Code.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "BertNoPrep.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "machine_shape": "hm"
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8JL4OAlSip-W",
        "colab_type": "text"
      },
      "source": [
        "# Connect to drive & import libraries"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "caEGNDMzfRAh",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        },
        "outputId": "426f6971-6162-4865-8649-08fc60b6d242"
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('drive')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Drive already mounted at drive; to attempt to forcibly remount, call drive.mount(\"drive\", force_remount=True).\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "tEzV-luAJRsZ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "%tensorflow_version 2.x"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iuOEBW9Liwfv",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from sklearn.model_selection import train_test_split\n",
        "import pandas as pd\n",
        "import tensorflow as tf\n",
        "import os\n",
        "from keras.engine import Layer\n",
        "import matplotlib.pyplot as plt\n",
        "import keras\n",
        "import numpy as np\n",
        "import nltk\n",
        "import tqdm\n",
        "import tensorflow_hub as hub\n",
        "import string\n",
        "from keras import backend as K"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BGuIwsXXJekG",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 37
        },
        "outputId": "f2913553-572b-440c-b831-97f9920d1b33"
      },
      "source": [
        "tf.__version__"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "string"
            },
            "text/plain": [
              "'2.3.0'"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 16
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ps76V6J_edIs",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from sklearn.metrics import accuracy_score\n",
        "from sklearn.metrics import classification_report\n",
        "from sklearn.metrics import precision_score\n",
        "from sklearn.metrics import recall_score\n",
        "from sklearn.metrics import f1_score\n",
        "from sklearn import metrics"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VNTXCjhyQFp7",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "54a87485-7c91-4c07-f81d-a344baaebc44"
      },
      "source": [
        "device_name = tf.test.gpu_device_name()\n",
        "if device_name != '/device:GPU:0':\n",
        "  raise SystemError('GPU device not found')\n",
        "print('Found GPU at: {}'.format(device_name))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found GPU at: /device:GPU:0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1sQU4wuWtkaf",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 102
        },
        "outputId": "163f881b-9457-4077-fa9e-904e1ddca09d"
      },
      "source": [
        "nltk.download('stopwords')\n",
        "nltk.download('punkt')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[nltk_data] Downloading package stopwords to /root/nltk_data...\n",
            "[nltk_data]   Package stopwords is already up-to-date!\n",
            "[nltk_data] Downloading package punkt to /root/nltk_data...\n",
            "[nltk_data]   Package punkt is already up-to-date!\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "True"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 19
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pqHDaPXXop-k",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 151
        },
        "outputId": "5854328d-a7a7-4f7b-f4bd-0b66b6021d30"
      },
      "source": [
        "!pip install bert-for-tf2"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Requirement already satisfied: bert-for-tf2 in /usr/local/lib/python3.6/dist-packages (0.14.5)\n",
            "Requirement already satisfied: params-flow>=0.8.0 in /usr/local/lib/python3.6/dist-packages (from bert-for-tf2) (0.8.2)\n",
            "Requirement already satisfied: py-params>=0.9.6 in /usr/local/lib/python3.6/dist-packages (from bert-for-tf2) (0.9.7)\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.6/dist-packages (from params-flow>=0.8.0->bert-for-tf2) (1.18.5)\n",
            "Requirement already satisfied: tqdm in /usr/local/lib/python3.6/dist-packages (from params-flow>=0.8.0->bert-for-tf2) (4.41.1)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vybgl8VvoqOG",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import bert"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4b2ADngPjIZo",
        "colab_type": "text"
      },
      "source": [
        "# Constants"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zxO3Fs12ivVm",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# data parameters\n",
        "test_data_percentage = 0.2\n",
        "# model parameters\n",
        "learning_rate = 0.0002\n",
        "batch_size = 32\n",
        "max_sequence_length = 128\n",
        "number_of_epochs = [1, 10]\n",
        "# model links\n",
        "bert_model_hub_link = 'https://tfhub.dev/tensorflow/bert_en_uncased_L-12_H-768_A-12/1'\n",
        "elmo_model_hub_link = 'https://tfhub.dev/google/elmo/3'"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8NwVmAJGi0vh",
        "colab_type": "text"
      },
      "source": [
        "# Reading & Pre-processing the data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fI056jLjj0oU",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 134
        },
        "outputId": "054ecd87-561f-4f70-a93c-1683d44f6ca6"
      },
      "source": [
        "metaphor_sentences, literal_sentences = [], []\n",
        "\n",
        "metaphor_words = []\n",
        "\n",
        "import pandas as pd\n",
        "\n",
        "file_1_df = pd.read_csv('drive/My Drive/Metaphor Detection/Data-metaphoric-or-literal.txt', sep='\\t')\n",
        "file_1_df.head(3)\n",
        "\n",
        "file_1_df['sentence'] = [str(x).replace('</b>', '').replace('<b>', '') for x in file_1_df['sentence'].values]\n",
        "file_1_df.head()\n",
        "\n",
        "with open('drive/My Drive/Metaphor Detection/TroFi/metaphorical.txt', 'r') as fp:\n",
        "  lines = fp.read().split('./.\\n')\n",
        "print(lines[0:2])\n",
        "\n",
        "metaphor_sentences.extend(lines)\n",
        "\n",
        "metaphor_sentences.extend(file_1_df[file_1_df['class']=='metaphorical'].sentence.values.tolist())\n",
        "\n",
        "with open('drive/My Drive/Metaphor Detection/TroFi/literal.txt', 'r') as fp:\n",
        "  lines = fp.read().split('./.\\n')\n",
        "print(lines[1])\n",
        "\n",
        "literal_sentences.extend(lines)\n",
        "\n",
        "literal_sentences.extend(file_1_df[file_1_df['class']=='literal'].sentence.values.tolist())\n",
        "\n",
        "X = literal_sentences + metaphor_sentences\n",
        "y = [0]*len(literal_sentences)+[1]*len(metaphor_sentences)\n",
        "print(len(X), len(y))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[\"But in the short-term it will absorb a lot of top management 's energy and attention , '' says Philippe Haspeslagh , a business professor at the European management school , Insead , in Paris \", 'Shocks from one-time changes in the terms of trade should be absorbed by adjustments in exchange-rate terms -- not price levels ']\n",
            "The yellow beta carotene pigment absorbs blue -LRB- not yellow -RRB- laser light \n",
            "5033 5033\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_spPPJ7oj05q",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "all_Y = y\n",
        "all_X = X"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "RQ_r5tFLnH7O",
        "colab_type": "text"
      },
      "source": [
        "## Create bert tokenizer"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VO_FMTtnnMCY",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "bert_layer = hub.KerasLayer(bert_model_hub_link)\n",
        "vocab_file = bert_layer.resolved_object.vocab_file.asset_path.numpy()\n",
        "do_lower_case = bert_layer.resolved_object.do_lower_case.numpy()\n",
        "bert_tokenizer_instance = bert.bert_tokenization.FullTokenizer(vocab_file=vocab_file, do_lower_case=do_lower_case)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kYuXzRGY1DpG",
        "colab_type": "text"
      },
      "source": [
        "### Bert functions for mask and id and segment id"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "W-UjqYR71I5I",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def convert_text_to_number_and_fix_length(this_x, tokenizer_mod):\n",
        "  this_x_ids = tokenizer_mod.convert_tokens_to_ids(this_x,)\n",
        "  if len(this_x_ids) >= max_sequence_length:\n",
        "    this_x_fix_len = this_x_ids[:max_sequence_length]\n",
        "  else:\n",
        "    this_x_fix_len = this_x_ids + [0]*(max_sequence_length - len(this_x_ids)) \n",
        "  return this_x_fix_len\n",
        "\n",
        "def create_mask_for_input(this_x):\n",
        "  if len(this_x) >= max_sequence_length:\n",
        "    this_x_masks = [1]*max_sequence_length\n",
        "  else:\n",
        "    this_x_masks = [1]*len(this_x) + [0] * (max_sequence_length - len(this_x))\n",
        "  return this_x_masks\n",
        "\n",
        "def create_req_bert_inputs(all_X, tokenizer_mod):\n",
        "  X_input_ids, X_input_masks, X_segment_ids = [], [], []\n",
        "\n",
        "  for this_x in tqdm.tqdm(all_X, position=0, leave=True):\n",
        "    X_input_ids.append(convert_text_to_number_and_fix_length(this_x, tokenizer_mod))\n",
        "    X_input_masks.append(create_mask_for_input(this_x))\n",
        "    X_segment_ids.append([0 for _ in range(max_sequence_length)])\n",
        "\n",
        "  return X_input_ids, X_input_masks, X_segment_ids"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-8T7Dfndvf0Z",
        "colab_type": "text"
      },
      "source": [
        "## Clean and tokenize "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Ruj1PKWZj005",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def clean_text(all_X, tokenizer_mod=None):\n",
        "  nltk_eng_stopwords = nltk.corpus.stopwords.words('english')\n",
        "  converted_X = []\n",
        "  for this_x in all_X:\n",
        "    this_x_tokenized = tokenizer_mod.tokenize(this_x)\n",
        "    converted_X.append(this_x_tokenized)\n",
        "  return converted_X\n",
        "\n",
        "def convert_y_to_0_and_1(all_Y):\n",
        "  label_number_mapping = {0: [1, 0], 1: [0, 1]}\n",
        "  converted_y = [label_number_mapping[val] for val in all_Y]\n",
        "  return converted_y"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Uqp90zJbtdG7",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "all_X_cleaned_using_bert = clean_text(all_X, bert_tokenizer_instance)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "77rTdgAdvnaU",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "all_y_converted = convert_y_to_0_and_1(all_Y)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "HgQX9l3xxzUX",
        "colab_type": "text"
      },
      "source": [
        "### Testing BERT results"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eMAxUp7Cthrz",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 111
        },
        "outputId": "dc7057f7-6ad4-457c-d133-dd4c28a23e1c"
      },
      "source": [
        "print('X[0] => before: {} and after: {}'.format(all_X[0], all_X_cleaned_using_bert[0]))\n",
        "print('Y[0] => before: {} and after: {}'.format(all_Y[0], all_y_converted[0]))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "X[0] => before: An Energy Department spokesman says the sulfur dioxide might be simultaneously recoverable through the use of powdered limestone , which tends to absorb the sulfur  and after: ['an', 'energy', 'department', 'spokesman', 'says', 'the', 'sulfur', 'dioxide', 'might', 'be', 'simultaneously', 'recover', '##able', 'through', 'the', 'use', 'of', 'powder', '##ed', 'limestone', ',', 'which', 'tends', 'to', 'absorb', 'the', 'sulfur']\n",
            "Y[0] => before: 0 and after: [1, 0]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Z8BQlRI2zL2j",
        "colab_type": "text"
      },
      "source": [
        "## Test Train split"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sk2kYytNzLdn",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "X_train, X_test, y_train, y_test = train_test_split(all_X_cleaned_using_bert, all_y_converted, test_size=test_data_percentage, random_state=42)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MztG_Ei6x3ul",
        "colab_type": "text"
      },
      "source": [
        "# Creating Model"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CX9WnCmUx58B",
        "colab_type": "text"
      },
      "source": [
        "## Using Bert"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "680fZdfa2qun",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        },
        "outputId": "995a0806-5e75-4584-eb31-77b83ca4a540"
      },
      "source": [
        "train_input_ids, train_input_masks, train_segment_ids = create_req_bert_inputs(X_train, bert_tokenizer_instance)\n",
        "test_input_ids, test_input_masks, test_segment_ids = create_req_bert_inputs(X_test, bert_tokenizer_instance)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "100%|██████████| 4026/4026 [00:00<00:00, 55464.28it/s]\n",
            "100%|██████████| 1007/1007 [00:00<00:00, 43244.23it/s]\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "thrzRZ04R5Vt",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 136
        },
        "outputId": "1fbf8be4-4d90-4491-b348-c14fa0751892"
      },
      "source": [
        "type(train_input_masks)\n",
        "np.asarray(train_input_masks, dtype=np.int32)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[1, 1, 1, ..., 0, 0, 0],\n",
              "       [1, 1, 1, ..., 0, 0, 0],\n",
              "       [1, 1, 1, ..., 0, 0, 0],\n",
              "       ...,\n",
              "       [1, 1, 1, ..., 0, 0, 0],\n",
              "       [1, 1, 1, ..., 0, 0, 0],\n",
              "       [1, 1, 1, ..., 0, 0, 0]], dtype=int32)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 42
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fRJQNWXDKWyw",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "0994376e-34cc-425e-fe5a-fda62284ef7e"
      },
      "source": [
        "bert_model_preped_input = [np.asarray(train_input_ids, dtype=np.int32), np.asarray(train_input_masks, dtype=np.int32), np.asarray(train_segment_ids, dtype=np.int32)]\n",
        "len(bert_model_preped_input), bert_model_preped_input[0].shape, bert_model_preped_input[0][1].shape"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(3, (4026, 128), (128,))"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 43
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_xNby7WbLSw5",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 323
        },
        "outputId": "34b1ec9e-fddf-468e-b6d8-b28fe181cc48"
      },
      "source": [
        "y_train = np.asarray(y_train)\n",
        "type(bert_model_preped_input), type(bert_model_preped_input[0]), type(y_train), bert_model_preped_input[0][0]"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(list,\n",
              " numpy.ndarray,\n",
              " numpy.ndarray,\n",
              " array([21497,  4187, 12629,  2000,  3071,  2556,     0,     0,     0,\n",
              "            0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
              "            0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
              "            0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
              "            0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
              "            0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
              "            0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
              "            0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
              "            0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
              "            0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
              "            0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
              "            0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
              "            0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
              "            0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
              "            0,     0], dtype=int32))"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 44
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iREClz0e25nV",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "this_optimizer = tf.keras.optimizers.Adam(\n",
        "    learning_rate=learning_rate, beta_1=0.9, beta_2=0.999, epsilon=1e-07, amsgrad=False,\n",
        "    name='Adam')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HpHXcNdNvUvm",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def create_model():\n",
        "  input_word_ids = tf.keras.layers.Input(shape=(max_sequence_length,), dtype=tf.int32,name=\"input_word_ids\")\n",
        "  input_mask = tf.keras.layers.Input(shape=(max_sequence_length,), dtype=tf.int32,name=\"input_mask\")\n",
        "  segment_ids = tf.keras.layers.Input(shape=(max_sequence_length,), dtype=tf.int32,name=\"segment_ids\")\n",
        "  bert_layer = hub.KerasLayer(bert_model_hub_link,trainable=False)\n",
        "  pooled_output, sequence_output = bert_layer([input_word_ids, input_mask, segment_ids])\n",
        "  pooling_lay = tf.keras.layers.GlobalAveragePooling1D()(sequence_output)\n",
        "  dense = tf.keras.layers.Dense(768, activation='relu')(pooling_lay)\n",
        "  pred = tf.keras.layers.Dense(2, activation='softmax')(dense)\n",
        "  model = tf.keras.models.Model(inputs=[input_word_ids, input_mask, segment_ids], outputs=pred)\n",
        "  return model "
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CQHP6X_h4Pnh",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def plot_accuracy_and_loss(file_save_name, history):\n",
        "  plt.plot(history.history['accuracy'])\n",
        "  plt.title('model accuracy')\n",
        "  plt.ylabel('accuracy')\n",
        "  plt.xlabel('epoch')\n",
        "  plt.show()\n",
        "  plt.cla()\n",
        "  plt.plot(history.history['loss'])\n",
        "  plt.title('model loss')\n",
        "  plt.ylabel('loss')\n",
        "  plt.xlabel('epoch')\n",
        "  plt.show()\n",
        "  plt.cla()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rGC5twGSdw4D",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def evaluate_model(model):\n",
        "  bert_model_test_input = [np.asarray(test_input_ids, dtype=np.int32), np.asarray(test_input_masks, dtype=np.int32), np.asarray(test_segment_ids, dtype=np.int32)]\n",
        "  predicted_y = model.predict(bert_model_test_input)\n",
        "  number_label_mapping = {0: 'ham', 1:'spam'}\n",
        "  y_test_conv = [number_label_mapping[np.argmax(val)] for val in y_test]\n",
        "  predicted_y_conv = [number_label_mapping[np.argmax(val)] for val in predicted_y]\n",
        "  print('Accuracy:', accuracy_score(y_test_conv, predicted_y_conv))\n",
        "  print('Classification report:', classification_report(y_test_conv, predicted_y_conv))\n",
        "  print('F1 (Micro):', f1_score(y_test_conv, predicted_y_conv, average='micro'))\n",
        "  print('F1 (Macro):', f1_score(y_test_conv, predicted_y_conv, average='macro'))\n",
        "  print('F1 (Weighted):', f1_score(y_test_conv, predicted_y_conv, average='weighted'))\n",
        "  print('Recall (Micro):', recall_score(y_test_conv, predicted_y_conv, average='micro'))\n",
        "  print('Recall (Macro):', recall_score(y_test_conv, predicted_y_conv, average='macro'))\n",
        "  print('Recall (Weighted):', recall_score(y_test_conv, predicted_y_conv, average='weighted'))\n",
        "  print('Precision (Micro):', precision_score(y_test_conv, predicted_y_conv, average='micro'))\n",
        "  print('Precision (Macro):', precision_score(y_test_conv, predicted_y_conv, average='macro'))\n",
        "  print('Precision (Weighted):', precision_score(y_test_conv, predicted_y_conv, average='weighted'))\n",
        "  y_test_for_auc = [np.argmax(val) for val in y_test]\n",
        "  y_predicted_for_auc = [np.argmax(val) for val in predicted_y]\n",
        "  fpr, tpr, thresholds = metrics.roc_curve(y_test_for_auc, y_predicted_for_auc, pos_label=1)\n",
        "  print('AUC:', metrics.auc(fpr, tpr))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eqAW2g-ZCbGi",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "b93a66f5-78a4-4b36-b05e-529570a73f4f"
      },
      "source": [
        "val_counts_in_train_y = {}\n",
        "for val in y_train:\n",
        "  if not str(val) in val_counts_in_train_y:\n",
        "    val_counts_in_train_y[str(val)] = 0\n",
        "  val_counts_in_train_y[str(val)] += 1\n",
        "val_counts_in_train_y"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "{'[0 1]': 1898, '[1 0]': 2128}"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 49
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "oPQQaGlFoBda",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "09573709-ef88-4cb7-d207-6b46551faeee"
      },
      "source": [
        "val_counts_in_test_y = {}\n",
        "for val in y_test:\n",
        "  if not str(val) in val_counts_in_test_y:\n",
        "    val_counts_in_test_y[str(val)] = 0\n",
        "  val_counts_in_test_y[str(val)] += 1\n",
        "val_counts_in_test_y"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "{'[0, 1]': 447, '[1, 0]': 560}"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 50
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "mUksObv2zJZ3",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "e0229074-ae82-4192-9dc7-7799a80307e2"
      },
      "source": [
        "for epochs in [10, 20, 100]:\n",
        "  print('Number of epochs:', epochs)\n",
        "  model = create_model()\n",
        "  print(model.summary())\n",
        "  model.compile(loss='binary_crossentropy', optimizer=this_optimizer, metrics=['accuracy'])\n",
        "  history = model.fit(\n",
        "      bert_model_preped_input, \n",
        "      y_train,\n",
        "      epochs=epochs,\n",
        "      batch_size=batch_size\n",
        "  )\n",
        "  print('Done training')\n",
        "  image_save_file_name = 'drive/My Drive/CA5_Sabri_810198312/Results/Q1_bert_model‌_no_preprocess_{}_epochs'.format(epochs)\n",
        "  plot_accuracy_and_loss(image_save_file_name, history)\n",
        "  evaluate_model(model)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Number of epochs: 10\n",
            "Model: \"functional_7\"\n",
            "__________________________________________________________________________________________________\n",
            "Layer (type)                    Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            "input_word_ids (InputLayer)     [(None, 128)]        0                                            \n",
            "__________________________________________________________________________________________________\n",
            "input_mask (InputLayer)         [(None, 128)]        0                                            \n",
            "__________________________________________________________________________________________________\n",
            "segment_ids (InputLayer)        [(None, 128)]        0                                            \n",
            "__________________________________________________________________________________________________\n",
            "keras_layer_4 (KerasLayer)      [(None, 768), (None, 109482241   input_word_ids[0][0]             \n",
            "                                                                 input_mask[0][0]                 \n",
            "                                                                 segment_ids[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "global_average_pooling1d_3 (Glo (None, 768)          0           keras_layer_4[0][1]              \n",
            "__________________________________________________________________________________________________\n",
            "dense_6 (Dense)                 (None, 768)          590592      global_average_pooling1d_3[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "dense_7 (Dense)                 (None, 2)            1538        dense_6[0][0]                    \n",
            "==================================================================================================\n",
            "Total params: 110,074,371\n",
            "Trainable params: 592,130\n",
            "Non-trainable params: 109,482,241\n",
            "__________________________________________________________________________________________________\n",
            "None\n",
            "Epoch 1/10\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.5774 - accuracy: 0.7144\n",
            "Epoch 2/10\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.4706 - accuracy: 0.7819\n",
            "Epoch 3/10\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.4267 - accuracy: 0.8120\n",
            "Epoch 4/10\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.3775 - accuracy: 0.8390\n",
            "Epoch 5/10\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.3304 - accuracy: 0.8661\n",
            "Epoch 6/10\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.2877 - accuracy: 0.8920\n",
            "Epoch 7/10\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.2518 - accuracy: 0.9066\n",
            "Epoch 8/10\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.2118 - accuracy: 0.9391\n",
            "Epoch 9/10\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.1797 - accuracy: 0.9493\n",
            "Epoch 10/10\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.1534 - accuracy: 0.9630\n",
            "Done training\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Accuracy: 0.7686196623634558\n",
            "Classification report:               precision    recall  f1-score   support\n",
            "\n",
            "         ham       0.79      0.79      0.79       560\n",
            "        spam       0.74      0.74      0.74       447\n",
            "\n",
            "    accuracy                           0.77      1007\n",
            "   macro avg       0.77      0.77      0.77      1007\n",
            "weighted avg       0.77      0.77      0.77      1007\n",
            "\n",
            "F1 (Micro): 0.7686196623634558\n",
            "F1 (Macro): 0.7655080537241415\n",
            "F1 (Weighted): 0.7685391897262321\n",
            "Recall (Micro): 0.7686196623634558\n",
            "Recall (Macro): 0.7653303771172899\n",
            "Recall (Weighted): 0.7686196623634558\n",
            "Precision (Micro): 0.7686196623634558\n",
            "Precision (Macro): 0.7656997583729377\n",
            "Precision (Weighted): 0.7684724403135137\n",
            "AUC: 0.7653303771172899\n",
            "Number of epochs: 20\n",
            "Model: \"functional_9\"\n",
            "__________________________________________________________________________________________________\n",
            "Layer (type)                    Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            "input_word_ids (InputLayer)     [(None, 128)]        0                                            \n",
            "__________________________________________________________________________________________________\n",
            "input_mask (InputLayer)         [(None, 128)]        0                                            \n",
            "__________________________________________________________________________________________________\n",
            "segment_ids (InputLayer)        [(None, 128)]        0                                            \n",
            "__________________________________________________________________________________________________\n",
            "keras_layer_5 (KerasLayer)      [(None, 768), (None, 109482241   input_word_ids[0][0]             \n",
            "                                                                 input_mask[0][0]                 \n",
            "                                                                 segment_ids[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "global_average_pooling1d_4 (Glo (None, 768)          0           keras_layer_5[0][1]              \n",
            "__________________________________________________________________________________________________\n",
            "dense_8 (Dense)                 (None, 768)          590592      global_average_pooling1d_4[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "dense_9 (Dense)                 (None, 2)            1538        dense_8[0][0]                    \n",
            "==================================================================================================\n",
            "Total params: 110,074,371\n",
            "Trainable params: 592,130\n",
            "Non-trainable params: 109,482,241\n",
            "__________________________________________________________________________________________________\n",
            "None\n",
            "Epoch 1/20\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.5809 - accuracy: 0.7064\n",
            "Epoch 2/20\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.4691 - accuracy: 0.7899\n",
            "Epoch 3/20\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.4180 - accuracy: 0.8095\n",
            "Epoch 4/20\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.3705 - accuracy: 0.8390\n",
            "Epoch 5/20\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.3284 - accuracy: 0.8631\n",
            "Epoch 6/20\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.2802 - accuracy: 0.8942\n",
            "Epoch 7/20\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.2427 - accuracy: 0.9146\n",
            "Epoch 8/20\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.2009 - accuracy: 0.9421\n",
            "Epoch 9/20\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.1708 - accuracy: 0.9565\n",
            "Epoch 10/20\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.1419 - accuracy: 0.9704\n",
            "Epoch 11/20\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.1190 - accuracy: 0.9786\n",
            "Epoch 12/20\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.1005 - accuracy: 0.9836\n",
            "Epoch 13/20\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0872 - accuracy: 0.9883\n",
            "Epoch 14/20\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0731 - accuracy: 0.9918\n",
            "Epoch 15/20\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0619 - accuracy: 0.9953\n",
            "Epoch 16/20\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0549 - accuracy: 0.9960\n",
            "Epoch 17/20\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0456 - accuracy: 0.9968\n",
            "Epoch 18/20\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0431 - accuracy: 0.9968\n",
            "Epoch 19/20\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0364 - accuracy: 0.9975\n",
            "Epoch 20/20\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0354 - accuracy: 0.9970\n",
            "Done training\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Accuracy: 0.7487586891757696\n",
            "Classification report:               precision    recall  f1-score   support\n",
            "\n",
            "         ham       0.76      0.80      0.78       560\n",
            "        spam       0.73      0.69      0.71       447\n",
            "\n",
            "    accuracy                           0.75      1007\n",
            "   macro avg       0.75      0.74      0.74      1007\n",
            "weighted avg       0.75      0.75      0.75      1007\n",
            "\n",
            "F1 (Micro): 0.7487586891757696\n",
            "F1 (Macro): 0.7438068719461732\n",
            "F1 (Weighted): 0.747803695852919\n",
            "Recall (Micro): 0.7487586891757696\n",
            "Recall (Macro): 0.742507590284436\n",
            "Recall (Weighted): 0.7487586891757696\n",
            "Precision (Micro): 0.7487586891757696\n",
            "Precision (Macro): 0.7462257645818122\n",
            "Precision (Weighted): 0.7479396596663463\n",
            "AUC: 0.742507590284436\n",
            "Number of epochs: 100\n",
            "Model: \"functional_11\"\n",
            "__________________________________________________________________________________________________\n",
            "Layer (type)                    Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            "input_word_ids (InputLayer)     [(None, 128)]        0                                            \n",
            "__________________________________________________________________________________________________\n",
            "input_mask (InputLayer)         [(None, 128)]        0                                            \n",
            "__________________________________________________________________________________________________\n",
            "segment_ids (InputLayer)        [(None, 128)]        0                                            \n",
            "__________________________________________________________________________________________________\n",
            "keras_layer_6 (KerasLayer)      [(None, 768), (None, 109482241   input_word_ids[0][0]             \n",
            "                                                                 input_mask[0][0]                 \n",
            "                                                                 segment_ids[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "global_average_pooling1d_5 (Glo (None, 768)          0           keras_layer_6[0][1]              \n",
            "__________________________________________________________________________________________________\n",
            "dense_10 (Dense)                (None, 768)          590592      global_average_pooling1d_5[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "dense_11 (Dense)                (None, 2)            1538        dense_10[0][0]                   \n",
            "==================================================================================================\n",
            "Total params: 110,074,371\n",
            "Trainable params: 592,130\n",
            "Non-trainable params: 109,482,241\n",
            "__________________________________________________________________________________________________\n",
            "None\n",
            "Epoch 1/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.5783 - accuracy: 0.7089\n",
            "Epoch 2/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.4734 - accuracy: 0.7814\n",
            "Epoch 3/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.4187 - accuracy: 0.8147\n",
            "Epoch 4/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.3703 - accuracy: 0.8373\n",
            "Epoch 5/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.3229 - accuracy: 0.8693\n",
            "Epoch 6/100\n",
            "126/126 [==============================] - 19s 152ms/step - loss: 0.2820 - accuracy: 0.8942\n",
            "Epoch 7/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.2357 - accuracy: 0.9225\n",
            "Epoch 8/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.2007 - accuracy: 0.9399\n",
            "Epoch 9/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.1713 - accuracy: 0.9548\n",
            "Epoch 10/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.1429 - accuracy: 0.9685\n",
            "Epoch 11/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.1202 - accuracy: 0.9801\n",
            "Epoch 12/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0986 - accuracy: 0.9856\n",
            "Epoch 13/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0841 - accuracy: 0.9901\n",
            "Epoch 14/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0705 - accuracy: 0.9918\n",
            "Epoch 15/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0597 - accuracy: 0.9950\n",
            "Epoch 16/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0537 - accuracy: 0.9960\n",
            "Epoch 17/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0466 - accuracy: 0.9968\n",
            "Epoch 18/100\n",
            "126/126 [==============================] - 19s 152ms/step - loss: 0.0417 - accuracy: 0.9970\n",
            "Epoch 19/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0380 - accuracy: 0.9970\n",
            "Epoch 20/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0317 - accuracy: 0.9975\n",
            "Epoch 21/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0295 - accuracy: 0.9973\n",
            "Epoch 22/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0285 - accuracy: 0.9978\n",
            "Epoch 23/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0256 - accuracy: 0.9970\n",
            "Epoch 24/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0243 - accuracy: 0.9970\n",
            "Epoch 25/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0214 - accuracy: 0.9978\n",
            "Epoch 26/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0200 - accuracy: 0.9978\n",
            "Epoch 27/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0218 - accuracy: 0.9975\n",
            "Epoch 28/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0217 - accuracy: 0.9973\n",
            "Epoch 29/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0211 - accuracy: 0.9975\n",
            "Epoch 30/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0207 - accuracy: 0.9970\n",
            "Epoch 31/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0188 - accuracy: 0.9975\n",
            "Epoch 32/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0154 - accuracy: 0.9978\n",
            "Epoch 33/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0180 - accuracy: 0.9973\n",
            "Epoch 34/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0171 - accuracy: 0.9975\n",
            "Epoch 35/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0178 - accuracy: 0.9975\n",
            "Epoch 36/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0195 - accuracy: 0.9970\n",
            "Epoch 37/100\n",
            "126/126 [==============================] - 19s 152ms/step - loss: 0.0157 - accuracy: 0.9975\n",
            "Epoch 38/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0162 - accuracy: 0.9978\n",
            "Epoch 39/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0156 - accuracy: 0.9978\n",
            "Epoch 40/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0183 - accuracy: 0.9973\n",
            "Epoch 41/100\n",
            "126/126 [==============================] - 19s 152ms/step - loss: 0.0159 - accuracy: 0.9975\n",
            "Epoch 42/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0167 - accuracy: 0.9975\n",
            "Epoch 43/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0149 - accuracy: 0.9978\n",
            "Epoch 44/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0174 - accuracy: 0.9973\n",
            "Epoch 45/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0173 - accuracy: 0.9973\n",
            "Epoch 46/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0165 - accuracy: 0.9973\n",
            "Epoch 47/100\n",
            "126/126 [==============================] - 19s 152ms/step - loss: 0.0183 - accuracy: 0.9973\n",
            "Epoch 48/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0131 - accuracy: 0.9980\n",
            "Epoch 49/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0136 - accuracy: 0.9980\n",
            "Epoch 50/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0145 - accuracy: 0.9978\n",
            "Epoch 51/100\n",
            "126/126 [==============================] - 19s 152ms/step - loss: 0.0133 - accuracy: 0.9980\n",
            "Epoch 52/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0171 - accuracy: 0.9975\n",
            "Epoch 53/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0136 - accuracy: 0.9980\n",
            "Epoch 54/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0143 - accuracy: 0.9978\n",
            "Epoch 55/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0149 - accuracy: 0.9975\n",
            "Epoch 56/100\n",
            "126/126 [==============================] - 19s 152ms/step - loss: 0.0171 - accuracy: 0.9975\n",
            "Epoch 57/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0190 - accuracy: 0.9973\n",
            "Epoch 58/100\n",
            "126/126 [==============================] - 19s 152ms/step - loss: 0.0138 - accuracy: 0.9978\n",
            "Epoch 59/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0177 - accuracy: 0.9973\n",
            "Epoch 60/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0165 - accuracy: 0.9973\n",
            "Epoch 61/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0176 - accuracy: 0.9978\n",
            "Epoch 62/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0170 - accuracy: 0.9975\n",
            "Epoch 63/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0181 - accuracy: 0.9973\n",
            "Epoch 64/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0181 - accuracy: 0.9970\n",
            "Epoch 65/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0152 - accuracy: 0.9978\n",
            "Epoch 66/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0149 - accuracy: 0.9975\n",
            "Epoch 67/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0158 - accuracy: 0.9973\n",
            "Epoch 68/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0162 - accuracy: 0.9975\n",
            "Epoch 69/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0156 - accuracy: 0.9978\n",
            "Epoch 70/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0168 - accuracy: 0.9973\n",
            "Epoch 71/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0121 - accuracy: 0.9980\n",
            "Epoch 72/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0142 - accuracy: 0.9980\n",
            "Epoch 73/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0172 - accuracy: 0.9975\n",
            "Epoch 74/100\n",
            "126/126 [==============================] - 19s 152ms/step - loss: 0.0139 - accuracy: 0.9978\n",
            "Epoch 75/100\n",
            "126/126 [==============================] - 19s 152ms/step - loss: 0.0141 - accuracy: 0.9973\n",
            "Epoch 76/100\n",
            "126/126 [==============================] - 19s 152ms/step - loss: 0.0140 - accuracy: 0.9978\n",
            "Epoch 77/100\n",
            "126/126 [==============================] - 19s 152ms/step - loss: 0.0127 - accuracy: 0.9978\n",
            "Epoch 78/100\n",
            "126/126 [==============================] - 19s 152ms/step - loss: 0.0170 - accuracy: 0.9975\n",
            "Epoch 79/100\n",
            "126/126 [==============================] - 19s 152ms/step - loss: 0.0172 - accuracy: 0.9975\n",
            "Epoch 80/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0143 - accuracy: 0.9978\n",
            "Epoch 81/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0175 - accuracy: 0.9973\n",
            "Epoch 82/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0164 - accuracy: 0.9973\n",
            "Epoch 83/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0146 - accuracy: 0.9978\n",
            "Epoch 84/100\n",
            "126/126 [==============================] - 19s 152ms/step - loss: 0.0151 - accuracy: 0.9978\n",
            "Epoch 85/100\n",
            "126/126 [==============================] - 19s 152ms/step - loss: 0.0136 - accuracy: 0.9978\n",
            "Epoch 86/100\n",
            "126/126 [==============================] - 19s 152ms/step - loss: 0.0123 - accuracy: 0.9980\n",
            "Epoch 87/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0168 - accuracy: 0.9975\n",
            "Epoch 88/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0179 - accuracy: 0.9973\n",
            "Epoch 89/100\n",
            "126/126 [==============================] - 19s 152ms/step - loss: 0.0126 - accuracy: 0.9980\n",
            "Epoch 90/100\n",
            "126/126 [==============================] - 19s 152ms/step - loss: 0.0142 - accuracy: 0.9980\n",
            "Epoch 91/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0148 - accuracy: 0.9978\n",
            "Epoch 92/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0168 - accuracy: 0.9973\n",
            "Epoch 93/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0158 - accuracy: 0.9975\n",
            "Epoch 94/100\n",
            "126/126 [==============================] - 19s 152ms/step - loss: 0.0137 - accuracy: 0.9980\n",
            "Epoch 95/100\n",
            "126/126 [==============================] - 19s 152ms/step - loss: 0.0179 - accuracy: 0.9970\n",
            "Epoch 96/100\n",
            "126/126 [==============================] - 19s 152ms/step - loss: 0.0149 - accuracy: 0.9973\n",
            "Epoch 97/100\n",
            "126/126 [==============================] - 19s 153ms/step - loss: 0.0123 - accuracy: 0.9975\n",
            "Epoch 98/100\n",
            "126/126 [==============================] - 19s 152ms/step - loss: 0.0145 - accuracy: 0.9975\n",
            "Epoch 99/100\n",
            "126/126 [==============================] - 19s 152ms/step - loss: 0.0166 - accuracy: 0.9973\n",
            "Epoch 100/100\n",
            "126/126 [==============================] - 19s 152ms/step - loss: 0.0144 - accuracy: 0.9980\n",
            "Done training\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Accuracy: 0.7477656405163853\n",
            "Classification report:               precision    recall  f1-score   support\n",
            "\n",
            "         ham       0.78      0.76      0.77       560\n",
            "        spam       0.71      0.74      0.72       447\n",
            "\n",
            "    accuracy                           0.75      1007\n",
            "   macro avg       0.74      0.75      0.75      1007\n",
            "weighted avg       0.75      0.75      0.75      1007\n",
            "\n",
            "F1 (Micro): 0.7477656405163853\n",
            "F1 (Macro): 0.7455957827730257\n",
            "F1 (Weighted): 0.7482322765902262\n",
            "Recall (Micro): 0.7477656405163853\n",
            "Recall (Macro): 0.7468060882070949\n",
            "Recall (Weighted): 0.7477656405163853\n",
            "Precision (Micro): 0.7477656405163853\n",
            "Precision (Macro): 0.7449857244825124\n",
            "Precision (Weighted): 0.7492888821787814\n",
            "AUC: 0.7468060882070949\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Rpcei2cLyV5n",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 374
        },
        "outputId": "cb35eaf3-f5e1-4615-a468-482ea190185f"
      },
      "source": [
        "train_input_ids, train_input_masks, train_segment_ids = create_req_bert_inputs(all_X_cleaned_using_bert, bert_tokenizer_instance)\n",
        "bert_model_preped_input = [np.asarray(train_input_ids, dtype=np.int32), np.asarray(train_input_masks, dtype=np.int32), np.asarray(train_segment_ids, dtype=np.int32)]\n",
        "all_Y = np.asarray(all_Y)\n",
        "model = create_model()\n",
        "model.compile(loss='binary_crossentropy', optimizer=this_optimizer, metrics=['accuracy'])\n",
        "history = model.fit(\n",
        "    bert_model_preped_input, \n",
        "    all_Y,\n",
        "    epochs=10, \n",
        "    batch_size = 32\n",
        ")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "100%|██████████| 5033/5033 [00:00<00:00, 58747.55it/s]\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/10\n",
            "158/158 [==============================] - 24s 153ms/step - loss: 0.7019 - accuracy: 0.5037\n",
            "Epoch 2/10\n",
            "158/158 [==============================] - 24s 152ms/step - loss: 0.6944 - accuracy: 0.5053\n",
            "Epoch 3/10\n",
            "158/158 [==============================] - 24s 153ms/step - loss: 0.6936 - accuracy: 0.5053\n",
            "Epoch 4/10\n",
            "158/158 [==============================] - 24s 152ms/step - loss: 0.6934 - accuracy: 0.5098\n",
            "Epoch 5/10\n",
            "158/158 [==============================] - 24s 153ms/step - loss: 0.6933 - accuracy: 0.4965\n",
            "Epoch 6/10\n",
            "158/158 [==============================] - 24s 153ms/step - loss: 0.6933 - accuracy: 0.4991\n",
            "Epoch 7/10\n",
            "158/158 [==============================] - 24s 153ms/step - loss: 0.6933 - accuracy: 0.5067\n",
            "Epoch 8/10\n",
            "158/158 [==============================] - 24s 153ms/step - loss: 0.6932 - accuracy: 0.5035\n",
            "Epoch 9/10\n",
            "158/158 [==============================] - 24s 153ms/step - loss: 0.6932 - accuracy: 0.4957\n",
            "Epoch 10/10\n",
            "158/158 [==============================] - 24s 153ms/step - loss: 0.6932 - accuracy: 0.4924\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2Osr0LYh-cBf",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "bfd65a4b-c26b-44aa-f424-bf336d6bbb05"
      },
      "source": [
        "covid_texts = []\n",
        "df1 = pd.read_csv('drive/My Drive/Metaphor Detection/Data-Texts/corona_tweets_01.csv', header=None)\n",
        "df2 = pd.read_csv('drive/My Drive/Metaphor Detection/Data-Texts/corona_tweets_32.csv', header=None)\n",
        "\n",
        "covid_texts = df1[1].values.tolist()+df2[1].values.tolist()\n",
        "covid_ids = df1[0].values.tolist()+df2[0].values.tolist()\n",
        "\n",
        "test_input_ids, test_input_masks, test_segment_ids = create_req_bert_inputs(clean_text(covid_texts, bert_tokenizer_instance), bert_tokenizer_instance)\n",
        "bert_model_test_input = [np.asarray(test_input_ids, dtype=np.int32), np.asarray(test_input_masks, dtype=np.int32), np.asarray(test_segment_ids, dtype=np.int32)]\n",
        "predicted_y = model.predict(bert_model_test_input)\n",
        "predicted_y_conv = [np.argmax(val) for val in predicted_y]"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "100%|██████████| 9561/9561 [00:00<00:00, 64246.67it/s]\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pr6Hj09j2PVg",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "dd36151a-ce02-4b5f-ce21-c24c05ca2d79"
      },
      "source": [
        "lable_counts = {0:0, 1:0}\n",
        "for item in predicted_y_conv:\n",
        "  lable_counts[item]+=1\n",
        "lable_counts"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "{0: 3647, 1: 5914}"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 66
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5MCzByMj23D1",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "3d28e258-2df2-4f3c-aa99-539bfff296d3"
      },
      "source": [
        "5914/9661"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.612151951143774"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 67
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vUv71F9W3YMl",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        },
        "outputId": "ed04c879-5694-42e8-cbe4-87bbc9745dd2"
      },
      "source": [
        "metaphor_sentiment_mapping = {'pos':{0:0, 1:0}, 'neg':{0:0, 1:0}}\n",
        "df11 = pd.read_csv('drive/My Drive/Metaphor Detection/corona_tweets_01.csv', header=None)\n",
        "df11 = df11[df11[0].isin(covid_ids)]\n",
        "df21 = pd.read_csv('drive/My Drive/Metaphor Detection/corona_tweets_32.csv', header=None)\n",
        "df21 = df21[df21[0].isin(covid_ids)]\n",
        "print(df11.shape, df21.shape)\n",
        "for _, row in df11.iterrows():\n",
        "  if row[0] in covid_ids:\n",
        "    ind = covid_ids.index(row[0])\n",
        "    if row[1]<0:\n",
        "      metaphor_sentiment_mapping['neg'][predicted_y_conv[ind]] += 1\n",
        "    else:\n",
        "      metaphor_sentiment_mapping['pos'][predicted_y_conv[ind]] += 1\n",
        "for _, row in df21.iterrows():\n",
        "  if row[0] in covid_ids:\n",
        "    ind = covid_ids.index(row[0])\n",
        "    if row[1]<0:\n",
        "      metaphor_sentiment_mapping['neg'][predicted_y_conv[ind]] += 1\n",
        "    else:\n",
        "      metaphor_sentiment_mapping['pos'][predicted_y_conv[ind]] += 1\n",
        "metaphor_sentiment_mapping"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(4804, 2) (4757, 2)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "{'neg': {0: 865, 1: 1394}, 'pos': {0: 2783, 1: 4519}}"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 76
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "upGjKZUu799h",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}