Sound_recognition.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "import numpy as np\n",
    "import random\n",
    "import scipy.io.wavfile as wav\n",
    "import matplotlib.pyplot as plt\n",
    "from IPython.display import Markdown, display\n",
    "from IPython.display import Audio\n",
    "from itertools import islice\n",
    "from pathlib import Path\n",
    "import pandas as pd\n",
    "from scipy.fftpack import fft\n",
    "from scipy.signal import get_window\n",
    "import librosa\n",
    "import librosa.display"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.rcParams['figure.figsize'] = (12, 3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Задача 1: Построение модели распознавания речи."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Алгоритм распознавания речи состоит из следующих этапов:\n",
    "- Этап предварительной обработки, который преобразует необработанный звук в представление признаков, которое обычно используются для ASR (Automatic Speech Recognition);\n",
    "- Этап создания акустической модели, которая принимает аудио признаки в качестве входных данных и возвращает распределение вероятностей по всем потенциальным транскрипциям;\n",
    "- Этап принятия выходных данных из акустической модели и возвращения предсказанной транскрипции."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Вы можете загрузить свой аудиофайл и файлы с текстовым описанием\n",
    "\n",
    "AUDIO = '/Users/Protector/Sounds/8a7aa386717c.mp3'\n",
    "CSV = 'C:/Users/Protector/Sounds/example.csv'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Protector\\AppData\\Roaming\\Python\\Python37\\site-packages\\librosa\\core\\audio.py:146: UserWarning: PySoundFile failed. Trying audioread instead.\n",
      "  warnings.warn('PySoundFile failed. Trying audioread instead.')\n"
     ]
    }
   ],
   "source": [
    "# открываем аудиофайл\n",
    "clip, sample_rate = librosa.load(AUDIO, sr=None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Частота дискретизации   16000 Гц\n",
      "Длина аудиофайла   0.34 секунд\n"
     ]
    }
   ],
   "source": [
    "print('Частота дискретизации   {} Гц'.format(sample_rate))\n",
    "print('Длина аудиофайла   {:3.2f} секунд'.format(len(clip)/sample_rate))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Первый график ниже отображает весь сигнал полностью\n",
    "\n",
    "Второй график уже увеличивает какой-то частотный диапозон, где мы можем наблюдать какую амплитудуимеет каждый объект\n",
    "из массива clip (данный массив представлен в виде временненого ряда), который представлен в виде одномерного массива типа NumPy с плавающей точкой NumPy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "timesteps = np.arange(len(clip)) / sample_rate  # in seconds\n",
    "\n",
    "fig, ax = plt.subplots(2, figsize=(12, 5))\n",
    "fig.subplots_adjust(hspace=0.5)\n",
    "\n",
    "# изобразить всю запись\n",
    "ax[0].plot(timesteps, clip)\n",
    "ax[0].set_xlabel('Time (s)')\n",
    "ax[0].set_ylabel('Amplitude')\n",
    "ax[0].set_title('Raw Audio: {} samples'.format(len(clip)))\n",
    "\n",
    "\n",
    "n_fft = 100 # длина кадра\n",
    "start = 1100 # начать с какой-то части записи \n",
    "x = clip[start:start+n_fft]\n",
    "\n",
    "ax[0].axvline(start/sample_rate, c='r') \n",
    "ax[0].axvline((start+n_fft)/sample_rate, c='r')\n",
    " \n",
    "ax[1].plot(x)\n",
    "ax[1].set_xlabel('Samples')\n",
    "ax[1].set_ylabel('Amplitude')\n",
    "ax[1].set_title('Raw Audio: {} samples'.format(len(x)));"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Признаковое описание сигнала"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Существует два основных способа представления сигнала в виде признакового описания. Спектрограмма – изображение, показывающее зависимость спектральной плотности мощности сигнала от времени. MFCC (Мел-частотные кепстральные коэффициенты). Основная идея MFCC аналогична спектрограммам: в каждом временном окне MFCC выдает вектор признаков, который характеризует звук внутри окна. Стоит обратить внимание, что признаки, полученные с помощью MFCC, имеют намного меньшую размерность, чем признаки, полученные с помощью спектрограмм. Это поможет нашей модели не переобучиться на тренировочном датасете. \n",
    "\n",
    "Далле в своей работе я буду использовать MFCC"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Использование Short-Term Fourier Transform\n",
    "\n",
    "hop_length = 512\n",
    "stft = librosa.stft(clip, n_fft=n_fft, hop_length=hop_length)\n",
    "stft_magnitude, stft_phase = librosa.magphase(stft)\n",
    "stft_magnitude_db = librosa.amplitude_to_db(stft_magnitude, ref=np.max)\n",
    "\n",
    "plt.figure(figsize=(12, 6))\n",
    "librosa.display.specshow(stft_magnitude_db, x_axis='time', y_axis='linear', \n",
    "                         sr=sample_rate, hop_length=hop_length)\n",
    "\n",
    "title = 'n_fft={},  hop_length={},  time_steps={},  fft_bins={}  (2D resulting shape: {})'\n",
    "plt.title(title.format(n_fft, hop_length, \n",
    "                       stft_magnitude_db.shape[1], \n",
    "                       stft_magnitude_db.shape[0], \n",
    "                       stft_magnitude_db.shape));"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Использование mel-спектограмм (MFCC)\n",
    "n_mels = 13\n",
    "\n",
    "fmin = 20           \n",
    "fmax = 8000\n",
    "mel_spec = librosa.feature.mfcc(clip, n_fft=n_fft, hop_length=hop_length,\n",
    "                                          n_mels=n_mels, sr=sample_rate, power=1.0, \n",
    "                                          fmin=fmin, fmax=fmax)\n",
    "mel_spec_db = librosa.amplitude_to_db(mel_spec, ref=np.max)\n",
    "\n",
    "plt.figure(figsize=(12, 6))\n",
    "librosa.display.specshow(mel_spec_db, x_axis='time',  y_axis='mel', \n",
    "                         sr=sample_rate, hop_length=hop_length, \n",
    "                         fmin=fmin, fmax=fmax)\n",
    "\n",
    "title = 'n_mels=13, fmin=20, fmax=8000'\n",
    "plt.title('n_mels=13, fmin=20, fmax=8000');\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Теперь подготовим нашу акустическую модель"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import tensorflow.compat.v1 as tf \n",
    "import tensorflow.compat.v1.keras\n",
    "from keras.backend.tensorflow_backend import set_session\n",
    "from tensorflow.compat.v1.keras import backend as K\n",
    "from tensorflow.compat.v1.keras.models import Model\n",
    "from tensorflow.compat.v1.keras.layers import (BatchNormalization, Conv1D, Dense, Input, \n",
    "    TimeDistributed, Activation, Bidirectional, SimpleRNN, GRU, LSTM, Lambda)\n",
    "\n",
    "\n",
    "gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction = 0.4)\n",
    "session = tf.Session(config=tf.ConfigProto(gpu_options=gpu_options))\n",
    "set_session(session)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "import wave\n",
    "import _pickle as pickle"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Необходимо определить два словаря для преобразований между текстовыми и целочисленными последовательностями"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "char_map_str = \"\"\"\n",
    "<SPACE> 0\n",
    "а 1\n",
    "б 2\n",
    "в 3\n",
    "г 4\n",
    "д 5\n",
    "е 6\n",
    "ё 7\n",
    "ж 8\n",
    "з 9\n",
    "и 10\n",
    "й 11\n",
    "к 12\n",
    "л 13\n",
    "м 14\n",
    "н 15\n",
    "о 16\n",
    "п 17\n",
    "р 18\n",
    "с 19\n",
    "т 20\n",
    "у 21\n",
    "ф 22\n",
    "х 23\n",
    "ц 24\n",
    "ч 25\n",
    "ш 26\n",
    "щ 27\n",
    "ъ 28\n",
    "ы 29\n",
    "ь 30\n",
    "э 31\n",
    "ю 32\n",
    "я 33\n",
    "0 34\n",
    "1 35\n",
    "2 36\n",
    "3 37\n",
    "4 38\n",
    "5 39\n",
    "6 40\n",
    "7 41\n",
    "8 42\n",
    "9 43\n",
    "№ 44\n",
    "s 45\n",
    "a 46\n",
    "p 47\n",
    "e 48\n",
    "w 49\n",
    "o 50\n",
    "d 51\n",
    "v 52\n",
    "m 53\n",
    "- 54\n",
    "\n",
    "\"\"\"\n",
    "\n",
    "char_map = {}\n",
    "index_map = {}\n",
    "for line in char_map_str.strip().split('\\n'):\n",
    "    ch, index = line.split()\n",
    "    char_map[ch] = int(index)\n",
    "    index_map[int(index)+1] = ch\n",
    "index_map[1] = ' '"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "def cnn_output_length(input_length, filter_size, border_mode, stride,\n",
    "                       dilation=1):\n",
    "    if input_length is None:\n",
    "        return None\n",
    "    assert border_mode in {'same', 'valid'}\n",
    "    dilated_filter_size = filter_size + (filter_size - 1) * (dilation - 1)\n",
    "    if border_mode == 'same':\n",
    "        output_length = input_length\n",
    "    elif border_mode == 'valid':\n",
    "        output_length = input_length - dilated_filter_size + 1\n",
    "    return (output_length + stride - 1) // stride"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Выбор модели нейронной сети"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Далее необходимо было выбрать архитектуру нейронной сети. Очевидно, что для решения задачи распознавания речи нужно \n",
    "использовать RNN слои. На каждом временном шаге один из 33 букв русского алфавита и возможно пару английских букв. \n",
    "Также стоит учитывать пробел и дефис. Выход RNN слоя на каждом временном шаге представляет собой \n",
    "вектор вероятностей с каким-то количеством вхождений, которое можно задать. В данном векторе i-ая запись кодирует вероятность \n",
    "того, что i-ый символ произнес во временной последовательности (дополнительный пустой символ или пробел используется, \n",
    "чтобы дополнять тренировочные примеры в батчях, содержащих неравную длину)."
   ]
  },
  {
   "attachments": {
    "image.png": {
     "image/png": "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"
    }
   },
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![image.png](attachment:image.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Однако обычные RNN слои малоэффективные, поэтому лучше использовать LSTM или GRU слои. LSTM слои имеют более продвинутый \n",
    "и более сложный способ вычисления. Данный способ, помимо входных значений и предыдущего состояния сети, использует также \n",
    "фильтры (gates), которые определяют каким образом информация будет использоваться для вычисления как входных значений на \n",
    "текущем слое, так и значений скрытого слоя на следующем шаге. Модель GRU основана на тех же принципах, но использует меньше \n",
    "фильтров и операций для вычисления итогового значения "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "В своей работе и использую LSTM слои, так как они показывают лучшую точность и меньшее значение функции потерь."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Однако просто слоёв LSTM недостаточно. Точность модели все равно крайне низкая. Чтобы улучшить точность необходимо добавить слои TimeDistributed, необходимый для поиска более сложных паттернов в наборе данных, и  BatchNormalization, необходимый чтобы сократить время обучения. Далее стоит усложнить архитектуру, а именно добавить слой 1D convolution."
   ]
  },
  {
   "attachments": {
    "image.png": {
     "image/png": "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"
    }
   },
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![image.png](attachment:image.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Если говорить об итоговой архитектуре, которую я использовал то это двухслойная LSTM сеть плюс сверточный слой вначале.\n",
    "Причина, по которой данная архитектура показывает лучшие результаты это то, что  сверточный слой отлично выбирает важные \n",
    "признаки из звукового представления сигнала в виде MFCC, а затем уже передает эти признаки в LSTM сеть. \n",
    "Значения функции потерь составили примерно 150 – 200."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "input_data = Input(name='the_input', shape=(None, 13))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'Conv1D' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-1-4c22fac34e08>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[1;32m----> 1\u001b[1;33m conv_1d = Conv1D(200, 11, strides=2, padding='valid', activation='tanh',\n\u001b[0m\u001b[0;32m      2\u001b[0m                  name='conv1d', dilation_rate=1)(input_data)\n\u001b[0;32m      3\u001b[0m \u001b[0mbn_cnn\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mBatchNormalization\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mname\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;34m'bn_conv_1d'\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mconv_1d\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      4\u001b[0m \u001b[0msimp_rnn\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mLSTM\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;36m250\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mactivation\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;34m'tanh'\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mreturn_sequences\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;32mTrue\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mimplementation\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m2\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mname\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;34m'rnn_1'\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mdropout\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m0.9\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mbn_cnn\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      5\u001b[0m \u001b[0mbn_rnn\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mBatchNormalization\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mname\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;34m'bn_rnn_1'\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0msimp_rnn\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mNameError\u001b[0m: name 'Conv1D' is not defined"
     ]
    }
   ],
   "source": [
    "conv_1d = Conv1D(200, 11, strides=2, padding='valid', activation='tanh',\n",
    "                 name='conv1d', dilation_rate=1)(input_data)\n",
    "bn_cnn = BatchNormalization(name='bn_conv_1d')(conv_1d)\n",
    "simp_rnn = LSTM(250, activation='tanh', return_sequences=True, implementation=2, name='rnn_1', dropout=0.9)(bn_cnn)\n",
    "bn_rnn = BatchNormalization(name='bn_rnn_1')(simp_rnn)\n",
    "simp_rnn_fin = LSTM(250, activation='tanh', return_sequences=True, implementation=2, name='final_layer_of_rnn', dropout=0.9)(bn_rnn)\n",
    "bn_rnn_fin = BatchNormalization(name='bt_rnn_final')(simp_rnn_fin)\n",
    "time_dense = TimeDistributed(Dense(56))(bn_rnn_fin)\n",
    "y_pred = Activation('softmax', name='softmax')(time_dense)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = Model(inputs=input_data, outputs=y_pred)\n",
    "model.output_length = lambda x: cnn_output_length(x, 11, 'valid', 2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "the_input (InputLayer)       [(None, None, 13)]        0         \n",
      "_________________________________________________________________\n",
      "conv1d (Conv1D)              (None, None, 200)         28800     \n",
      "_________________________________________________________________\n",
      "bn_conv_1d (BatchNormalizati (None, None, 200)         800       \n",
      "_________________________________________________________________\n",
      "rnn_1 (LSTM)                 (None, None, 250)         451000    \n",
      "_________________________________________________________________\n",
      "bn_rnn_1 (BatchNormalization (None, None, 250)         1000      \n",
      "_________________________________________________________________\n",
      "final_layer_of_rnn (LSTM)    (None, None, 250)         501000    \n",
      "_________________________________________________________________\n",
      "bt_rnn_final (BatchNormaliza (None, None, 250)         1000      \n",
      "_________________________________________________________________\n",
      "time_distributed (TimeDistri (None, None, 56)          14056     \n",
      "_________________________________________________________________\n",
      "softmax (Activation)         (None, None, 56)          0         \n",
      "=================================================================\n",
      "Total params: 997,656\n",
      "Trainable params: 996,256\n",
      "Non-trainable params: 1,400\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "print(model.summary())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Теперь поговорим о настройке дополнительных параметрах в данной сети. К слоям LSTM я добавил dropout, равный 0.9. На самом деле сильно он не повлиял на точность, однако всё-таки улучшил её. Dropout - это популярная методика регуляризации с глубокими сетями, где некоторые случайные нейроны не принимают участия в обучение (отбрасываются). Также для улучшения точности модели можно использовать расширенные сверкти (dilated convolutions). Расширенная свертка - это свертка, в которой фильтр применяется к области, превышающей его длину, путем пропуска входных значений с определенным шагом. Это эквивалентно свертке с более крупным фильтром, полученным из исходного фильтра путем расширения его нулями, но значительно более эффективно. Расширенная свертка позволяет сети работать в более широком масштабе, чем при обычной свертке. Это похоже на объединение или пошаговые свертки, но здесь выход имеет тот же размер, что и вход. В своей работе я выставил значение равным единице. Количество эпох было выбрано равным 20. Количество выходных фильтров в CNN слое равно 400. Число, указывающее длину окна свертки равно 11. Размерность выходного пространства для LSTM слоев равно 250. Функцию активации, которую я использовал, была гиперболический тангенс. Её природа нелинейна, она хорошо подходит для комбинации слоёв, а диапазон значений функции - (-1, 1). Она была выбрана, так как в ходе обучения возникала проблема взрывающего градиента, когда градиент принимает слишком большие значения и становится nan. Для данной функции нет смысла беспокоиться, что она перегрузится от больших значений."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.compat.v1.keras.optimizers import SGD, RMSprop, Adam\n",
    "from tensorflow.compat.v1.keras.callbacks import ModelCheckpoint   \n",
    "import os\n",
    "from python_speech_features import mfcc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "def text_to_int_sequence(text):\n",
    "    # Преобразовать текст в целочисленную последовательность\n",
    "    int_sequence = []\n",
    "    for c in text:\n",
    "        if c == ' ':\n",
    "            ch = char_map['<SPACE>']\n",
    "        else:\n",
    "            ch = char_map[c]\n",
    "        int_sequence.append(ch)\n",
    "    return int_sequence\n",
    "\n",
    "def int_sequence_to_text(int_sequence):\n",
    "    # Преобразовать целочисленную последовательность в текст\n",
    "    text = []\n",
    "    for c in int_sequence:\n",
    "        ch = index_map[c]\n",
    "        text.append(ch)\n",
    "    return text"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [],
   "source": [
    "def calc_feat_dim(window, max_freq):\n",
    "    return int(0.001 * window * max_freq) + 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [],
   "source": [
    "# класс для работы с аудиофайлами\n",
    "\n",
    "class AudioReceive():\n",
    "    def __init__(self, step=10, window=10, max_freq=8000, mfcc_dim=13,\n",
    "                minibatch_size=20, desc_file=None, max_duration=20.0):\n",
    "        self.feat_dim = calc_feat_dim(window, max_freq)\n",
    "        self.mfcc_dim = mfcc_dim\n",
    "        self.feats_mean = np.zeros((self.feat_dim,))\n",
    "        self.feats_std = np.ones((self.feat_dim,))\n",
    "        self.rng = random.Random(123)\n",
    "        if desc_file is not None:\n",
    "            self.load_metadata_from_desc_file(desc_file)\n",
    "        self.step = step\n",
    "        self.window = window\n",
    "        #self.max_freq = max_freq\n",
    "        self.cur_train_index = 0\n",
    "        self.cur_valid_index = 0\n",
    "        self.cur_test_index = 0\n",
    "        self.max_duration=max_duration\n",
    "        self.minibatch_size = minibatch_size\n",
    "\n",
    "    def get_batch(self, partition):\n",
    "        if partition == 'train':\n",
    "            audio_paths = self.train_audio_paths\n",
    "            cur_index = self.cur_train_index\n",
    "            texts = self.train_texts\n",
    "        elif partition == 'valid':\n",
    "            audio_paths = self.valid_audio_paths\n",
    "            cur_index = self.cur_valid_index\n",
    "            texts = self.valid_texts\n",
    "        elif partition == 'test':\n",
    "            audio_paths = self.test_audio_paths\n",
    "            cur_index = self.test_valid_index\n",
    "            texts = self.test_texts\n",
    "        else:\n",
    "            raise Exception(\"Invalid partition. Must be train/validation\")\n",
    "\n",
    "        features = [self.normalize(self.featurize(a)) for a in audio_paths[cur_index:cur_index+self.minibatch_size]]\n",
    "\n",
    "        # рассчитываем необходимые размеры\n",
    "        max_length = max([features[i].shape[0] for i in range(0, self.minibatch_size)])\n",
    "        max_string_length = max([len(texts[cur_index+i]) for i in range(0, self.minibatch_size)])\n",
    "\n",
    "        # инициализируем массивы\n",
    "        X_data = np.zeros([self.minibatch_size, max_length, self.mfcc_dim])\n",
    "        labels = np.ones([self.minibatch_size, max_string_length]) * 55\n",
    "        input_length = np.zeros([self.minibatch_size, 1])\n",
    "        label_length = np.zeros([self.minibatch_size, 1])\n",
    "\n",
    "        for i in range(0, self.minibatch_size):\n",
    "            # считаем X_data & input_length\n",
    "            feat = features[i]\n",
    "            input_length[i] = feat.shape[0]\n",
    "            X_data[i, :feat.shape[0], :] = feat\n",
    "\n",
    "            # считаем labels & label_length\n",
    "            label = np.array(text_to_int_sequence(texts[cur_index+i])) \n",
    "            labels[i, :len(label)] = label\n",
    "            label_length[i] = len(label)\n",
    "\n",
    "        # возвращаем массивы\n",
    "        outputs = {'ctc': np.zeros([self.minibatch_size])}\n",
    "        inputs = {'the_input': X_data, \n",
    "                      'the_labels': labels, \n",
    "                      'input_length': input_length, \n",
    "                      'label_length': label_length \n",
    "                     }\n",
    "\n",
    "        return (inputs, outputs)  \n",
    "\n",
    "    def shuffle_data_by_partition(self, partition):\n",
    "        if partition == 'train':\n",
    "            self.train_audio_paths, self.train_durations, self.train_texts = shuffle_data(\n",
    "                self.train_audio_paths, self.train_durations, self.train_texts)\n",
    "        elif partition == 'valid':\n",
    "            self.valid_audio_paths, self.valid_durations, self.valid_texts = shuffle_data(\n",
    "                self.valid_audio_paths, self.valid_durations, self.valid_texts)\n",
    "        else:\n",
    "            raise Exception(\"Invalid partition. Must be train/validation\")\n",
    "            \n",
    "            \n",
    "    def next_train(self):\n",
    "        # Получаем батч тренировочных данных\n",
    "        while True:\n",
    "            ret = self.get_batch('train')\n",
    "            self.cur_train_index += self.minibatch_size\n",
    "            if self.cur_train_index >= len(self.train_texts) - self.minibatch_size:\n",
    "                self.cur_train_index = 0\n",
    "                #self.shuffle_data_by_partition('train')\n",
    "            yield ret \n",
    "            \n",
    "    def next_valid(self):\n",
    "        # Получаем батч валидационных данных\n",
    "        while True:\n",
    "            ret = self.get_batch('valid')\n",
    "            self.cur_valid_index += self.minibatch_size\n",
    "            if self.cur_valid_index >= len(self.valid_texts) - self.minibatch_size:\n",
    "                self.cur_valid_index = 0\n",
    "                #self.shuffle_data_by_partition('valid')\n",
    "            yield ret\n",
    "             \n",
    "    def next_test(self):\n",
    "       # Получаем батч тестовых данных\n",
    "        while True:\n",
    "            ret = self.get_batch('test')\n",
    "            self.cur_test_index += self.minibatch_size\n",
    "            if self.cur_test_index >= len(self.test_texts) - self.minibatch_size:\n",
    "                self.cur_test_index = 0\n",
    "            yield ret\n",
    "            \n",
    "            \n",
    "    def load_train_data(self, desc_file='train.json'):\n",
    "        self.load_metadata_from_desc_file(desc_file, 'train')\n",
    "        self.fit_train()\n",
    "\n",
    "    def load_validation_data(self, desc_file='valid.json'):\n",
    "        self.load_metadata_from_desc_file(desc_file, 'validation')\n",
    "\n",
    "    def load_test_data(self, desc_file='test.json'):\n",
    "        self.load_metadata_from_desc_file(desc_file, 'test')\n",
    "        \n",
    "    def load_metadata_from_desc_file(self, desc_file, partition):\n",
    "        audio_paths, durations, texts = [], [], []\n",
    "        with open(desc_file) as json_line_file:\n",
    "            for line_num, json_line in enumerate(json_line_file):\n",
    "                try:\n",
    "                    spec = json.loads(json_line)\n",
    "                    #if float(spec['duration']) > self.max_duration:\n",
    "                    #    continue\n",
    "                    audio_paths.append(spec['key'])\n",
    "                    durations.append(float(spec['duration']))\n",
    "                    texts.append(spec['text'])\n",
    "                except Exception as e:\n",
    "                    print('Error reading line #{}: {}'.format(line_num, json_line))\n",
    "\n",
    "        if partition == 'train':\n",
    "            self.train_audio_paths = audio_paths\n",
    "            self.train_durations = durations\n",
    "            self.train_texts = texts\n",
    "        elif partition == 'validation':\n",
    "            self.valid_audio_paths = audio_paths\n",
    "            self.valid_durations = durations\n",
    "            self.valid_texts = texts\n",
    "        elif partition == 'test':\n",
    "            self.test_audio_paths = audio_paths\n",
    "            self.test_durations = durations\n",
    "            self.test_texts = texts\n",
    "        else:\n",
    "            raise Exception(\"Invalid partition to load metadata. Must be train/validation/test\")\n",
    "            \n",
    "    def fit_train(self, k_samples=20):\n",
    "        k_samples = min(k_samples, len(self.train_audio_paths))\n",
    "        samples = self.rng.sample(self.train_audio_paths, k_samples)\n",
    "        feats = [self.featurize(s) for s in samples]\n",
    "        feats = np.vstack(feats)\n",
    "        self.feats_mean = np.mean(feats, axis=0)\n",
    "        self.feats_std = np.std(feats, axis=0)\n",
    "     \n",
    "    def featurize(self, audio_clip):\n",
    "        (rate, sig) = wav.read(audio_clip)\n",
    "        return mfcc(sig, rate, numcep=self.mfcc_dim, nfft = 1103)\n",
    "    \n",
    "    def normalize(self, feature, eps=1e-14):\n",
    "        return (feature - self.feats_mean) / (self.feats_std + eps)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "def shuffle_data(audio_paths, durations, texts):\n",
    "        \"\"\" Перемешать данные (вызывается после полного прохождения \n",
    "            training или validation данных в течении процесса обучения)\n",
    "        Params:\n",
    "            audio_paths (list): Пути к аудиофайлам\n",
    "            durations (list): Продолжительность высказываний для каждого аудиоклипа\n",
    "            texts (list): Предложения, произнесенные в каждом аудиофайле\n",
    "        \"\"\"\n",
    "        p = np.random.permutation(len(audio_paths))\n",
    "        audio_paths = [audio_paths[i] for i in p] \n",
    "        durations = [durations[i] for i in p] \n",
    "        texts = [texts[i] for i in p]\n",
    "        return audio_paths, durations, texts"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Немного нужно рассказать про функцию потерь, которую я использовал. Эта функция потерь называется CTC (Connectionist Temporal Classification). Важные преимущества данной функции потерь заключается в том, что происходит устранение необходимости определять неоднозначные границы меток (например, в речи или рукописном вводе) и есть возможность группировать прогнозы меток, если это оказывается полезным (например, если несколько меток обычно встречаются вместе)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Функция потерь CTC\n",
    "\n",
    "def ctc_lambda_func(args):\n",
    "    y_pred, labels, input_length, label_length = args\n",
    "    return K.ctc_batch_cost(labels, y_pred, input_length, label_length)\n",
    "\n",
    "def add_ctc_loss(input_to_softmax):\n",
    "    the_labels = Input(name='the_labels', shape=(None,), dtype='float32')\n",
    "    input_lengths = Input(name='input_length', shape=(1,), dtype='int64')\n",
    "    label_lengths = Input(name='label_length', shape=(1,), dtype='int64')\n",
    "    output_lengths = Lambda(input_to_softmax.output_length)(input_lengths)\n",
    "    # output_length = BatchNormalization()(input_lengths)\n",
    "    # CTC loss is implemented in a lambda layer\n",
    "    loss_out = Lambda(ctc_lambda_func, output_shape=(1,), name='ctc')(\n",
    "        [input_to_softmax.output, the_labels, output_lengths, label_lengths])\n",
    "    model = Model(\n",
    "        inputs=[input_to_softmax.input, the_labels, input_lengths, label_lengths], \n",
    "        outputs=loss_out)\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "def train_model(model, \n",
    "                pickle_path,\n",
    "                save_model_path,\n",
    "                train_json='train.json',\n",
    "                valid_json='valid.json',\n",
    "                minibatch_size=20,\n",
    "                mfcc_dim=13,\n",
    "                #optimizer=SGD(lr=0.03, decay=1e-6, momentum=0.9, nesterov=True, clipnorm=5),\n",
    "                optimizer=Adam(lr=0.03, beta_1=0.9, beta_2=0.999, amsgrad=True),\n",
    "                epochs=20,\n",
    "                verbose=1,\n",
    "                max_duration=20.0):\n",
    "    \n",
    "    audio_rec = AudioReceive(minibatch_size=minibatch_size, mfcc_dim=mfcc_dim, max_duration=max_duration)\n",
    "    audio_rec.load_train_data(train_json)\n",
    "    audio_rec.load_validation_data(valid_json)\n",
    "\n",
    "    #считаем шаги за эпоху\n",
    "    num_train_examples=len(audio_rec.train_audio_paths)\n",
    "    steps_per_epoch = num_train_examples//minibatch_size\n",
    "    num_valid_samples = len(audio_rec.valid_audio_paths) \n",
    "    validation_steps = num_valid_samples//minibatch_size\n",
    "    \n",
    "    model1 = add_ctc_loss(model)\n",
    "    model1.compile(loss={'ctc': lambda y_true, y_pred: y_pred}, optimizer=optimizer)\n",
    "    \n",
    "    if not os.path.exists('results'):\n",
    "        os.makedirs('results')\n",
    "        \n",
    "    checkpointer = ModelCheckpoint(filepath='results/'+save_model_path, verbose=0)\n",
    "    \n",
    "    #тренируем моедль\n",
    "    hist = model1.fit_generator(generator=audio_rec.next_train(), steps_per_epoch=steps_per_epoch,\n",
    "        epochs=epochs, validation_data=audio_rec.next_valid(), validation_steps=validation_steps,\n",
    "        callbacks=[checkpointer], verbose=verbose)\n",
    "    \n",
    "    # сохраняем потери модели\n",
    "    with open('results/'+pickle_path, 'wb') as f:\n",
    "        pickle.dump(hist.history, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "12/12 [==============================] - 256s 21s/step - loss: 374.5694 - val_loss: 357.6669\n",
      "Epoch 2/20\n",
      "12/12 [==============================] - 254s 21s/step - loss: 304.0705 - val_loss: 326.5615\n",
      "Epoch 3/20\n",
      "12/12 [==============================] - 245s 20s/step - loss: 274.4922 - val_loss: 271.9562\n",
      "Epoch 4/20\n",
      "12/12 [==============================] - 243s 20s/step - loss: 251.3850 - val_loss: 276.1869\n",
      "Epoch 5/20\n",
      "12/12 [==============================] - 248s 21s/step - loss: 247.2293 - val_loss: 266.6015\n",
      "Epoch 6/20\n",
      "12/12 [==============================] - 245s 20s/step - loss: 243.9274 - val_loss: 265.8563\n",
      "Epoch 7/20\n",
      "12/12 [==============================] - 250s 21s/step - loss: 242.3600 - val_loss: 265.2929\n",
      "Epoch 8/20\n",
      "12/12 [==============================] - 251s 21s/step - loss: 243.1519 - val_loss: 266.1812\n",
      "Epoch 9/20\n",
      "12/12 [==============================] - 252s 21s/step - loss: 242.7371 - val_loss: 266.9386\n",
      "Epoch 10/20\n",
      "12/12 [==============================] - 254s 21s/step - loss: 241.7019 - val_loss: 266.3876\n",
      "Epoch 11/20\n",
      "12/12 [==============================] - 254s 21s/step - loss: 241.2103 - val_loss: 265.4185\n",
      "Epoch 12/20\n",
      "12/12 [==============================] - 250s 21s/step - loss: 240.8432 - val_loss: 264.7383\n",
      "Epoch 13/20\n",
      "12/12 [==============================] - 255s 21s/step - loss: 240.1438 - val_loss: 264.7210\n",
      "Epoch 14/20\n",
      "12/12 [==============================] - 249s 21s/step - loss: 240.6398 - val_loss: 265.2297\n",
      "Epoch 15/20\n",
      "12/12 [==============================] - 249s 21s/step - loss: 239.9267 - val_loss: 264.3293\n",
      "Epoch 16/20\n",
      "12/12 [==============================] - 253s 21s/step - loss: 241.1893 - val_loss: 265.5235\n",
      "Epoch 17/20\n",
      "12/12 [==============================] - 250s 21s/step - loss: 240.3602 - val_loss: 265.3988\n",
      "Epoch 18/20\n",
      "12/12 [==============================] - 247s 21s/step - loss: 240.0344 - val_loss: 265.2034\n",
      "Epoch 19/20\n",
      "12/12 [==============================] - 250s 21s/step - loss: 239.9670 - val_loss: 265.2125\n",
      "Epoch 20/20\n",
      "12/12 [==============================] - 249s 21s/step - loss: 240.4588 - val_loss: 264.5522\n"
     ]
    },
    {
     "ename": "NameError",
     "evalue": "name 'pickle' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-113-4b3139a9abaa>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[0;32m      1\u001b[0m train_model(model=model, \n\u001b[0;32m      2\u001b[0m             \u001b[0mpickle_path\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;34m'model.pickle'\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m----> 3\u001b[1;33m             save_model_path='model.h5') \n\u001b[0m",
      "\u001b[1;32m<ipython-input-112-33903b1043fc>\u001b[0m in \u001b[0;36mtrain_model\u001b[1;34m(model, pickle_path, save_model_path, train_json, valid_json, minibatch_size, mfcc_dim, optimizer, epochs, verbose, max_duration)\u001b[0m\n\u001b[0;32m     37\u001b[0m     \u001b[1;31m# сохраняем потери модели\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     38\u001b[0m     \u001b[1;32mwith\u001b[0m \u001b[0mopen\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m'results/'\u001b[0m\u001b[1;33m+\u001b[0m\u001b[0mpickle_path\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;34m'wb'\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;32mas\u001b[0m \u001b[0mf\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 39\u001b[1;33m         \u001b[0mpickle\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mdump\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mhist\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mhistory\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mf\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[1;31mNameError\u001b[0m: name 'pickle' is not defined"
     ]
    }
   ],
   "source": [
    "# не обращайте внимание на ошибку\n",
    "# просто по окончанию тренировки забыл установить модуль Pickle, который создаст файл для хранения значения потерь\n",
    "\n",
    "train_model(model=model, \n",
    "            pickle_path='model.pickle', \n",
    "            save_model_path='model.h5') "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Данная ошибка пока считается более менее приемлемой так как изначально она достигала значений около 3000.\n",
    "#На выходе получались достаточно запутанные фразы или \"ничего\". Но мы планиурем продолжить обучать данную модель."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Использованная мной метрика для измерения качества была WER (Коэффициент ошибок в словах). Данная метрика является общей для измерения производительности системы распознавания речи или машинного перевода."
   ]
  },
  {
   "attachments": {
    "image.png": {
     "image/png": "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"
    }
   },
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![image.png](attachment:image.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "В данной формуле:\n",
    "- S – число замен;\n",
    "- D – число удалений;\n",
    "- I – число вставок;\n",
    "- N – общее число слов в оригинальной фразе;\n",
    "- C – количество правильных слов."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def wer(r, h):\n",
    "    \"\"\"\n",
    "    Расчет WER, используя расстояние Левенштейна.\n",
    "\n",
    "    Параметры\n",
    "    ----------\n",
    "    r : list\n",
    "    h : list\n",
    "\n",
    "    Returns\n",
    "    -------\n",
    "    int\n",
    "\n",
    "    \"\"\"\n",
    "    # инициализация\n",
    "    d = np.zeros((len(r)+1)*(len(h)+1), dtype=np.uint8)\n",
    "    d = d.reshape((len(r)+1, len(h)+1))\n",
    "    for i in range(len(r)+1):\n",
    "        for j in range(len(h)+1):\n",
    "            if i == 0:\n",
    "                d[0][j] = j\n",
    "            elif j == 0:\n",
    "                d[i][0] = i\n",
    "\n",
    "    # вычисление\n",
    "    for i in range(1, len(r)+1):\n",
    "        for j in range(1, len(h)+1):\n",
    "            if r[i-1] == h[j-1]:\n",
    "                d[i][j] = d[i-1][j-1]\n",
    "            else:\n",
    "                substitution = d[i-1][j-1] + 1\n",
    "                insertion    = d[i][j-1] + 1\n",
    "                deletion     = d[i-1][j] + 1\n",
    "                d[i][j] = min(substitution, insertion, deletion)\n",
    "\n",
    "    return d[len(r)][len(h)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "wer(\"Не удалось обновить акт\".split(), \"Не удалось обновить акт\".split())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Вот такие приблизительно фразы могли получаться в ходе тренировки (чаще даже хуже)\n",
    "# \"Ен тка обновить удалось что сделать нжн\"\n",
    "# Нужно будет больше тренировать модель и дорабатывать архитектуру.\n",
    "wer(\"нет роли мол\".split(), \"нет роли мол\".split())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "В моей работе точность по данной метрике вариьровалась от 5 до 15. Предсказанные транскрипции, которые возвращала моя модель были такими же как можете видеть выше справа (были варианты и хуже).\n",
    "Разумеется если дольше обучать модель и улучшать архитектуру, то можно прийти к лучшему результату."
   ]
  }
 ],
 "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": "3.7.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}