BERT Features with Keras.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n",
      "/home/akoksal/.local/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:526: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_qint8 = np.dtype([(\"qint8\", np.int8, 1)])\n",
      "/home/akoksal/.local/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:527: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_quint8 = np.dtype([(\"quint8\", np.uint8, 1)])\n",
      "/home/akoksal/.local/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:528: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_qint16 = np.dtype([(\"qint16\", np.int16, 1)])\n",
      "/home/akoksal/.local/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:529: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_quint16 = np.dtype([(\"quint16\", np.uint16, 1)])\n",
      "/home/akoksal/.local/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:530: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_qint32 = np.dtype([(\"qint32\", np.int32, 1)])\n",
      "/home/akoksal/.local/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:535: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  np_resource = np.dtype([(\"resource\", np.ubyte, 1)])\n",
      "/home/akoksal/.local/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:541: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_qint8 = np.dtype([(\"qint8\", np.int8, 1)])\n",
      "/home/akoksal/.local/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:542: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_quint8 = np.dtype([(\"quint8\", np.uint8, 1)])\n",
      "/home/akoksal/.local/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:543: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_qint16 = np.dtype([(\"qint16\", np.int16, 1)])\n",
      "/home/akoksal/.local/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:544: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_quint16 = np.dtype([(\"quint16\", np.uint16, 1)])\n",
      "/home/akoksal/.local/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:545: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_qint32 = np.dtype([(\"qint32\", np.int32, 1)])\n",
      "/home/akoksal/.local/lib/python3.6/site-packages/tensorboard/compat/tensorflow_stub/dtypes.py:550: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  np_resource = np.dtype([(\"resource\", np.ubyte, 1)])\n"
     ]
    }
   ],
   "source": [
    "import json\n",
    "import random\n",
    "import warnings\n",
    "from datetime import datetime\n",
    "\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import torch\n",
    "from keras.callbacks import EarlyStopping\n",
    "from keras.layers import Dense, Dropout\n",
    "from keras.models import Sequential\n",
    "from keras.optimizers import Adagrad\n",
    "from sklearn.decomposition import PCA\n",
    "from sklearn.metrics import classification_report, f1_score\n",
    "from sklearn.utils import class_weight\n",
    "from tqdm import tqdm\n",
    "from transformers import AutoModel, AutoTokenizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_path = 'Boun Twitter Data/train.json'\n",
    "val_path = 'Boun Twitter Data/validation.json'\n",
    "test_path = 'Boun Twitter Data/test.json'\n",
    "cappy_path = 'Other Data/cappy.json'\n",
    "bege_path = 'Other Data/berkcan.json'\n",
    "netflix_path = 'Other Data/netflix.json'\n",
    "device = 'cuda'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def filter(text):\n",
    "    final_text = ''\n",
    "    for word in text.split():\n",
    "        if word.startswith('@'):\n",
    "            continue\n",
    "        elif word[-3:] in ['com', 'org']:\n",
    "            continue\n",
    "        elif word.startswith('pic') or word.startswith('http') or word.startswith('www'):\n",
    "            continue\n",
    "        else:\n",
    "            final_text += word+' '\n",
    "    return final_text"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "tokenizer = AutoTokenizer.from_pretrained(\"dbmdz/bert-base-turkish-128k-uncased\")\n",
    "bert = AutoModel.from_pretrained(\"dbmdz/bert-base-turkish-128k-uncased\").to(device)\n",
    "\n",
    "def feature_extraction(text):\n",
    "    x = tokenizer.encode(filter(text))\n",
    "    with torch.no_grad():\n",
    "        x, _ = bert(torch.stack([torch.tensor(x)]).to(device))\n",
    "        return list(x[0][0].cpu().numpy())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open(train_path, 'r') as f:\n",
    "    train = json.load(f)\n",
    "with open(val_path, 'r') as f:\n",
    "    val = json.load(f)\n",
    "with open(test_path, 'r') as f:\n",
    "    test = json.load(f)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5733/5733 [00:39<00:00, 144.07it/s]\n",
      "100%|██████████| 639/639 [00:04<00:00, 144.77it/s]\n",
      "100%|██████████| 1592/1592 [00:11<00:00, 144.64it/s]\n"
     ]
    }
   ],
   "source": [
    "mapping = {'negative':0, 'neutral':1, 'positive':2}\n",
    "def data_prep(dataset):\n",
    "    X = []\n",
    "    y = []\n",
    "    for element in tqdm(dataset):\n",
    "        X.append(feature_extraction(element['sentence']))\n",
    "        y_val = np.zeros(3)\n",
    "        y_val[mapping[element['value']]] = 1\n",
    "        y.append(y_val)\n",
    "    return np.array(X), np.array(y)\n",
    "\n",
    "X_train, y_train = data_prep(train)\n",
    "X_val, y_val = data_prep(val)\n",
    "X_test, y_test = data_prep(test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Training a Model with Keras"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /home/akoksal/.local/lib/python3.6/site-packages/tensorflow/python/ops/resource_variable_ops.py:435: 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",
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_1 (Dense)              (None, 512)               393728    \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 128)               65664     \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 32)                4128      \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 32)                0         \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              (None, 3)                 99        \n",
      "=================================================================\n",
      "Total params: 463,619\n",
      "Trainable params: 463,619\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "class_counts = [0, 0, 0]\n",
    "for el in y_train:\n",
    "    class_counts[np.argmax(el)]+=1\n",
    "class_weights = {idx:sum(class_counts)/el for idx, el in enumerate(class_counts)}\n",
    "\n",
    "es = EarlyStopping(monitor='val_loss', mode='min', verbose=1, patience=50)\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Dense(512, activation='tanh', input_shape=(768,)))\n",
    "model.add(Dropout(0.5))\n",
    "model.add(Dense(128, activation='tanh'))\n",
    "model.add(Dropout(0.5))\n",
    "model.add(Dense(32, activation='tanh'))\n",
    "model.add(Dropout(0.5))\n",
    "model.add(Dense(3, activation='softmax'))\n",
    "\n",
    "model.summary()\n",
    "\n",
    "model.compile(loss='categorical_crossentropy',\n",
    "              optimizer=Adagrad(),\n",
    "              metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /home/akoksal/.local/lib/python3.6/site-packages/tensorflow/python/ops/math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use tf.cast instead.\n",
      "Train on 5733 samples, validate on 639 samples\n",
      "Epoch 1/500\n",
      "5733/5733 [==============================] - 0s 73us/step - loss: 4.1559 - accuracy: 0.3373 - val_loss: 1.0409 - val_accuracy: 0.5290\n",
      "Epoch 2/500\n",
      "5733/5733 [==============================] - 0s 41us/step - loss: 3.6342 - accuracy: 0.3937 - val_loss: 0.9153 - val_accuracy: 0.5540\n",
      "Epoch 3/500\n",
      "5733/5733 [==============================] - 0s 58us/step - loss: 3.3256 - accuracy: 0.4629 - val_loss: 0.9112 - val_accuracy: 0.6166\n",
      "Epoch 4/500\n",
      "5733/5733 [==============================] - 0s 86us/step - loss: 3.1395 - accuracy: 0.5022 - val_loss: 0.9958 - val_accuracy: 0.5008\n",
      "Epoch 5/500\n",
      "5733/5733 [==============================] - 1s 121us/step - loss: 3.0775 - accuracy: 0.5245 - val_loss: 1.0622 - val_accuracy: 0.5696\n",
      "Epoch 6/500\n",
      "5733/5733 [==============================] - 1s 121us/step - loss: 2.9340 - accuracy: 0.5538 - val_loss: 0.9417 - val_accuracy: 0.6056\n",
      "Epoch 7/500\n",
      "5733/5733 [==============================] - 0s 36us/step - loss: 2.8529 - accuracy: 0.5754 - val_loss: 0.9369 - val_accuracy: 0.6119\n",
      "Epoch 8/500\n",
      "5733/5733 [==============================] - 1s 99us/step - loss: 2.7958 - accuracy: 0.5924 - val_loss: 0.8897 - val_accuracy: 0.6275\n",
      "Epoch 9/500\n",
      "5733/5733 [==============================] - 1s 122us/step - loss: 2.7413 - accuracy: 0.6040 - val_loss: 0.9039 - val_accuracy: 0.6228\n",
      "Epoch 10/500\n",
      "5733/5733 [==============================] - 1s 117us/step - loss: 2.6895 - accuracy: 0.6156 - val_loss: 0.8862 - val_accuracy: 0.6322\n",
      "Epoch 11/500\n",
      "5733/5733 [==============================] - 0s 34us/step - loss: 2.6260 - accuracy: 0.6306 - val_loss: 0.8586 - val_accuracy: 0.6510\n",
      "Epoch 12/500\n",
      "5733/5733 [==============================] - 1s 111us/step - loss: 2.6063 - accuracy: 0.6347 - val_loss: 0.8779 - val_accuracy: 0.6510\n",
      "Epoch 13/500\n",
      "5733/5733 [==============================] - 1s 121us/step - loss: 2.5589 - accuracy: 0.6407 - val_loss: 0.8756 - val_accuracy: 0.6338\n",
      "Epoch 14/500\n",
      "5733/5733 [==============================] - 1s 98us/step - loss: 2.5251 - accuracy: 0.6494 - val_loss: 0.8788 - val_accuracy: 0.6385\n",
      "Epoch 15/500\n",
      "5733/5733 [==============================] - 0s 66us/step - loss: 2.4717 - accuracy: 0.6548 - val_loss: 0.8711 - val_accuracy: 0.6416\n",
      "Epoch 16/500\n",
      "5733/5733 [==============================] - 1s 122us/step - loss: 2.4613 - accuracy: 0.6625 - val_loss: 0.8646 - val_accuracy: 0.6541\n",
      "Epoch 17/500\n",
      "5733/5733 [==============================] - 1s 122us/step - loss: 2.4175 - accuracy: 0.6717 - val_loss: 0.8896 - val_accuracy: 0.6401\n",
      "Epoch 18/500\n",
      "5733/5733 [==============================] - 0s 57us/step - loss: 2.4205 - accuracy: 0.6682 - val_loss: 0.8939 - val_accuracy: 0.6401\n",
      "Epoch 19/500\n",
      "5733/5733 [==============================] - 0s 84us/step - loss: 2.4095 - accuracy: 0.6674 - val_loss: 0.8991 - val_accuracy: 0.6416\n",
      "Epoch 20/500\n",
      "5733/5733 [==============================] - 1s 122us/step - loss: 2.3869 - accuracy: 0.6808 - val_loss: 0.8867 - val_accuracy: 0.6369\n",
      "Epoch 21/500\n",
      "5733/5733 [==============================] - 1s 120us/step - loss: 2.3936 - accuracy: 0.6745 - val_loss: 0.9922 - val_accuracy: 0.5994\n",
      "Epoch 22/500\n",
      "5733/5733 [==============================] - 0s 37us/step - loss: 2.3934 - accuracy: 0.6702 - val_loss: 0.8823 - val_accuracy: 0.6385\n",
      "Epoch 23/500\n",
      "5733/5733 [==============================] - 1s 89us/step - loss: 2.3473 - accuracy: 0.6785 - val_loss: 0.8710 - val_accuracy: 0.6588\n",
      "Epoch 24/500\n",
      "5733/5733 [==============================] - 1s 121us/step - loss: 2.3367 - accuracy: 0.6900 - val_loss: 0.9288 - val_accuracy: 0.6213\n",
      "Epoch 25/500\n",
      "5733/5733 [==============================] - 1s 121us/step - loss: 2.3352 - accuracy: 0.6866 - val_loss: 0.8851 - val_accuracy: 0.6495\n",
      "Epoch 26/500\n",
      "5733/5733 [==============================] - 0s 39us/step - loss: 2.3056 - accuracy: 0.6874 - val_loss: 0.8646 - val_accuracy: 0.6604\n",
      "Epoch 27/500\n",
      "5733/5733 [==============================] - 1s 108us/step - loss: 2.2848 - accuracy: 0.6949 - val_loss: 0.8825 - val_accuracy: 0.6463\n",
      "Epoch 28/500\n",
      "5733/5733 [==============================] - 1s 121us/step - loss: 2.2773 - accuracy: 0.6932 - val_loss: 0.8703 - val_accuracy: 0.6667\n",
      "Epoch 29/500\n",
      "5733/5733 [==============================] - 1s 101us/step - loss: 2.3006 - accuracy: 0.6942 - val_loss: 0.8974 - val_accuracy: 0.6416\n",
      "Epoch 30/500\n",
      "5733/5733 [==============================] - 0s 64us/step - loss: 2.2592 - accuracy: 0.6911 - val_loss: 0.8985 - val_accuracy: 0.6448\n",
      "Epoch 31/500\n",
      "5733/5733 [==============================] - 1s 121us/step - loss: 2.2409 - accuracy: 0.6965 - val_loss: 0.8831 - val_accuracy: 0.6557\n",
      "Epoch 32/500\n",
      "5733/5733 [==============================] - 1s 121us/step - loss: 2.2444 - accuracy: 0.7026 - val_loss: 0.8785 - val_accuracy: 0.6588\n",
      "Epoch 33/500\n",
      "5733/5733 [==============================] - 0s 61us/step - loss: 2.2155 - accuracy: 0.7063 - val_loss: 0.8825 - val_accuracy: 0.6620\n",
      "Epoch 34/500\n",
      "5733/5733 [==============================] - 0s 74us/step - loss: 2.2354 - accuracy: 0.7045 - val_loss: 0.8856 - val_accuracy: 0.6573\n",
      "Epoch 35/500\n",
      "5733/5733 [==============================] - 1s 111us/step - loss: 2.2194 - accuracy: 0.7057 - val_loss: 0.8768 - val_accuracy: 0.6667\n",
      "Epoch 36/500\n",
      "5733/5733 [==============================] - 1s 112us/step - loss: 2.2320 - accuracy: 0.7071 - val_loss: 0.8711 - val_accuracy: 0.6635\n",
      "Epoch 37/500\n",
      "5733/5733 [==============================] - 0s 61us/step - loss: 2.1809 - accuracy: 0.7136 - val_loss: 0.8749 - val_accuracy: 0.6620\n",
      "Epoch 38/500\n",
      "5733/5733 [==============================] - 0s 57us/step - loss: 2.2107 - accuracy: 0.7068 - val_loss: 0.8856 - val_accuracy: 0.6620\n",
      "Epoch 39/500\n",
      "5733/5733 [==============================] - 1s 121us/step - loss: 2.1796 - accuracy: 0.7185 - val_loss: 0.8950 - val_accuracy: 0.6448\n",
      "Epoch 40/500\n",
      "5733/5733 [==============================] - 1s 121us/step - loss: 2.1813 - accuracy: 0.7143 - val_loss: 0.8897 - val_accuracy: 0.6573\n",
      "Epoch 41/500\n",
      "5733/5733 [==============================] - 0s 61us/step - loss: 2.1585 - accuracy: 0.7199 - val_loss: 0.8875 - val_accuracy: 0.6604\n",
      "Epoch 42/500\n",
      "5733/5733 [==============================] - 0s 85us/step - loss: 2.1400 - accuracy: 0.7190 - val_loss: 0.9132 - val_accuracy: 0.6401\n",
      "Epoch 43/500\n",
      "5733/5733 [==============================] - 1s 121us/step - loss: 2.1458 - accuracy: 0.7209 - val_loss: 0.8968 - val_accuracy: 0.6479\n",
      "Epoch 44/500\n",
      "5733/5733 [==============================] - 1s 122us/step - loss: 2.1339 - accuracy: 0.7190 - val_loss: 0.8763 - val_accuracy: 0.6620\n",
      "Epoch 45/500\n",
      "5733/5733 [==============================] - 0s 37us/step - loss: 2.1294 - accuracy: 0.7181 - val_loss: 0.8958 - val_accuracy: 0.6588\n",
      "Epoch 46/500\n",
      "5733/5733 [==============================] - 1s 107us/step - loss: 2.0995 - accuracy: 0.7242 - val_loss: 0.9053 - val_accuracy: 0.6495\n",
      "Epoch 47/500\n",
      "5733/5733 [==============================] - 1s 121us/step - loss: 2.1031 - accuracy: 0.7173 - val_loss: 0.8739 - val_accuracy: 0.6620\n",
      "Epoch 48/500\n",
      "5733/5733 [==============================] - 1s 112us/step - loss: 2.0951 - accuracy: 0.7221 - val_loss: 0.8995 - val_accuracy: 0.6541\n",
      "Epoch 49/500\n",
      "5733/5733 [==============================] - 0s 49us/step - loss: 2.0928 - accuracy: 0.7207 - val_loss: 0.8727 - val_accuracy: 0.6620\n",
      "Epoch 50/500\n",
      "5733/5733 [==============================] - 1s 121us/step - loss: 2.0876 - accuracy: 0.7265 - val_loss: 0.8870 - val_accuracy: 0.6557\n",
      "Epoch 51/500\n",
      "5733/5733 [==============================] - 1s 122us/step - loss: 2.0968 - accuracy: 0.7234 - val_loss: 0.8801 - val_accuracy: 0.6667\n",
      "Epoch 52/500\n",
      "5733/5733 [==============================] - 0s 76us/step - loss: 2.1211 - accuracy: 0.7193 - val_loss: 0.8957 - val_accuracy: 0.6526\n",
      "Epoch 53/500\n",
      "5733/5733 [==============================] - 0s 70us/step - loss: 2.0756 - accuracy: 0.7291 - val_loss: 0.8756 - val_accuracy: 0.6667\n",
      "Epoch 54/500\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5733/5733 [==============================] - 1s 113us/step - loss: 2.0729 - accuracy: 0.7268 - val_loss: 0.9299 - val_accuracy: 0.6322\n",
      "Epoch 55/500\n",
      "5733/5733 [==============================] - 1s 121us/step - loss: 2.0570 - accuracy: 0.7316 - val_loss: 0.8855 - val_accuracy: 0.6651\n",
      "Epoch 56/500\n",
      "5733/5733 [==============================] - 0s 58us/step - loss: 2.0233 - accuracy: 0.7328 - val_loss: 0.8755 - val_accuracy: 0.6588\n",
      "Epoch 57/500\n",
      "5733/5733 [==============================] - 0s 61us/step - loss: 2.0803 - accuracy: 0.7228 - val_loss: 0.8805 - val_accuracy: 0.6651\n",
      "Epoch 58/500\n",
      "5733/5733 [==============================] - 1s 121us/step - loss: 2.0341 - accuracy: 0.7319 - val_loss: 0.9308 - val_accuracy: 0.6369\n",
      "Epoch 59/500\n",
      "5733/5733 [==============================] - 1s 121us/step - loss: 2.0171 - accuracy: 0.7323 - val_loss: 0.8759 - val_accuracy: 0.6667\n",
      "Epoch 60/500\n",
      "5733/5733 [==============================] - 0s 56us/step - loss: 2.0372 - accuracy: 0.7324 - val_loss: 0.8936 - val_accuracy: 0.6604\n",
      "Epoch 61/500\n",
      "5733/5733 [==============================] - 1s 101us/step - loss: 2.0213 - accuracy: 0.7384 - val_loss: 0.8713 - val_accuracy: 0.6682\n",
      "Epoch 00061: early stopping\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(np.array(X_train), np.array(y_train),\n",
    "                    batch_size=64,\n",
    "                    epochs=500,\n",
    "                    verbose=1,\n",
    "                    validation_data=(X_val, y_val),\n",
    "                    class_weight=class_weights,\n",
    "                    callbacks = [es])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0      0.507     0.668     0.576       280\n",
      "           1      0.839     0.667     0.743       843\n",
      "           2      0.566     0.667     0.613       469\n",
      "\n",
      "    accuracy                          0.667      1592\n",
      "   macro avg      0.637     0.667     0.644      1592\n",
      "weighted avg      0.700     0.667     0.675      1592\n",
      "\n"
     ]
    }
   ],
   "source": [
    "y_true, y_pred = np.argmax(y_test, 1), np.argmax(model.predict(X_test), 1)\n",
    "print(classification_report(y_true, y_pred, digits=3))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Analysis"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Cappy Analysis"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "13717it [02:09, 105.96it/s]\n"
     ]
    }
   ],
   "source": [
    "cappy_df = pd.read_json(cappy_path).query('is_retweet==False').drop_duplicates(['text'])[['text', 'datetime']]\n",
    "cappy_df['value'] = 0\n",
    "#-1 negative 0 neutral 1 positive\n",
    "for idx, row in tqdm(cappy_df.iterrows()):\n",
    "    X = feature_extraction(row['text'])\n",
    "    cappy_df.at[idx, 'value'] = np.argmax(model.predict(np.array(X).reshape(1, -1)))-1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "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": [
    "incident1 = datetime.strptime('2020-01-28 15:20:00', '%Y-%m-%d %H:%M:%S')\n",
    "incident2 = datetime.strptime('2020-04-07 19:34:00', '%Y-%m-%d %H:%M:%S')\n",
    "dates = np.array(cappy_df['datetime'])\n",
    "indices = np.argsort(dates)[5300:]\n",
    "window = 750\n",
    "\n",
    "dates = dates[indices][window:]\n",
    "values = np.array(cappy_df['value'])[indices]\n",
    "windows = pd.Series(values).rolling(window)\n",
    "moving_averages = windows.mean()[window:]\n",
    "\n",
    "\n",
    "plt.figure(figsize=(12,6))\n",
    "plt.plot(dates, moving_averages, color='blue', label='Average Sentiment')\n",
    "\n",
    "plt.axvline(incident1, 0, 1, label='Date of the First Incident', color='red', alpha=0.5)\n",
    "plt.axvline(incident2, 0, 1, label='Date of the Second Incident', color='red', alpha=0.7)\n",
    "plt.title('Analysis of Turkish Tweets about Cappy')\n",
    "plt.xlabel('Date')\n",
    "plt.ylabel('Sentiment Score')\n",
    "plt.legend();"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Netflix Analysis"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "13948it [02:12, 105.28it/s]\n"
     ]
    }
   ],
   "source": [
    "netflix_df = pd.read_json(netflix_path).query('is_retweet==False').drop_duplicates(['text'])[['text', 'datetime']]\n",
    "netflix_df['value'] = 0\n",
    "#-1 negative 0 neutral 1 positive\n",
    "for idx, row in tqdm(netflix_df.iterrows()):\n",
    "    X = feature_extraction(row['text'])\n",
    "    netflix_df.at[idx, 'value'] = np.argmax(model.predict(np.array(X).reshape(1, -1)))-1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "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": [
    "incident1 = datetime.strptime('2020-04-08 13:48:00', '%Y-%m-%d %H:%M:%S')\n",
    "dates = np.array(netflix_df['datetime'])\n",
    "indices = np.argsort(dates)\n",
    "window = 750\n",
    "\n",
    "dates = dates[indices][window:]\n",
    "values = np.array(netflix_df['value'])[indices]\n",
    "windows = pd.Series(values).rolling(window)\n",
    "moving_averages = windows.mean()[window:]\n",
    "\n",
    "\n",
    "plt.figure(figsize=(12,6))\n",
    "plt.plot(dates, moving_averages, color='blue', label='Average Sentiment')\n",
    "\n",
    "plt.axvline(incident1, 0, 1, label='Date of the First Incident', color='red', alpha=0.5)\n",
    "plt.title('Analysis of Turkish Tweets about Netflix')\n",
    "plt.xlabel('Date')\n",
    "plt.ylabel('Sentiment Score')\n",
    "plt.legend();"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Berkcan Guven Analysis"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "14274it [02:12, 107.38it/s]\n"
     ]
    }
   ],
   "source": [
    "bege_df = pd.read_json(bege_path).query('is_retweet==False').drop_duplicates(['text'])[['text', 'datetime']]\n",
    "bege_df['value'] = 0\n",
    "#-1 negative 0 neutral 1 positive\n",
    "for idx, row in tqdm(bege_df.iterrows()):\n",
    "    X = feature_extraction(row['text'])\n",
    "    bege_df.at[idx, 'value'] = np.argmax(model.predict(np.array(X).reshape(1, -1)))-1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "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": [
    "incident1 = datetime.strptime('2020-03-07 23:11:00', '%Y-%m-%d %H:%M:%S')\n",
    "dates = np.array(bege_df['datetime'])\n",
    "indices = np.argsort(dates)\n",
    "window = 750\n",
    "\n",
    "dates = dates[indices][window:]\n",
    "values = np.array(bege_df['value'])[indices]\n",
    "windows = pd.Series(values).rolling(window)\n",
    "moving_averages = windows.mean()[window:]\n",
    "\n",
    "\n",
    "plt.figure(figsize=(12,6))\n",
    "plt.plot(dates, moving_averages, color='blue', label='Average Sentiment')\n",
    "\n",
    "plt.axvline(incident1, 0, 1, label='Date of the Video', color='red', alpha=0.5)\n",
    "plt.title('Analysis of Turkish Tweets about Berkcan Guven')\n",
    "plt.xlabel('Date')\n",
    "plt.ylabel('Sentiment Score')\n",
    "plt.legend();"
   ]
  },
  {
   "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": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}