n",
    "                     justify='right', \n",
    "                     rmargin=20,\n",
    "                     offset=10,\n",
    "                     borderwidth=10)\n",
    "\n",
    "text_msg = Text(root, width=61, height=6, bd=0)\n",
    "text_msg.place(x=0, y=275)\n",
    "\n",
    "\n",
    "def show():\n",
    "    seq = text_msg.get(\"0.0\", \"end\")\n",
    "    msg_send = seq\n",
    "    print(msg_send)\n",
    "    text_recv.insert(END, msg_send, 'me_msg')\n",
    "    seq, sentence = input_question(seq)\n",
    "    answer = decode_greedy(seq, sentence)\n",
    "    print(answer)\n",
    "    answer = ''.join(answer.split(' ')[:-1])\n",
    "    msg_recv = '[小意]: '+ answer + '\\n'\n",
    "    text_recv.insert(END, msg_recv)\n",
    "    text_msg.delete(\"0.0\", \"end\")\n",
    "Button(root,text='发送信息',width=10,command=show, \n",
    "       bd=0, bg='#00BFFF', fg='#FFFFFF',\n",
    "       activeforeground='#1E90FF').place(x=340,y=330)\n",
    "Button(root,text='quit',width=10,command=root.destroy, \n",
    "       bd=0, bg='#00BFFF', fg='#FFFFFF',\n",
    "       activeforeground='#1E90FF').place(x=20,y=330)\n",
    "Button(root, text=\"清屏\",width=10,command = lambda : text_recv.delete(0.0, END),\n",
    "       bd=0, bg='#00BFFF', fg='#FFFFFF',\n",
    "       activeforeground='#1E90FF').place(x=240,y=330)\n",
    "root.mainloop()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}