Main.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Main.ipynb",
      "provenance": [],
      "toc_visible": true,
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/Positron23/ignitionhacks/blob/master/Main.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "hKoGvGaxqlPY",
        "colab_type": "text"
      },
      "source": [
        "# Initialization\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "G2o8eYVOqwC2",
        "colab_type": "text"
      },
      "source": [
        "Check GPU"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "i9-h3tbNLJQX",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 384
        },
        "outputId": "a9606217-feff-488f-c7da-e6d11c86d6e5"
      },
      "source": [
        "!nvidia-smi"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Mon Aug 24 03:58:23 2020       \n",
            "+-----------------------------------------------------------------------------+\n",
            "| NVIDIA-SMI 450.57       Driver Version: 418.67       CUDA Version: 10.1     |\n",
            "|-------------------------------+----------------------+----------------------+\n",
            "| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |\n",
            "| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |\n",
            "|                               |                      |               MIG M. |\n",
            "|===============================+======================+======================|\n",
            "|   0  Tesla K80           Off  | 00000000:00:04.0 Off |                    0 |\n",
            "| N/A   52C    P8    31W / 149W |      0MiB / 11441MiB |      0%      Default |\n",
            "|                               |                      |                 ERR! |\n",
            "+-------------------------------+----------------------+----------------------+\n",
            "                                                                               \n",
            "+-----------------------------------------------------------------------------+\n",
            "| Processes:                                                                  |\n",
            "|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |\n",
            "|        ID   ID                                                   Usage      |\n",
            "|=============================================================================|\n",
            "|  No running processes found                                                 |\n",
            "+-----------------------------------------------------------------------------+\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7yQvx1mDq3mP",
        "colab_type": "text"
      },
      "source": [
        "Install Tensorflow-GPU \n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BY05zCipJMZW",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "051f350a-94d0-4740-c24d-981baa0cebeb"
      },
      "source": [
        "!pip install tensorflow-gpu\n",
        "!pip install tensorflow_text"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Collecting tensorflow-gpu\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/0f/11/763f55d3d15efd778ef24453f126e6c33635680e5a2bb346da3fab5997cb/tensorflow_gpu-2.3.0-cp36-cp36m-manylinux2010_x86_64.whl (320.4MB)\n",
            "\u001b[K     |████████████████████████████████| 320.4MB 52kB/s \n",
            "\u001b[?25hRequirement already satisfied: opt-einsum>=2.3.2 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (3.3.0)\n",
            "Requirement already satisfied: h5py<2.11.0,>=2.10.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (2.10.0)\n",
            "Requirement already satisfied: gast==0.3.3 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (0.3.3)\n",
            "Requirement already satisfied: keras-preprocessing<1.2,>=1.1.1 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (1.1.2)\n",
            "Requirement already satisfied: protobuf>=3.9.2 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (3.12.4)\n",
            "Requirement already satisfied: six>=1.12.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (1.15.0)\n",
            "Requirement already satisfied: grpcio>=1.8.6 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (1.31.0)\n",
            "Requirement already satisfied: astunparse==1.6.3 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (1.6.3)\n",
            "Requirement already satisfied: termcolor>=1.1.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (1.1.0)\n",
            "Requirement already satisfied: wrapt>=1.11.1 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (1.12.1)\n",
            "Requirement already satisfied: wheel>=0.26 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (0.34.2)\n",
            "Requirement already satisfied: absl-py>=0.7.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (0.9.0)\n",
            "Requirement already satisfied: tensorboard<3,>=2.3.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (2.3.0)\n",
            "Requirement already satisfied: scipy==1.4.1 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (1.4.1)\n",
            "Requirement already satisfied: tensorflow-estimator<2.4.0,>=2.3.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (2.3.0)\n",
            "Requirement already satisfied: google-pasta>=0.1.8 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (0.2.0)\n",
            "Requirement already satisfied: numpy<1.19.0,>=1.16.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (1.18.5)\n",
            "Requirement already satisfied: setuptools in /usr/local/lib/python3.6/dist-packages (from protobuf>=3.9.2->tensorflow-gpu) (49.2.0)\n",
            "Requirement already satisfied: requests<3,>=2.21.0 in /usr/local/lib/python3.6/dist-packages (from tensorboard<3,>=2.3.0->tensorflow-gpu) (2.23.0)\n",
            "Requirement already satisfied: markdown>=2.6.8 in /usr/local/lib/python3.6/dist-packages (from tensorboard<3,>=2.3.0->tensorflow-gpu) (3.2.2)\n",
            "Requirement already satisfied: google-auth-oauthlib<0.5,>=0.4.1 in /usr/local/lib/python3.6/dist-packages (from tensorboard<3,>=2.3.0->tensorflow-gpu) (0.4.1)\n",
            "Requirement already satisfied: tensorboard-plugin-wit>=1.6.0 in /usr/local/lib/python3.6/dist-packages (from tensorboard<3,>=2.3.0->tensorflow-gpu) (1.7.0)\n",
            "Requirement already satisfied: werkzeug>=0.11.15 in /usr/local/lib/python3.6/dist-packages (from tensorboard<3,>=2.3.0->tensorflow-gpu) (1.0.1)\n",
            "Requirement already satisfied: google-auth<2,>=1.6.3 in /usr/local/lib/python3.6/dist-packages (from tensorboard<3,>=2.3.0->tensorflow-gpu) (1.17.2)\n",
            "Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.6/dist-packages (from requests<3,>=2.21.0->tensorboard<3,>=2.3.0->tensorflow-gpu) (3.0.4)\n",
            "Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.6/dist-packages (from requests<3,>=2.21.0->tensorboard<3,>=2.3.0->tensorflow-gpu) (2.10)\n",
            "Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.6/dist-packages (from requests<3,>=2.21.0->tensorboard<3,>=2.3.0->tensorflow-gpu) (1.24.3)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.6/dist-packages (from requests<3,>=2.21.0->tensorboard<3,>=2.3.0->tensorflow-gpu) (2020.6.20)\n",
            "Requirement already satisfied: importlib-metadata; python_version < \"3.8\" in /usr/local/lib/python3.6/dist-packages (from markdown>=2.6.8->tensorboard<3,>=2.3.0->tensorflow-gpu) (1.7.0)\n",
            "Requirement already satisfied: requests-oauthlib>=0.7.0 in /usr/local/lib/python3.6/dist-packages (from google-auth-oauthlib<0.5,>=0.4.1->tensorboard<3,>=2.3.0->tensorflow-gpu) (1.3.0)\n",
            "Requirement already satisfied: cachetools<5.0,>=2.0.0 in /usr/local/lib/python3.6/dist-packages (from google-auth<2,>=1.6.3->tensorboard<3,>=2.3.0->tensorflow-gpu) (4.1.1)\n",
            "Requirement already satisfied: pyasn1-modules>=0.2.1 in /usr/local/lib/python3.6/dist-packages (from google-auth<2,>=1.6.3->tensorboard<3,>=2.3.0->tensorflow-gpu) (0.2.8)\n",
            "Requirement already satisfied: rsa<5,>=3.1.4; python_version >= \"3\" in /usr/local/lib/python3.6/dist-packages (from google-auth<2,>=1.6.3->tensorboard<3,>=2.3.0->tensorflow-gpu) (4.6)\n",
            "Requirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.6/dist-packages (from importlib-metadata; python_version < \"3.8\"->markdown>=2.6.8->tensorboard<3,>=2.3.0->tensorflow-gpu) (3.1.0)\n",
            "Requirement already satisfied: oauthlib>=3.0.0 in /usr/local/lib/python3.6/dist-packages (from requests-oauthlib>=0.7.0->google-auth-oauthlib<0.5,>=0.4.1->tensorboard<3,>=2.3.0->tensorflow-gpu) (3.1.0)\n",
            "Requirement already satisfied: pyasn1<0.5.0,>=0.4.6 in /usr/local/lib/python3.6/dist-packages (from pyasn1-modules>=0.2.1->google-auth<2,>=1.6.3->tensorboard<3,>=2.3.0->tensorflow-gpu) (0.4.8)\n",
            "Installing collected packages: tensorflow-gpu\n",
            "Successfully installed tensorflow-gpu-2.3.0\n",
            "Collecting tensorflow_text\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/28/b2/2dbd90b93913afd07e6101b8b84327c401c394e60141c1e98590038060b3/tensorflow_text-2.3.0-cp36-cp36m-manylinux1_x86_64.whl (2.6MB)\n",
            "\u001b[K     |████████████████████████████████| 2.6MB 2.8MB/s \n",
            "\u001b[?25hRequirement already satisfied: tensorflow<2.4,>=2.3.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow_text) (2.3.0)\n",
            "Requirement already satisfied: absl-py>=0.7.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow<2.4,>=2.3.0->tensorflow_text) (0.9.0)\n",
            "Requirement already satisfied: six>=1.12.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow<2.4,>=2.3.0->tensorflow_text) (1.15.0)\n",
            "Requirement already satisfied: astunparse==1.6.3 in /usr/local/lib/python3.6/dist-packages (from tensorflow<2.4,>=2.3.0->tensorflow_text) (1.6.3)\n",
            "Requirement already satisfied: protobuf>=3.9.2 in /usr/local/lib/python3.6/dist-packages (from tensorflow<2.4,>=2.3.0->tensorflow_text) (3.12.4)\n",
            "Requirement already satisfied: tensorboard<3,>=2.3.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow<2.4,>=2.3.0->tensorflow_text) (2.3.0)\n",
            "Requirement already satisfied: grpcio>=1.8.6 in /usr/local/lib/python3.6/dist-packages (from tensorflow<2.4,>=2.3.0->tensorflow_text) (1.31.0)\n",
            "Requirement already satisfied: numpy<1.19.0,>=1.16.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow<2.4,>=2.3.0->tensorflow_text) (1.18.5)\n",
            "Requirement already satisfied: gast==0.3.3 in /usr/local/lib/python3.6/dist-packages (from tensorflow<2.4,>=2.3.0->tensorflow_text) (0.3.3)\n",
            "Requirement already satisfied: wrapt>=1.11.1 in /usr/local/lib/python3.6/dist-packages (from tensorflow<2.4,>=2.3.0->tensorflow_text) (1.12.1)\n",
            "Requirement already satisfied: scipy==1.4.1 in /usr/local/lib/python3.6/dist-packages (from tensorflow<2.4,>=2.3.0->tensorflow_text) (1.4.1)\n",
            "Requirement already satisfied: wheel>=0.26 in /usr/local/lib/python3.6/dist-packages (from tensorflow<2.4,>=2.3.0->tensorflow_text) (0.34.2)\n",
            "Requirement already satisfied: keras-preprocessing<1.2,>=1.1.1 in /usr/local/lib/python3.6/dist-packages (from tensorflow<2.4,>=2.3.0->tensorflow_text) (1.1.2)\n",
            "Requirement already satisfied: opt-einsum>=2.3.2 in /usr/local/lib/python3.6/dist-packages (from tensorflow<2.4,>=2.3.0->tensorflow_text) (3.3.0)\n",
            "Requirement already satisfied: termcolor>=1.1.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow<2.4,>=2.3.0->tensorflow_text) (1.1.0)\n",
            "Requirement already satisfied: h5py<2.11.0,>=2.10.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow<2.4,>=2.3.0->tensorflow_text) (2.10.0)\n",
            "Requirement already satisfied: tensorflow-estimator<2.4.0,>=2.3.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow<2.4,>=2.3.0->tensorflow_text) (2.3.0)\n",
            "Requirement already satisfied: google-pasta>=0.1.8 in /usr/local/lib/python3.6/dist-packages (from tensorflow<2.4,>=2.3.0->tensorflow_text) (0.2.0)\n",
            "Requirement already satisfied: setuptools in /usr/local/lib/python3.6/dist-packages (from protobuf>=3.9.2->tensorflow<2.4,>=2.3.0->tensorflow_text) (49.2.0)\n",
            "Requirement already satisfied: requests<3,>=2.21.0 in /usr/local/lib/python3.6/dist-packages (from tensorboard<3,>=2.3.0->tensorflow<2.4,>=2.3.0->tensorflow_text) (2.23.0)\n",
            "Requirement already satisfied: google-auth-oauthlib<0.5,>=0.4.1 in /usr/local/lib/python3.6/dist-packages (from tensorboard<3,>=2.3.0->tensorflow<2.4,>=2.3.0->tensorflow_text) (0.4.1)\n",
            "Requirement already satisfied: tensorboard-plugin-wit>=1.6.0 in /usr/local/lib/python3.6/dist-packages (from tensorboard<3,>=2.3.0->tensorflow<2.4,>=2.3.0->tensorflow_text) (1.7.0)\n",
            "Requirement already satisfied: markdown>=2.6.8 in /usr/local/lib/python3.6/dist-packages (from tensorboard<3,>=2.3.0->tensorflow<2.4,>=2.3.0->tensorflow_text) (3.2.2)\n",
            "Requirement already satisfied: google-auth<2,>=1.6.3 in /usr/local/lib/python3.6/dist-packages (from tensorboard<3,>=2.3.0->tensorflow<2.4,>=2.3.0->tensorflow_text) (1.17.2)\n",
            "Requirement already satisfied: werkzeug>=0.11.15 in /usr/local/lib/python3.6/dist-packages (from tensorboard<3,>=2.3.0->tensorflow<2.4,>=2.3.0->tensorflow_text) (1.0.1)\n",
            "Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.6/dist-packages (from requests<3,>=2.21.0->tensorboard<3,>=2.3.0->tensorflow<2.4,>=2.3.0->tensorflow_text) (2.10)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.6/dist-packages (from requests<3,>=2.21.0->tensorboard<3,>=2.3.0->tensorflow<2.4,>=2.3.0->tensorflow_text) (2020.6.20)\n",
            "Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.6/dist-packages (from requests<3,>=2.21.0->tensorboard<3,>=2.3.0->tensorflow<2.4,>=2.3.0->tensorflow_text) (3.0.4)\n",
            "Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.6/dist-packages (from requests<3,>=2.21.0->tensorboard<3,>=2.3.0->tensorflow<2.4,>=2.3.0->tensorflow_text) (1.24.3)\n",
            "Requirement already satisfied: requests-oauthlib>=0.7.0 in /usr/local/lib/python3.6/dist-packages (from google-auth-oauthlib<0.5,>=0.4.1->tensorboard<3,>=2.3.0->tensorflow<2.4,>=2.3.0->tensorflow_text) (1.3.0)\n",
            "Requirement already satisfied: importlib-metadata; python_version < \"3.8\" in /usr/local/lib/python3.6/dist-packages (from markdown>=2.6.8->tensorboard<3,>=2.3.0->tensorflow<2.4,>=2.3.0->tensorflow_text) (1.7.0)\n",
            "Requirement already satisfied: pyasn1-modules>=0.2.1 in /usr/local/lib/python3.6/dist-packages (from google-auth<2,>=1.6.3->tensorboard<3,>=2.3.0->tensorflow<2.4,>=2.3.0->tensorflow_text) (0.2.8)\n",
            "Requirement already satisfied: cachetools<5.0,>=2.0.0 in /usr/local/lib/python3.6/dist-packages (from google-auth<2,>=1.6.3->tensorboard<3,>=2.3.0->tensorflow<2.4,>=2.3.0->tensorflow_text) (4.1.1)\n",
            "Requirement already satisfied: rsa<5,>=3.1.4; python_version >= \"3\" in /usr/local/lib/python3.6/dist-packages (from google-auth<2,>=1.6.3->tensorboard<3,>=2.3.0->tensorflow<2.4,>=2.3.0->tensorflow_text) (4.6)\n",
            "Requirement already satisfied: oauthlib>=3.0.0 in /usr/local/lib/python3.6/dist-packages (from requests-oauthlib>=0.7.0->google-auth-oauthlib<0.5,>=0.4.1->tensorboard<3,>=2.3.0->tensorflow<2.4,>=2.3.0->tensorflow_text) (3.1.0)\n",
            "Requirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.6/dist-packages (from importlib-metadata; python_version < \"3.8\"->markdown>=2.6.8->tensorboard<3,>=2.3.0->tensorflow<2.4,>=2.3.0->tensorflow_text) (3.1.0)\n",
            "Requirement already satisfied: pyasn1<0.5.0,>=0.4.6 in /usr/local/lib/python3.6/dist-packages (from pyasn1-modules>=0.2.1->google-auth<2,>=1.6.3->tensorboard<3,>=2.3.0->tensorflow<2.4,>=2.3.0->tensorflow_text) (0.4.8)\n",
            "Installing collected packages: tensorflow-text\n",
            "Successfully installed tensorflow-text-2.3.0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "rH7UcCkJq7dt",
        "colab_type": "text"
      },
      "source": [
        "Import Packages"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HrbJenMbhD4N",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# DATA PROCESSING\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import gzip # To utilize previously encoded data\n",
        "\n",
        "# DEEP LEARNING\n",
        "import tensorflow as tf \n",
        "from tensorflow import keras\n",
        "from sklearn.model_selection import train_test_split\n",
        "import tensorflow_hub as hub\n",
        "import tensorflow_text\n",
        "\n",
        "# PREPROCESSING\n",
        "import re\n",
        "import inflect\n",
        "import string, unicodedata\n",
        "import nltk\n",
        "from nltk.corpus import stopwords\n",
        "from nltk.stem import PorterStemmer\n",
        "from nltk.tokenize import sent_tokenize, word_tokenize\n",
        "\n",
        "# DATA DISPLAY\n",
        "import matplotlib.pyplot as plt\n",
        "from tqdm import tqdm #For showing progress bars"
      ],
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "NCeZg4vJSVQ2",
        "colab_type": "text"
      },
      "source": [
        "# Preprocessing"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WQBs-iJ9kIX9",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "e5f54048-538d-4e00-ef22-a10639f88682"
      },
      "source": [
        "# READ ORIGINAL DATA\n",
        "df = pd.read_csv('training_data.csv', usecols=['Text', 'Sentiment'])\n",
        "print(df.shape)"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(218598, 2)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "V6v6A8WroHXV",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# SET RANDOM SEED FOR REPRODUCIBLE TRAIN/DEV/TEST SETS\n",
        "RANDOM_SEED = 42\n",
        "np.random.seed(RANDOM_SEED)\n",
        "\n",
        "#tf.random.set_seed(RANDOM_SEED)"
      ],
      "execution_count": 34,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1_m0B6sHoHIp",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# SPLIT DATA\n",
        "train, test, y_train, y_test =\\\n",
        "  train_test_split(\n",
        "    df.Text, \n",
        "    df.Sentiment, \n",
        "    test_size=.1, \n",
        "    random_state=RANDOM_SEED)"
      ],
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2_7ktKK8lfTn",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# DOWNLOAD NATURAL LANGUAGE PREPROCESSING TOOLKITS\n",
        "nltk.download('punkt') # for stemming \n",
        "nltk.download('stopwords') # for stopwords\n",
        "porter = PorterStemmer() # create porter stemming algorithm object"
      ],
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bThxyrWJDvQ6",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 164
        },
        "outputId": "18caa284-e5f6-4b95-af1c-c99c23b54476"
      },
      "source": [
        "# CHECK SHAPES AND DATA\n",
        "print(train.shape)\n",
        "print(test.shape)\n",
        "train.head()"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(900000,)\n",
            "(100000,)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "378046    @plsdontgogurl That's the one. I wish I have a...\n",
              "510062               I just woke up and I feel soooo tired \n",
              "534815    @nainargrl I wish I could do that   I am not t...\n",
              "342590      @MariahCarey  lol hey i am in bed too cuddling \n",
              "559414    today i'm going to friend's house  it's gonna ...\n",
              "Name: Text, dtype: object"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 9
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8JlIhUNlq_Bw",
        "colab_type": "text"
      },
      "source": [
        "# Encoding\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0FZFXbYYrTsU",
        "colab_type": "text"
      },
      "source": [
        "## Extract from previously encoded data"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "wonvW7sBfYt_",
        "colab_type": "text"
      },
      "source": [
        "Add training and test datasets to your drive by right clicking and selecting 'make a shortcut to your drive'. Afterwards, the output will prompt for an authentication password, which you can receive by allowing the file stream to access the drive in which the datasets are stored."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7VbuSXP0pHEN",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# from google.colab import drive\n",
        "# drive.mount(\"/content/drive\")"
      ],
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "K2iPd_VSuJm6",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "f=gzip.GzipFile('/content/drive/My Drive/train.npy.gz','r')\n",
        "X_train = np.load(f)\n",
        "X_train = np.array(X_train)\n",
        "f2=gzip.GzipFile('/content/drive/My Drive/test.npy.gz','r')\n",
        "X_test = np.load(f2)\n",
        "X_test = np.array(X_test)\n"
      ],
      "execution_count": 11,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "oqAfus6sraAq",
        "colab_type": "text"
      },
      "source": [
        "## Re-encode data (~3 to 4 hours)\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pitgN_EpC2WJ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# LOAD UNIVERAL SERIAL ENCODER\n",
        "use = hub.load(\"https://tfhub.dev/google/universal-sentence-encoder-multilingual-large/3\")\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "szEzsQJ0rwiO",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "## OLD ENCODING ALGORITHM ##\n",
        "# splitlist = np.array_split(trainlist,50)\n",
        "# # GET EMBEDDINGS\n",
        "# embedlist=[]\n",
        "# j=0\n",
        "# for i in splitlist:\n",
        "#   s = embed(i)\n",
        "#   for k in s:\n",
        "#     embedlist.append(k)\n",
        "#     j+=1\n",
        "#     print(j)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vpHsrBC6r2ny",
        "colab_type": "text"
      },
      "source": [
        "Encoding Training and Test Sets\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "l8bdM-kbDCz6",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "1a29bd29-c2aa-4e23-8596-77c817a26caa"
      },
      "source": [
        "X_train = np.zeros((train.shape[0],512))\n",
        "i=0\n",
        "for r in tq.tqdm(trains):\n",
        "  X_train[i] = tf.reshape(use(r),[-1]).numpy()\n",
        "  i+=1"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "100%|██████████| 900000/900000 [3:10:52<00:00, 78.59it/s]\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kKEAbLBsL4y0",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "39dc6280-57eb-444b-811f-27e3dfb3c464"
      },
      "source": [
        "X_test = np.zeros((test.shape[0],512))\n",
        "i=0\n",
        "for r in tq.tqdm(tests):\n",
        "  X_test[i] = tf.reshape(use(r),[-1]).numpy()\n",
        "  i+=1"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "100%|██████████| 100000/100000 [20:38<00:00, 80.74it/s]\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3GNwxoZUstro",
        "colab_type": "text"
      },
      "source": [
        "## Check Encodings\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "g_chHEudr9J0",
        "colab_type": "text"
      },
      "source": [
        "Check if shapes and values are correct: Expected (n,512) 32-bit float vectors for feature train/test arrays and (n,) for labels"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0R9UhEMgL_QR",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 274
        },
        "outputId": "3bc38158-27d8-447f-d63a-829ee80479c1"
      },
      "source": [
        "print(X_train.shape,X_test.shape,y_train.shape)\n",
        "print(X_test)"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(900000, 512) (100000, 512) (900000,)\n",
            "[[-0.04484706 -0.0484597  -0.02886226 ... -0.01150637  0.01021791\n",
            "  -0.07051698]\n",
            " [ 0.01337283  0.0073399   0.03012324 ... -0.04162316  0.03671977\n",
            "  -0.03627334]\n",
            " [ 0.06179673 -0.02251715  0.02605925 ...  0.0276723  -0.04356809\n",
            "   0.05065024]\n",
            " ...\n",
            " [-0.00890851  0.01316289  0.01039192 ... -0.01220796 -0.0712463\n",
            "   0.10207573]\n",
            " [-0.05828437 -0.00819871  0.02271808 ... -0.03066864 -0.02567255\n",
            "   0.00929535]\n",
            " [-0.0683042   0.04698984  0.02167922 ... -0.05713771  0.00492626\n",
            "   0.04706876]]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zMOo-gqhsSqI",
        "colab_type": "text"
      },
      "source": [
        "Save Encodings and Compress"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SkV2M3tDBOYo",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "f = gzip.GzipFile(\"train.npy.gz\", \"w\")\n",
        "np.save(file=f, arr=X_train)\n",
        "f.close()\n",
        "f2 = gzip.GzipFile(\"test.npy.gz\",\"w\")\n",
        "np.save(file=f2, arr=X_test)\n",
        "f2.close()\n",
        "\n",
        "## Non-compressed saving creates 12GB in filesize\n",
        "#np.savetxt('test.csv',X_train,delimiter=',')\n",
        "#np.savetxt('train.csv',X_test,delimiter=',')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "IuLnuWQirGCe",
        "colab_type": "text"
      },
      "source": [
        "\n",
        "# Dimensionality Reshaping for Alternate Models"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Ad-2LG_JF4gP",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# from sklearn.preprocessing import OneHotEncoder\n",
        "# y_train = OneHotEncoder(sparse=False).fit_transform(y_train.to_numpy().reshape(-1,1))\n",
        "# y_test = OneHotEncoder(sparse=False).fit_transform(y_test.to_numpy().reshape(-1,1))\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PaXBEF1Ue3fM",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# maxpolxtr = X_train.reshape(-1, 1, 512)\n",
        "# maxpolxte  = X_test.reshape(-1, 1, 512)\n",
        "# maxpolytr = np.array(y_train).reshape(-1, 1, 1)\n",
        "# maxpolyte = np.array(y_test).reshape(-1, 1, 1)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "s6RjvlsJrMMh",
        "colab_type": "text"
      },
      "source": [
        "# Model\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "B1oi5lYA6eZg",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model = keras.Sequential()\n",
        "\n",
        "model.add(\n",
        "  keras.layers.Dense(\n",
        "    units=128,\n",
        "    input_shape=(X_train.shape[1], ),\n",
        "    activation='relu'\n",
        "  )\n",
        ")\n",
        "model.add(\n",
        "  keras.layers.Dropout(rate=0.5)\n",
        ")\n",
        "\n",
        "model.add(\n",
        "  keras.layers.Dense(\n",
        "    units=64,\n",
        "    kernel_regularizer='l2',\n",
        "    activation='relu'\n",
        "  )\n",
        ")\n",
        "model.add(\n",
        "  keras.layers.Dropout(rate=0.5)\n",
        ")\n",
        "\n",
        "# model.add(\n",
        "#   keras.layers.Dense(\n",
        "#     units=64,\n",
        "#     kernel_regularizer=keras.regularizers.l2(0.001),\n",
        "#     activation='relu'\n",
        "#   )\n",
        "# )\n",
        "# model.add(\n",
        "#   keras.layers.Dropout(rate=0.5)\n",
        "# )\n",
        "\n",
        "model.add(keras.layers.Dense(1, activation='sigmoid'))\n",
        "model.compile(\n",
        "    loss='binary_crossentropy', \n",
        "    optimizer=keras.optimizers.Adam(0.001),\n",
        "    metrics=['accuracy']\n",
        ")\n",
        "\n"
      ],
      "execution_count": 59,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "74Dwg9W0bytC",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# model = keras.Sequential([\n",
        "#     keras.layers.GlobalMaxPooling1D(),\n",
        "#     keras.layers.Dense(256,activation='relu'),\n",
        "#     keras.layers.Dropout(0.3),\n",
        "#     keras.layers.Dense(128,activation='relu'),\n",
        "#     keras.layers.Dropout(0.5),\n",
        "#     keras.layers.Dense(64,activation='relu'),\n",
        "#     keras.layers.Dense(1, activation='sigmoid')\n",
        "# ])\n",
        "# model.compile(loss=\"binary_crossentropy\", optimizer=keras.optimizers,Adam(0.01), metrics=['accuracy'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "M1dfRaWrcE9a",
        "colab_type": "text"
      },
      "source": [
        "#### CNN"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "P7Dih_lhb0yI",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# model = tf.keras.Sequential([\n",
        "#     tf.keras.layers.Embedding(vocab_size, 16, input_length=max_len),\n",
        "#     tf.keras.layers.SeparableConv1D(16, 5, activation='relu', bias_initializer='random_uniform', depthwise_initializer='random_uniform', padding='same'),\n",
        "#     tf.keras.layers.GlobalMaxPooling1D(),\n",
        "#     tf.keras.layers.Dense(128, activation='relu'),\n",
        "#     tf.keras.layers.Dropout(0.5),\n",
        "#     tf.keras.layers.Dense(1, activation='sigmoid')\n",
        "# ])\n",
        "\n",
        "# model.compile(loss='binary_crossentropy', optimizer=tf.keras.optimizers.Adam(0.001), metrics=['accuracy'])\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "r0c7miyLcB26",
        "colab_type": "text"
      },
      "source": [
        "#### Bidirectional LSTM"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SETa347Jb5Yz",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# model = tf.keras.Sequential([\n",
        "#     tf.keras.layers.Embedding(vocab_size, 16, input_length=max_len),\n",
        "#     tf.keras.layers.Bidirectional(tf.keras.layers.LSTM(16, return_sequences=True)),\n",
        "#     tf.keras.layers.Bidirectional(tf.keras.layers.LSTM(16)),\n",
        "#     tf.keras.layers.Dense(1, activation='sigmoid')\n",
        "# ])\n",
        "# model.compile(loss='binary_crossentropy', optimizer=tf.keras.optimizers.Adam(0.001), metrics=['accuracy'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "GYaYEVYjbIi8",
        "colab_type": "text"
      },
      "source": [
        "####Early Stopping\n",
        "Determines which epoch model starts to overfit and lose validation accuracy"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "06gm06pmbtvT",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "class ResetStatesCallback(tf.keras.callbacks.Callback):\n",
        "    def on_epoch_begin(self, epoch, logs):\n",
        "        self.model.reset_states()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "FZX-ok6zbvOv",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "reset_states = ResetStatesCallback()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5oG8seDQbHOE",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "early_stopping = tf.keras.callbacks.EarlyStopping(patience=50)\n",
        "model_checkpoint = tf.keras.callbacks.ModelCheckpoint(\"maincheckpoint.h5\", save_best_only=True)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "dchUPdcsbMN-",
        "colab_type": "text"
      },
      "source": [
        "Train\n",
        "\n",
        "\n",
        "*   8-10 Epochs\n",
        "*   Default learning rate (0.001)\n",
        "*   16 Batch Size"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8W6Al0qO62M-",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 404
        },
        "outputId": "b4810d42-70e1-4f1b-8f87-9261a8e6acf2"
      },
      "source": [
        "history = model.fit(\n",
        "    X_train,y_train, \n",
        "    epochs=10, \n",
        "    batch_size=16, \n",
        "    validation_data=(X_test,y_test), \n",
        "    verbose=1, \n",
        "    shuffle=True\n",
        ")"
      ],
      "execution_count": 70,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/10\n",
            "56250/56250 [==============================] - 116s 2ms/step - loss: 0.4439 - accuracy: 0.8086 - val_loss: 0.4328 - val_accuracy: 0.8069\n",
            "Epoch 2/10\n",
            "56250/56250 [==============================] - 118s 2ms/step - loss: 0.4433 - accuracy: 0.8093 - val_loss: 0.4304 - val_accuracy: 0.8072\n",
            "Epoch 3/10\n",
            "56250/56250 [==============================] - 117s 2ms/step - loss: 0.4432 - accuracy: 0.8094 - val_loss: 0.4314 - val_accuracy: 0.8070\n",
            "Epoch 4/10\n",
            "56250/56250 [==============================] - 117s 2ms/step - loss: 0.4421 - accuracy: 0.8100 - val_loss: 0.4324 - val_accuracy: 0.8071\n",
            "Epoch 5/10\n",
            "56250/56250 [==============================] - 118s 2ms/step - loss: 0.4422 - accuracy: 0.8098 - val_loss: 0.4335 - val_accuracy: 0.8074\n",
            "Epoch 6/10\n",
            "56250/56250 [==============================] - 116s 2ms/step - loss: 0.4419 - accuracy: 0.8102 - val_loss: 0.4331 - val_accuracy: 0.8071\n",
            "Epoch 7/10\n",
            "56250/56250 [==============================] - 117s 2ms/step - loss: 0.4416 - accuracy: 0.8103 - val_loss: 0.4314 - val_accuracy: 0.8074\n",
            "Epoch 8/10\n",
            "56250/56250 [==============================] - 117s 2ms/step - loss: 0.4415 - accuracy: 0.8105 - val_loss: 0.4333 - val_accuracy: 0.8066\n",
            "Epoch 9/10\n",
            "56250/56250 [==============================] - 116s 2ms/step - loss: 0.4408 - accuracy: 0.8106 - val_loss: 0.4318 - val_accuracy: 0.8077\n",
            "Epoch 10/10\n",
            "56250/56250 [==============================] - 119s 2ms/step - loss: 0.4410 - accuracy: 0.8106 - val_loss: 0.4309 - val_accuracy: 0.8079\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "aDhIXBao6ZEX",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 348
        },
        "outputId": "7ebdc5a3-e1ee-4dab-ef24-a2bf923b9c64"
      },
      "source": [
        "model.summary()"
      ],
      "execution_count": 65,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_11\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "dense_24 (Dense)             (None, 128)               65664     \n",
            "_________________________________________________________________\n",
            "dropout_17 (Dropout)         (None, 128)               0         \n",
            "_________________________________________________________________\n",
            "dense_25 (Dense)             (None, 64)                8256      \n",
            "_________________________________________________________________\n",
            "dropout_18 (Dropout)         (None, 64)                0         \n",
            "_________________________________________________________________\n",
            "dense_26 (Dense)             (None, 1)                 65        \n",
            "=================================================================\n",
            "Total params: 73,985\n",
            "Trainable params: 73,985\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "2s99DAp2ber1",
        "colab_type": "text"
      },
      "source": [
        "### Plot Model Metrics"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xmNDYiR81A5y",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 297
        },
        "outputId": "ba4ee0e2-5df8-4696-f388-6bb131e41232"
      },
      "source": [
        "plt.plot(history.history['loss'], label='train loss')\n",
        "plt.plot(history.history['val_loss'], label='val loss')\n",
        "plt.xlabel(\"epoch\")\n",
        "plt.ylabel(\"accuracy\")\n",
        "plt.legend()"
      ],
      "execution_count": 71,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7f112d619908>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 71
        },
        {
          "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": "mTHBTVEoJfwa",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 297
        },
        "outputId": "07c5c4f7-e144-44fb-8014-bed0fb4ba256"
      },
      "source": [
        "plt.plot(history.history['accuracy'], label='train accuracy')\n",
        "plt.plot(history.history['val_accuracy'], label='val accuracy')\n",
        "plt.xlabel(\"epoch\")\n",
        "plt.ylabel(\"accuracy\")\n",
        "plt.legend()"
      ],
      "execution_count": 72,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7f112d526358>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 72
        },
        {
          "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": "markdown",
      "metadata": {
        "id": "_54GH2VGrQK_",
        "colab_type": "text"
      },
      "source": [
        "Evaluation on Test Data\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xiSaRp69fz60",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        },
        "outputId": "47c361fb-c480-425b-a2f6-d5952980812b"
      },
      "source": [
        "model.evaluate(X_test, y_test)"
      ],
      "execution_count": 68,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "3125/3125 [==============================] - 5s 1ms/step - loss: 0.4346 - accuracy: 0.8070\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[0.4346204698085785, 0.8069700002670288]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 68
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Ov9rkBVaI1_U",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Load the saved encoded dataset\n",
        "f3=gzip.GzipFile('/content/drive/My Drive/judgement.npy.gz','r')\n",
        "X_test = np.load(f)\n",
        "X_test = np.array(X_test)\n",
        "f4=gzip.GzipFile('/content/drive/My Drive/judgement2.npy.gz','r') # We split up our data to quickly encode it on multiple computers\n",
        "X_test2 = np.load(f)\n",
        "X_test2 = np.array(X_test2)\n",
        "# Concatentate Arrays Row-Wise\n",
        "X_test = np.concatenate((X_test,X_test2))\n",
        "# Load the original dataset\n",
        "testingdata = pd.read_csv('contestant_judgment.csv')\n",
        "# Predict sentiment from model\n",
        "pred=model.predict(X_test)\n",
        "# Write into csv\n",
        "testingdata['Sentiment']=pred\n",
        "testingdata.to_csv('predictions.csv')"
      ],
      "execution_count": 1,
      "outputs": []
    }
  ]
}