Seq2Seq_with_Attention.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Seq2Seq.ipynb",
      "provenance": [],
      "collapsed_sections": [
        "QpIdQP5hia8Q",
        "zpgivv-3zoBf",
        "U-70yAgUVxzE",
        "r99TFj4eDqD8",
        "6QKB2ccxJ_9L",
        "OwG4T9TRMvaQ",
        "ug4lJ29I_Kks",
        "fvtbnbXTXkJc",
        "MqzleT380osX",
        "lsBWWstumTyv",
        "lNEMyVQ1NvA4",
        "UUWMYuWHN22-",
        "cn4w0tV-N4MQ",
        "2M-GcQyUN6XN",
        "ktscwE43nzD0",
        "DHpt6Oa5w-BJ",
        "raVipwd0ljBO"
      ],
      "machine_shape": "hm"
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ke6uwaoIz6Az",
        "colab_type": "text"
      },
      "source": [
        "##Glove Embedding\n",
        "Embeding_D = 50"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "DqjH9ukPywr-",
        "colab_type": "text"
      },
      "source": [
        "#Seq2Seq attention"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "QpIdQP5hia8Q",
        "colab_type": "text"
      },
      "source": [
        "# Download Dataset"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "q_DmcYyHycfH",
        "colab_type": "code",
        "outputId": "ca2a36b2-5e3d-4801-9669-e576ed0ad28a",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "!curl --output dataset.zip https://archive.ics.uci.edu/ml/machine-learning-databases/00462/drugsCom_raw.zip\n",
        "!unzip dataset.zip\n",
        "# !rm dataset.zip"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current\n",
            "                                 Dload  Upload   Total   Spent    Left  Speed\n",
            "100 40.9M  100 40.9M    0     0  50.2M      0 --:--:-- --:--:-- --:--:-- 50.2M\n",
            "Archive:  dataset.zip\n",
            "  inflating: drugsComTest_raw.tsv    \n",
            "  inflating: drugsComTrain_raw.tsv   \n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "d6QBBCUnykJ8",
        "colab_type": "code",
        "outputId": "b5d14545-7faa-42b5-cfef-50a743b68ca6",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "import pandas as pd\n",
        "import html\n",
        "from nltk.tokenize import word_tokenize,RegexpTokenizer\n",
        "from nltk.corpus import stopwords \n",
        "import nltk\n",
        "nltk.download('stopwords')\n",
        "nltk.download('punkt')\n",
        "import gensim\n",
        "import numpy as np\n",
        "from keras.utils.np_utils import to_categorical\n",
        "from sklearn.model_selection import train_test_split\n",
        "from keras.preprocessing.sequence import pad_sequences\n",
        "import h5py"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[nltk_data] Downloading package stopwords to /root/nltk_data...\n",
            "[nltk_data]   Unzipping corpora/stopwords.zip.\n",
            "[nltk_data] Downloading package punkt to /root/nltk_data...\n",
            "[nltk_data]   Unzipping tokenizers/punkt.zip.\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "Using TensorFlow backend.\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<p style=\"color: red;\">\n",
              "The default version of TensorFlow in Colab will soon switch to TensorFlow 2.x.<br>\n",
              "We recommend you <a href=\"https://www.tensorflow.org/guide/migrate\" target=\"_blank\">upgrade</a> now \n",
              "or ensure your notebook will continue to use TensorFlow 1.x via the <code>%tensorflow_version 1.x</code> magic:\n",
              "<a href=\"https://colab.research.google.com/notebooks/tensorflow_version.ipynb\" target=\"_blank\">more info</a>.</p>\n"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zpgivv-3zoBf",
        "colab_type": "text"
      },
      "source": [
        "#Loading Data\n",
        "Tokenization,removing stopwords"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JlQqiCrizOjJ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Loading Data\n",
        "\n",
        "def load_dataset(filename,classes=0):\n",
        "    df = pd.read_csv(filename, sep='\\t')\n",
        "    df.drop(df.columns[0], axis=1, inplace=True)\n",
        "    df.drop(['drugName', 'condition', 'date', 'usefulCount'], axis=1, inplace=True)\n",
        "    df['review'] = df['review'].apply(lambda x: x[1:-1])\n",
        "    df['review'] = df['review'].apply(html.unescape)\n",
        "    df['rating'] = df['rating'].astype('int64')\n",
        "    df['label'] = 0\n",
        "\n",
        "    print(df.groupby('rating').size())\n",
        "\n",
        "\n",
        "    label_range = [0,4,6,10]\n",
        "    if(classes == 2):\n",
        "        label_range = [0,5,10]\n",
        "    for i,r in enumerate(label_range):\n",
        "        if(i == 0):\n",
        "            continue\n",
        "        df.label[(df['rating'] > label_range[i-1]) & (df['rating'] <=  label_range[i])] = i-1\n",
        "    return df\n",
        "\n",
        "# df = load_dataset(\"drugsComTrain_raw.tsv\")\n",
        "# df_test = load_dataset(\"drugsComTest_raw.tsv\")\n",
        "sequence_length = 100"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DH-MKgxtzwFF",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "stop_words = set(stopwords.words('english'))  \n",
        "tokenizer = RegexpTokenizer(r'\\w+')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UZP4w-Fozzb4",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def func(row):\n",
        "    l = tokenizer.tokenize(row['review'])\n",
        "    s = [i.lower() for i in l if i not in stop_words]\n",
        "    return s\n",
        "# df['processed'] = df.apply (lambda row: func(row), axis=1)\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "K7FnAub9Mzz6",
        "colab_type": "text"
      },
      "source": [
        "#Embedings"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "U-70yAgUVxzE",
        "colab_type": "text"
      },
      "source": [
        "## Glove embedings"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "r99TFj4eDqD8",
        "colab_type": "text"
      },
      "source": [
        "### Download Glove embeddings"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_ux6hL9qRGZW",
        "colab_type": "code",
        "outputId": "49e80f6e-0350-46da-f210-4becf24a62c4",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "!wget \"http://downloads.cs.stanford.edu/nlp/data/glove.6B.zip\"\n",
        "!unzip glove.6B.zip\n",
        "!rm glove.6B.zip"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--2019-11-24 16:17:36--  http://downloads.cs.stanford.edu/nlp/data/glove.6B.zip\n",
            "Resolving downloads.cs.stanford.edu (downloads.cs.stanford.edu)... 171.64.64.22\n",
            "Connecting to downloads.cs.stanford.edu (downloads.cs.stanford.edu)|171.64.64.22|:80... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 862182613 (822M) [application/zip]\n",
            "Saving to: ‘glove.6B.zip’\n",
            "\n",
            "glove.6B.zip        100%[===================>] 822.24M  1.85MB/s    in 6m 30s  \n",
            "\n",
            "2019-11-24 16:24:06 (2.11 MB/s) - ‘glove.6B.zip’ saved [862182613/862182613]\n",
            "\n",
            "Archive:  glove.6B.zip\n",
            "  inflating: glove.6B.50d.txt        \n",
            "  inflating: glove.6B.100d.txt       \n",
            "  inflating: glove.6B.200d.txt       \n",
            "  inflating: glove.6B.300d.txt       \n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6QKB2ccxJ_9L",
        "colab_type": "text"
      },
      "source": [
        "### Util functions"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9bl9-7Zsfm2j",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def embed(processed, w2v_dict, default_vec):\n",
        "    vectors = []\n",
        "    for w in processed:\n",
        "        if(w in w2v_dict):\n",
        "            vectors.append(w2v_dict[w])\n",
        "        else:\n",
        "            vectors.append(default_vec)\n",
        "    vectors = np.array(vectors)\n",
        "    return vectors\n",
        "\n",
        "def embed_using_glove(df, embed_dim=50):\n",
        "    # Load Glove embeddings.\n",
        "    GLOVE_FILE = 'glove.6B.' + str(embed_dim) + 'd.txt'\n",
        "    # Get number of vectors and hidden dimensions\n",
        "    with open(GLOVE_FILE, 'r') as f:\n",
        "        for i, line in enumerate(f):\n",
        "            pass\n",
        "    n_vec = i + 1\n",
        "    hidden_dim = len(line.split(' ')) - 1\n",
        "    # Find the average of all embeddings to be assigned to tokens not in the embeddings.\n",
        "    # Create a word to vector dictionary for embedding.\n",
        "    avg_vec = np.zeros((hidden_dim), dtype=np.float32)\n",
        "    w2v_dict = {}\n",
        "    count = 0\n",
        "    with open(GLOVE_FILE, 'r') as f:\n",
        "        for line in f:\n",
        "            w2v_dict[line.split(' ')[:1][0]] = np.array([float(n) for n in line.split(' ')[1:]], dtype=np.float32)\n",
        "            avg_vec += np.array([float(n) for n in line.split(' ')[1:]], dtype=np.float32)\n",
        "            count+=1\n",
        "    avg_vec/=count\n",
        "\n",
        "    df['vectors'] = df.apply(lambda row: embed(row['processed'], w2v_dict, avg_vec), axis=1)\n",
        "    return avg_vec"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0skbA-VnKDOc",
        "colab_type": "text"
      },
      "source": [
        "### Convert tokens in to Glove embeddings and save them to file."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rwjVr2tLJmNN",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "\n",
        "# embed_using_glove(df_train, 50)\n",
        "#df_train.to_pickle(\"train_glove_50.pckl\")\n",
        "# embed_using_glove(df_test, 50)\n",
        "# df_test.to_pickle(\"test_glove_50.pckl\")\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "heading_collapsed": true,
        "id": "OwG4T9TRMvaQ"
      },
      "source": [
        "## Word2Vec embeddings"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "heading_collapsed": true,
        "hidden": true,
        "id": "O5sARRHwM0Fl"
      },
      "source": [
        "### Download the Google word2vec embeddings"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "ExecuteTime": {
          "end_time": "2019-11-08T21:46:56.858133Z",
          "start_time": "2019-11-08T21:42:14.222728Z"
        },
        "colab_type": "code",
        "hidden": true,
        "id": "24KezgZuRbwk",
        "outputId": "6ddb23de-c9c5-429b-b66c-b29880e5f8f5",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "!curl --output word2vec.bin.gz https://s3.amazonaws.com/dl4j-distribution/GoogleNews-vectors-negative300.bin.gz\n",
        "!gunzip word2vec.bin.gz"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current\n",
            "                                 Dload  Upload   Total   Spent    Left  Speed\n",
            "100 1570M  100 1570M    0     0  44.6M      0  0:00:35  0:00:35 --:--:-- 45.8M\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "heading_collapsed": true,
        "hidden": true,
        "id": "f_Koey_WSjiY"
      },
      "source": [
        "### Load the word2vec embeddings into memory"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "ExecuteTime": {
          "end_time": "2019-11-08T21:59:19.615363Z",
          "start_time": "2019-11-08T21:58:09.951419Z"
        },
        "colab_type": "code",
        "hidden": true,
        "id": "3Iiw7NcRSqjA",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "heading_collapsed": true,
        "hidden": true,
        "id": "cNUQxRghW70T"
      },
      "source": [
        "### Util funcitons"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "ExecuteTime": {
          "end_time": "2019-11-08T21:59:33.641391Z",
          "start_time": "2019-11-08T21:59:33.634808Z"
        },
        "colab_type": "code",
        "hidden": true,
        "id": "yyQQerUzW68n",
        "colab": {}
      },
      "source": [
        "def embed_w2v(processed, w2v_dict, default_vec):\n",
        "    vectors = []\n",
        "    for w in processed:\n",
        "        if(w in w2v_dict):\n",
        "            vectors.append(w2v_dict[w])\n",
        "        else:\n",
        "            vectors.append(default_vec)\n",
        "    vectors = np.array(vectors)\n",
        "    return vectors\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "heading_collapsed": true,
        "hidden": true,
        "id": "1epAFPfIxWC-"
      },
      "source": [
        "### Get equal-class versions of train and test dataset"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pRK6atCUdLNN",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# df_test_w2v['label'].value_counts()\n",
        "# df_test_w2v"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "ExecuteTime": {
          "end_time": "2019-11-08T22:00:05.788109Z",
          "start_time": "2019-11-08T22:00:05.638070Z"
        },
        "colab_type": "code",
        "hidden": true,
        "id": "Aqw1UeLhxWDR",
        "colab": {}
      },
      "source": [
        "# df_train_w2v = pd.concat([df_train[df_train['label']==0].sample(n=4829).reset_index(drop=True),df_train[df_train['label']==1].sample(n=4829).reset_index(drop=True),df_train[df_train['label']==2].sample(n=4829).reset_index(drop=True)],ignore_index=True)\n",
        "# df_test_w2v = pd.concat([df_test[df_test['label']==0].sample(n=4829).reset_index(drop=True),df_test[df_test['label']==1].sample(n=4829).reset_index(drop=True),df_test[df_test['label']==2].sample(n=4829).reset_index(drop=True)],ignore_index=True)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "heading_collapsed": true,
        "hidden": true,
        "id": "HHE64eMvfUg9"
      },
      "source": [
        "### Convert the tokens into Word2Vec embeddings"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "ExecuteTime": {
          "end_time": "2019-11-08T22:00:32.891534Z",
          "start_time": "2019-11-08T22:00:21.978251Z"
        },
        "colab_type": "code",
        "hidden": true,
        "id": "sUw50_VeUU5c",
        "colab": {}
      },
      "source": [
        "# w2v_avg = np.average(w2v_model.vectors, axis=0)\n",
        "# df_train_w2v['vectors'] = df_train_w2v.apply(lambda row: embed_w2v(row['processed'], w2v_model, w2v_avg), axis=1)\n",
        "# df_test_w2v['vectors'] = df_test_w2v.apply(lambda row: embed_w2v(row['processed'], w2v_model, w2v_avg), axis=1)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "heading_collapsed": true,
        "hidden": true,
        "id": "unLYkdPwYLtR",
        "colab_type": "text"
      },
      "source": [
        "### Save the Embeddings to file"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "ExecuteTime": {
          "end_time": "2019-11-08T22:05:18.963018Z",
          "start_time": "2019-11-08T22:04:14.983253Z"
        },
        "colab_type": "code",
        "hidden": true,
        "id": "LH6rJMjryvzC",
        "colab": {}
      },
      "source": [
        "# df_train_w2v.to_pickle(\"train_w2v.pckl\")\n",
        "# df_test_w2v.to_pickle(\"test_w2v.pckl\")"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3ND-rucOSrUR",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def embeding_using_w2v(df):\n",
        "    from gensim.models import KeyedVectors\n",
        "    w2v_model = KeyedVectors.load_word2vec_format('word2vec.bin', binary=True)\n",
        "    w2v_avg = np.average(w2v_model.vectors, axis=0)\n",
        "\n",
        "    df['vectors'] = df.apply(lambda row: embed_w2v(row['processed'], w2v_model, w2v_avg), axis=1)\n",
        "    df['vectors'] = df.apply(lambda row: embed_w2v(row['processed'], w2v_model, w2v_avg), axis=1)    \n",
        "    return w2v_avg"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ug4lJ29I_Kks",
        "colab_type": "text"
      },
      "source": [
        "## Elmo embeddings"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fvtbnbXTXkJc",
        "colab_type": "text"
      },
      "source": [
        "### Install allennlp"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wX4DiLtJXnxI",
        "colab_type": "code",
        "outputId": "283cc7f8-99dc-4568-dc8b-e6f032f02a7a",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "!pip install allennlp"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Collecting allennlp\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/bb/bb/041115d8bad1447080e5d1e30097c95e4b66e36074277afce8620a61cee3/allennlp-0.9.0-py3-none-any.whl (7.6MB)\n",
            "\u001b[K     |████████████████████████████████| 7.6MB 4.7MB/s \n",
            "\u001b[?25hRequirement already satisfied: nltk in /usr/local/lib/python3.6/dist-packages (from allennlp) (3.2.5)\n",
            "Collecting jsonpickle\n",
            "  Downloading https://files.pythonhosted.org/packages/07/07/c157520a3ebd166c8c24c6ae0ecae7c3968eb4653ff0e5af369bb82f004d/jsonpickle-1.2-py2.py3-none-any.whl\n",
            "Requirement already satisfied: editdistance in /usr/local/lib/python3.6/dist-packages (from allennlp) (0.5.3)\n",
            "Requirement already satisfied: scipy in /usr/local/lib/python3.6/dist-packages (from allennlp) (1.3.2)\n",
            "Requirement already satisfied: pytz>=2017.3 in /usr/local/lib/python3.6/dist-packages (from allennlp) (2018.9)\n",
            "Requirement already satisfied: flask>=1.0.2 in /usr/local/lib/python3.6/dist-packages (from allennlp) (1.1.1)\n",
            "Collecting word2number>=1.1\n",
            "  Downloading https://files.pythonhosted.org/packages/4a/29/a31940c848521f0725f0df6b25dca8917f13a2025b0e8fcbe5d0457e45e6/word2number-1.1.zip\n",
            "Collecting ftfy\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/75/ca/2d9a5030eaf1bcd925dab392762b9709a7ad4bd486a90599d93cd79cb188/ftfy-5.6.tar.gz (58kB)\n",
            "\u001b[K     |████████████████████████████████| 61kB 7.9MB/s \n",
            "\u001b[?25hCollecting jsonnet>=0.10.0; sys_platform != \"win32\"\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/fe/a6/e69e38f1f259fcf8532d8bd2c4bc88764f42d7b35a41423a7f4b035cc5ce/jsonnet-0.14.0.tar.gz (253kB)\n",
            "\u001b[K     |████████████████████████████████| 256kB 50.2MB/s \n",
            "\u001b[?25hRequirement already satisfied: tqdm>=4.19 in /usr/local/lib/python3.6/dist-packages (from allennlp) (4.28.1)\n",
            "Collecting numpydoc>=0.8.0\n",
            "  Downloading https://files.pythonhosted.org/packages/6a/f3/7cfe4c616e4b9fe05540256cc9c6661c052c8a4cec2915732793b36e1843/numpydoc-0.9.1.tar.gz\n",
            "Requirement already satisfied: matplotlib>=2.2.3 in /usr/local/lib/python3.6/dist-packages (from allennlp) (3.1.1)\n",
            "Requirement already satisfied: boto3 in /usr/local/lib/python3.6/dist-packages (from allennlp) (1.10.18)\n",
            "Collecting overrides\n",
            "  Downloading https://files.pythonhosted.org/packages/8b/a0/0d4a9dcd28de9bf263e9436b5876ba199cc4e4a19d0dc4e85eed657b1b77/overrides-2.5.tar.gz\n",
            "Requirement already satisfied: spacy<2.2,>=2.1.0 in /usr/local/lib/python3.6/dist-packages (from allennlp) (2.1.9)\n",
            "Requirement already satisfied: h5py in /usr/local/lib/python3.6/dist-packages (from allennlp) (2.8.0)\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.6/dist-packages (from allennlp) (1.17.4)\n",
            "Requirement already satisfied: scikit-learn in /usr/local/lib/python3.6/dist-packages (from allennlp) (0.21.3)\n",
            "Collecting conllu==1.3.1\n",
            "  Downloading https://files.pythonhosted.org/packages/ae/54/b0ae1199f3d01666821b028cd967f7c0ac527ab162af433d3da69242cea2/conllu-1.3.1-py2.py3-none-any.whl\n",
            "Collecting parsimonious>=0.8.0\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/02/fc/067a3f89869a41009e1a7cdfb14725f8ddd246f30f63c645e8ef8a1c56f4/parsimonious-0.8.1.tar.gz (45kB)\n",
            "\u001b[K     |████████████████████████████████| 51kB 7.4MB/s \n",
            "\u001b[?25hCollecting flaky\n",
            "  Downloading https://files.pythonhosted.org/packages/fe/12/0f169abf1aa07c7edef4855cca53703d2e6b7ecbded7829588ac7e7e3424/flaky-3.6.1-py2.py3-none-any.whl\n",
            "Requirement already satisfied: requests>=2.18 in /usr/local/lib/python3.6/dist-packages (from allennlp) (2.21.0)\n",
            "Requirement already satisfied: pytest in /usr/local/lib/python3.6/dist-packages (from allennlp) (3.6.4)\n",
            "Requirement already satisfied: torch>=1.2.0 in /usr/local/lib/python3.6/dist-packages (from allennlp) (1.3.1)\n",
            "Collecting tensorboardX>=1.2\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/a6/5c/e918d9f190baab8d55bad52840d8091dd5114cc99f03eaa6d72d404503cc/tensorboardX-1.9-py2.py3-none-any.whl (190kB)\n",
            "\u001b[K     |████████████████████████████████| 194kB 39.0MB/s \n",
            "\u001b[?25hRequirement already satisfied: gevent>=1.3.6 in /usr/local/lib/python3.6/dist-packages (from allennlp) (1.4.0)\n",
            "Requirement already satisfied: sqlparse>=0.2.4 in /usr/local/lib/python3.6/dist-packages (from allennlp) (0.3.0)\n",
            "Collecting responses>=0.7\n",
            "  Downloading https://files.pythonhosted.org/packages/d1/5a/b887e89925f1de7890ef298a74438371ed4ed29b33def9e6d02dc6036fd8/responses-0.10.6-py2.py3-none-any.whl\n",
            "Collecting flask-cors>=3.0.7\n",
            "  Downloading https://files.pythonhosted.org/packages/78/38/e68b11daa5d613e3a91e4bf3da76c94ac9ee0d9cd515af9c1ab80d36f709/Flask_Cors-3.0.8-py2.py3-none-any.whl\n",
            "Collecting pytorch-pretrained-bert>=0.6.0\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/d7/e0/c08d5553b89973d9a240605b9c12404bcf8227590de62bae27acbcfe076b/pytorch_pretrained_bert-0.6.2-py3-none-any.whl (123kB)\n",
            "\u001b[K     |████████████████████████████████| 133kB 50.8MB/s \n",
            "\u001b[?25hCollecting pytorch-transformers==1.1.0\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/50/89/ad0d6bb932d0a51793eaabcf1617a36ff530dc9ab9e38f765a35dc293306/pytorch_transformers-1.1.0-py3-none-any.whl (158kB)\n",
            "\u001b[K     |████████████████████████████████| 163kB 55.0MB/s \n",
            "\u001b[?25hCollecting unidecode\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/d0/42/d9edfed04228bacea2d824904cae367ee9efd05e6cce7ceaaedd0b0ad964/Unidecode-1.1.1-py2.py3-none-any.whl (238kB)\n",
            "\u001b[K     |████████████████████████████████| 245kB 50.5MB/s \n",
            "\u001b[?25hRequirement already satisfied: six in /usr/local/lib/python3.6/dist-packages (from nltk->allennlp) (1.12.0)\n",
            "Requirement already satisfied: Jinja2>=2.10.1 in /usr/local/lib/python3.6/dist-packages (from flask>=1.0.2->allennlp) (2.10.3)\n",
            "Requirement already satisfied: Werkzeug>=0.15 in /usr/local/lib/python3.6/dist-packages (from flask>=1.0.2->allennlp) (0.16.0)\n",
            "Requirement already satisfied: itsdangerous>=0.24 in /usr/local/lib/python3.6/dist-packages (from flask>=1.0.2->allennlp) (1.1.0)\n",
            "Requirement already satisfied: click>=5.1 in /usr/local/lib/python3.6/dist-packages (from flask>=1.0.2->allennlp) (7.0)\n",
            "Requirement already satisfied: wcwidth in /usr/local/lib/python3.6/dist-packages (from ftfy->allennlp) (0.1.7)\n",
            "Requirement already satisfied: sphinx>=1.6.5 in /usr/local/lib/python3.6/dist-packages (from numpydoc>=0.8.0->allennlp) (1.8.5)\n",
            "Requirement already satisfied: python-dateutil>=2.1 in /usr/local/lib/python3.6/dist-packages (from matplotlib>=2.2.3->allennlp) (2.6.1)\n",
            "Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.6/dist-packages (from matplotlib>=2.2.3->allennlp) (0.10.0)\n",
            "Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.6/dist-packages (from matplotlib>=2.2.3->allennlp) (1.1.0)\n",
            "Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/local/lib/python3.6/dist-packages (from matplotlib>=2.2.3->allennlp) (2.4.5)\n",
            "Requirement already satisfied: jmespath<1.0.0,>=0.7.1 in /usr/local/lib/python3.6/dist-packages (from boto3->allennlp) (0.9.4)\n",
            "Requirement already satisfied: botocore<1.14.0,>=1.13.18 in /usr/local/lib/python3.6/dist-packages (from boto3->allennlp) (1.13.18)\n",
            "Requirement already satisfied: s3transfer<0.3.0,>=0.2.0 in /usr/local/lib/python3.6/dist-packages (from boto3->allennlp) (0.2.1)\n",
            "Requirement already satisfied: srsly<1.1.0,>=0.0.6 in /usr/local/lib/python3.6/dist-packages (from spacy<2.2,>=2.1.0->allennlp) (0.2.0)\n",
            "Requirement already satisfied: plac<1.0.0,>=0.9.6 in /usr/local/lib/python3.6/dist-packages (from spacy<2.2,>=2.1.0->allennlp) (0.9.6)\n",
            "Requirement already satisfied: preshed<2.1.0,>=2.0.1 in /usr/local/lib/python3.6/dist-packages (from spacy<2.2,>=2.1.0->allennlp) (2.0.1)\n",
            "Requirement already satisfied: blis<0.3.0,>=0.2.2 in /usr/local/lib/python3.6/dist-packages (from spacy<2.2,>=2.1.0->allennlp) (0.2.4)\n",
            "Requirement already satisfied: murmurhash<1.1.0,>=0.28.0 in /usr/local/lib/python3.6/dist-packages (from spacy<2.2,>=2.1.0->allennlp) (1.0.2)\n",
            "Requirement already satisfied: cymem<2.1.0,>=2.0.2 in /usr/local/lib/python3.6/dist-packages (from spacy<2.2,>=2.1.0->allennlp) (2.0.3)\n",
            "Requirement already satisfied: thinc<7.1.0,>=7.0.8 in /usr/local/lib/python3.6/dist-packages (from spacy<2.2,>=2.1.0->allennlp) (7.0.8)\n",
            "Requirement already satisfied: wasabi<1.1.0,>=0.2.0 in /usr/local/lib/python3.6/dist-packages (from spacy<2.2,>=2.1.0->allennlp) (0.4.0)\n",
            "Requirement already satisfied: joblib>=0.11 in /usr/local/lib/python3.6/dist-packages (from scikit-learn->allennlp) (0.14.0)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.6/dist-packages (from requests>=2.18->allennlp) (2019.9.11)\n",
            "Requirement already satisfied: urllib3<1.25,>=1.21.1 in /usr/local/lib/python3.6/dist-packages (from requests>=2.18->allennlp) (1.24.3)\n",
            "Requirement already satisfied: idna<2.9,>=2.5 in /usr/local/lib/python3.6/dist-packages (from requests>=2.18->allennlp) (2.8)\n",
            "Requirement already satisfied: chardet<3.1.0,>=3.0.2 in /usr/local/lib/python3.6/dist-packages (from requests>=2.18->allennlp) (3.0.4)\n",
            "Requirement already satisfied: attrs>=17.4.0 in /usr/local/lib/python3.6/dist-packages (from pytest->allennlp) (19.3.0)\n",
            "Requirement already satisfied: pluggy<0.8,>=0.5 in /usr/local/lib/python3.6/dist-packages (from pytest->allennlp) (0.7.1)\n",
            "Requirement already satisfied: py>=1.5.0 in /usr/local/lib/python3.6/dist-packages (from pytest->allennlp) (1.8.0)\n",
            "Requirement already satisfied: setuptools in /usr/local/lib/python3.6/dist-packages (from pytest->allennlp) (41.6.0)\n",
            "Requirement already satisfied: atomicwrites>=1.0 in /usr/local/lib/python3.6/dist-packages (from pytest->allennlp) (1.3.0)\n",
            "Requirement already satisfied: more-itertools>=4.0.0 in /usr/local/lib/python3.6/dist-packages (from pytest->allennlp) (7.2.0)\n",
            "Requirement already satisfied: protobuf>=3.8.0 in /usr/local/lib/python3.6/dist-packages (from tensorboardX>=1.2->allennlp) (3.10.0)\n",
            "Requirement already satisfied: greenlet>=0.4.14; platform_python_implementation == \"CPython\" in /usr/local/lib/python3.6/dist-packages (from gevent>=1.3.6->allennlp) (0.4.15)\n",
            "Collecting regex\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/e3/8e/cbf2295643d7265e7883326fb4654e643bfc93b3a8a8274d8010a39d8804/regex-2019.11.1-cp36-cp36m-manylinux1_x86_64.whl (643kB)\n",
            "\u001b[K     |████████████████████████████████| 645kB 42.3MB/s \n",
            "\u001b[?25hCollecting sentencepiece\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/14/3d/efb655a670b98f62ec32d66954e1109f403db4d937c50d779a75b9763a29/sentencepiece-0.1.83-cp36-cp36m-manylinux1_x86_64.whl (1.0MB)\n",
            "\u001b[K     |████████████████████████████████| 1.0MB 37.1MB/s \n",
            "\u001b[?25hRequirement already satisfied: MarkupSafe>=0.23 in /usr/local/lib/python3.6/dist-packages (from Jinja2>=2.10.1->flask>=1.0.2->allennlp) (1.1.1)\n",
            "Requirement already satisfied: snowballstemmer>=1.1 in /usr/local/lib/python3.6/dist-packages (from sphinx>=1.6.5->numpydoc>=0.8.0->allennlp) (2.0.0)\n",
            "Requirement already satisfied: Pygments>=2.0 in /usr/local/lib/python3.6/dist-packages (from sphinx>=1.6.5->numpydoc>=0.8.0->allennlp) (2.1.3)\n",
            "Requirement already satisfied: babel!=2.0,>=1.3 in /usr/local/lib/python3.6/dist-packages (from sphinx>=1.6.5->numpydoc>=0.8.0->allennlp) (2.7.0)\n",
            "Requirement already satisfied: alabaster<0.8,>=0.7 in /usr/local/lib/python3.6/dist-packages (from sphinx>=1.6.5->numpydoc>=0.8.0->allennlp) (0.7.12)\n",
            "Requirement already satisfied: imagesize in /usr/local/lib/python3.6/dist-packages (from sphinx>=1.6.5->numpydoc>=0.8.0->allennlp) (1.1.0)\n",
            "Requirement already satisfied: packaging in /usr/local/lib/python3.6/dist-packages (from sphinx>=1.6.5->numpydoc>=0.8.0->allennlp) (19.2)\n",
            "Requirement already satisfied: sphinxcontrib-websupport in /usr/local/lib/python3.6/dist-packages (from sphinx>=1.6.5->numpydoc>=0.8.0->allennlp) (1.1.2)\n",
            "Requirement already satisfied: docutils>=0.11 in /usr/local/lib/python3.6/dist-packages (from sphinx>=1.6.5->numpydoc>=0.8.0->allennlp) (0.15.2)\n",
            "Building wheels for collected packages: word2number, ftfy, jsonnet, numpydoc, overrides, parsimonious\n",
            "  Building wheel for word2number (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for word2number: filename=word2number-1.1-cp36-none-any.whl size=5588 sha256=e441dd3f106a9705b2a37836ee4194672275a95da52c307372442d69352f639c\n",
            "  Stored in directory: /root/.cache/pip/wheels/46/2f/53/5f5c1d275492f2fce1cdab9a9bb12d49286dead829a4078e0e\n",
            "  Building wheel for ftfy (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for ftfy: filename=ftfy-5.6-cp36-none-any.whl size=44553 sha256=969cfc2dea1ef7dd1b247c4bbde7d45dcc21e858e1e7b004deb9bf8de04a418b\n",
            "  Stored in directory: /root/.cache/pip/wheels/43/34/ce/cbb38d71543c408de56f3c5e26ce8ba495a0fa5a28eaaf1046\n",
            "  Building wheel for jsonnet (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for jsonnet: filename=jsonnet-0.14.0-cp36-cp36m-linux_x86_64.whl size=3320385 sha256=f40b1de696f171b4a846d2bf619dd72306772a385d01c51997ad09ff3a911807\n",
            "  Stored in directory: /root/.cache/pip/wheels/5b/b7/83/985f0f758fbb34f14989a0fab86d18890d1cc5ae12f26967bc\n",
            "  Building wheel for numpydoc (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for numpydoc: filename=numpydoc-0.9.1-cp36-none-any.whl size=31872 sha256=28508aff468fb6bb12c9a48cd83e115afc365c5651d74a0fbfae0ffe184db00d\n",
            "  Stored in directory: /root/.cache/pip/wheels/51/30/d1/92a39ba40f21cb70e53f8af96eb98f002a781843c065406500\n",
            "  Building wheel for overrides (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for overrides: filename=overrides-2.5-cp36-none-any.whl size=5407 sha256=51a858ecb333b3594f77bfde8c3e43e91bb96e12d1d29c4a37711eb597c883b0\n",
            "  Stored in directory: /root/.cache/pip/wheels/be/56/0c/f5ce8e2403ae7d881e5d638fca1c1c5d21dac8eb6df5a6950f\n",
            "  Building wheel for parsimonious (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for parsimonious: filename=parsimonious-0.8.1-cp36-none-any.whl size=42709 sha256=d33766a84bd205037ff74781e5c53277009339a94c386dd131586b45bbb472bd\n",
            "  Stored in directory: /root/.cache/pip/wheels/b7/8d/e7/a0e74217da5caeb3c1c7689639b6d28ddbf9985b840bc96a9a\n",
            "Successfully built word2number ftfy jsonnet numpydoc overrides parsimonious\n",
            "Installing collected packages: jsonpickle, word2number, ftfy, jsonnet, numpydoc, overrides, conllu, parsimonious, flaky, tensorboardX, responses, flask-cors, regex, pytorch-pretrained-bert, sentencepiece, pytorch-transformers, unidecode, allennlp\n",
            "Successfully installed allennlp-0.9.0 conllu-1.3.1 flaky-3.6.1 flask-cors-3.0.8 ftfy-5.6 jsonnet-0.14.0 jsonpickle-1.2 numpydoc-0.9.1 overrides-2.5 parsimonious-0.8.1 pytorch-pretrained-bert-0.6.2 pytorch-transformers-1.1.0 regex-2019.11.1 responses-0.10.6 sentencepiece-0.1.83 tensorboardX-1.9 unidecode-1.1.1 word2number-1.1\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Hv2O9HmcoSV7",
        "colab_type": "text"
      },
      "source": [
        "### Imports"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LhYYiddBjbfK",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from allennlp.commands.elmo import ElmoEmbedder\n",
        "import sys"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vne2Ej3koUES",
        "colab_type": "text"
      },
      "source": [
        "### Util functions"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uhMhSy3xXoeN",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "avg_elmo_embedding = ['1']\n",
        "def create_elmo_embeddings(elmo, documents, max_sentences = 1000):\n",
        "    num_sentences = min(max_sentences, len(documents)) if max_sentences > 0 else len(documents)\n",
        "    print(\"\\n\\n:: Lookup of \" + str(num_sentences) + \" ELMo representations. This takes a while ::\")\n",
        "    embeddings = []\n",
        "    tokens = documents['processed'].to_numpy()\n",
        "    \n",
        "    documentIdx = 0\n",
        "    for elmo_embedding in elmo.embed_sentences(tokens):  \n",
        "        document = documents.iloc[documentIdx]\n",
        "        # Average the 3 layers returned from ELMo\n",
        "        avg_elmo_embedding = np.average(elmo_embedding, axis=0)             \n",
        "        embeddings.append(avg_elmo_embedding)        \n",
        "        \n",
        "        # Some progress info\n",
        "        documentIdx += 1\n",
        "        percent = 100.0 * documentIdx / num_sentences\n",
        "        line = '[{0}{1}]'.format('=' * int(percent / 2), ' ' * (50 - int(percent / 2)))\n",
        "        status = '\\r{0:3.0f}%{1} {2:3d}/{3:3d} sentences'\n",
        "        sys.stdout.write(status.format(percent, line, documentIdx, num_sentences))\n",
        "        \n",
        "        if max_sentences > 0 and documentIdx >= max_sentences:\n",
        "            break\n",
        "            \n",
        "    return embeddings\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "qmrAEErXwpDA"
      },
      "source": [
        "### Get smaller versions of train and test dataset"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6dzCGQ9MZN7u",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# df_train_elmo = df_train.sample(n=int(df_train.shape[0]/20)).reset_index(drop=True)\n",
        "# df_test_elmo = df_test.sample(n=int(df_test.shape[0]/20)).reset_index(drop=True)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "uqmLae0uoXe7",
        "colab_type": "text"
      },
      "source": [
        "### Get ELMo embeddings"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2frbjy9yokoS",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# elmo = ElmoEmbedder(cuda_device=0) #Set cuda_device to the ID of your GPU if you have one\n",
        "# df_train_elmo['elmo_embeddings'] = create_elmo_embeddings(elmo, df_train_elmo, -1)\n",
        "# df_test_elmo['elmo_embeddings'] = create_elmo_embeddings(elmo, df_test_elmo, -1)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eUw1U_vTUi7A",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def embeding_using_elmo(df):\n",
        "    elmo = ElmoEmbedder(cuda_device=0) #Set cuda_device to the ID of your GPU if you have one\n",
        "    df['vectors'] = create_elmo_embeddings(elmo, df, -1)\n",
        "    return np.full_like(range(1024),0.0)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MqzleT380osX",
        "colab_type": "text"
      },
      "source": [
        "#Padding\n",
        "Sequence length = 100"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fssYrFmO0yVX",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "sequence_length = 100\n",
        "num_classes = 3"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4_q_6kQhBUfy",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def v2w(vec,avg_vec,w2v_dict):\n",
        "    if((vec==avg_vec).all()):\n",
        "        return 'UNK'\n",
        "    if(np.all(vec == 0)):\n",
        "        return 'None'\n",
        "    for k in w2v_dict:\n",
        "        if((w2v_dict[k]==vec).all()):\n",
        "            return k\n",
        "    return 'OOV'"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "lsBWWstumTyv",
        "colab_type": "text"
      },
      "source": [
        "# GetData"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "r83PZ5gIltcI",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def getData(normalize = False,k = -1,k_test = 1000,classes=3,embed = 'glove'):\n",
        "    df_train = load_dataset(\"drugsComTrain_raw.tsv\",classes)\n",
        "    df_test = load_dataset(\"drugsComTest_raw.tsv\",classes)\n",
        "    \n",
        "    print(df_train.groupby('label').size())\n",
        "    print(df_test.groupby('label').size())\n",
        "\n",
        "\n",
        "    if(normalize):\n",
        "        mn = df_train.groupby('label').size().min()\n",
        "        if(classes == 3):\n",
        "            df_train = pd.concat([df_train[df_train['label']==0].sample(n=mn).reset_index(drop=True),df_train[df_train['label']==1].sample(n=mn).reset_index(drop=True),df_train[df_train['label']==2].sample(n=mn).reset_index(drop=True)],ignore_index=True)\n",
        "        else:\n",
        "            df_train = pd.concat([df_train[df_train['label']==0].sample(n=mn).reset_index(drop=True),df_train[df_train['label']==1].sample(n=mn).reset_index(drop=True)],ignore_index=True)\n",
        "        mn = df_test.groupby('label').size().min()\n",
        "        if(classes == 3):\n",
        "            df_test = pd.concat([df_test[df_test['label']==0].sample(n=mn).reset_index(drop=True),df_test[df_test['label']==1].sample(n=mn).reset_index(drop=True),df_test[df_test['label']==2].sample(n=mn).reset_index(drop=True)],ignore_index=True)\n",
        "        else:\n",
        "            df_test = pd.concat([df_test[df_test['label']==0].sample(n=mn).reset_index(drop=True),df_test[df_test['label']==1].sample(n=mn).reset_index(drop=True)],ignore_index=True)\n",
        "\n",
        "    if(k!=-1):\n",
        "        df_train = df_train.sample(n=k).reset_index(drop=True)\n",
        "\n",
        "    if(k_test!=-1):\n",
        "        df_test = df_test.sample(n=k_test).reset_index(drop=True)\n",
        "\n",
        "    df_train['processed'] = df_train.apply (lambda row: func(row), axis=1)\n",
        "    df_test['processed'] = df_test.apply (lambda row: func(row), axis=1)\n",
        "\n",
        "    if(embed == 'glove'):\n",
        "        w2v_avg = embed_using_glove(df_train, 50)\n",
        "        w2v_avg = embed_using_glove(df_test, 50)\n",
        "\n",
        "    if(embed == 'w2v'):\n",
        "        w2v_avg = embeding_using_w2v(df_train)\n",
        "        w2v_avg = embeding_using_w2v(df_test)\n",
        "\n",
        "    if(embed == 'elmo'):\n",
        "        w2v_avg = embeding_using_elmo(df_train)\n",
        "        w2v_avg = embeding_using_elmo(df_test)\n",
        "\n",
        "    # df_train['vectors'] = df_train.apply (lambda row: foo(row['processed']), axis=1)\n",
        "    # df_test['vectors'] = df_test.apply (lambda row: foo(row['processed']), axis=1)\n",
        "\n",
        "    train_data = pad_sequences(df_train['vectors'], maxlen=sequence_length,dtype='float32')\n",
        "    train_labels = pd.get_dummies(df_train['label']).to_numpy().reshape(-1,1,classes)\n",
        "\n",
        "    test_data = pad_sequences(df_test['vectors'], maxlen=sequence_length,dtype='float32')\n",
        "    test_labels = pd.get_dummies(df_test['label']).to_numpy().reshape(-1,1,classes)\n",
        "\n",
        "    _GO=np.array([[range(classes)]]*train_labels.shape[0])\n",
        "    _GO_test=np.array([[range(classes)]]*test_labels.shape[0])\n",
        "    _GO.shape\n",
        "\n",
        "    print(df_train.groupby('label').size())\n",
        "    print(df_test.groupby('label').size())\n",
        "\n",
        "    xTrain = {'ex':train_data,'dx':_GO}\n",
        "    xTest = {'ex':test_data,'dx':_GO_test}\n",
        "    print(train_data.shape,test_data.shape)\n",
        "\n",
        "    return xTrain,xTest,train_labels,test_labels"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "C_s9YFvS8-jQ",
        "colab_type": "text"
      },
      "source": [
        "#Model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "GX_GSCLr89aa",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from tensorflow.keras.layers import Input,GRU,Attention,Bidirectional,Concatenate,RepeatVector,dot,Activation,concatenate,Dense,TimeDistributed\n",
        "from tensorflow.keras.models import Model"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QlUM3IIkiFUQ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def seq2seq(embed,classes=3):\n",
        "\n",
        "    if(embed == 'glove'):\n",
        "        EMBEDDING_D = 50\n",
        "    if(embed == 'w2v'):\n",
        "        EMBEDDING_D = 300\n",
        "    if(embed == 'elmo'):\n",
        "        EMBEDDING_D = 1024\n",
        "\n",
        "    encoder_inputs = Input(batch_shape=(None, sequence_length, EMBEDDING_D), name='ex')\n",
        "    decoder_inputs = Input(batch_shape=(None, 1, classes), name='dx')\n",
        "\n",
        "    # Encoder network\n",
        "    encoder_gru = Bidirectional(GRU(64, return_sequences=True, return_state=True, name='encoder_gru'),merge_mode='concat')\n",
        "    encoder_out, state_h,state_c = encoder_gru(encoder_inputs)\n",
        "    encoder_state = [state_h,state_c]\n",
        "    print(encoder_out.shape)\n",
        "    print(state_h.shape)\n",
        "    print(state_c.shape)\n",
        "\n",
        "\n",
        "    # Decoder network\n",
        "    decoder_gru = Bidirectional(GRU(64, return_sequences=True, return_state=True, name='decoder_gru'),merge_mode='concat')\n",
        "    decoder_out, _,_ = decoder_gru(decoder_inputs, initial_state=encoder_state)\n",
        "\n",
        "    print(decoder_out.shape)\n",
        "\n",
        "    #Attention Layer\n",
        "    attn_out = dot([decoder_out, encoder_out], axes=[2, 2])\n",
        "\n",
        "    print(attn_out.shape)\n",
        "\n",
        "    attention = Activation('softmax')(attn_out)\n",
        "\n",
        "    context = dot([attention, encoder_out], axes=[2,1])\n",
        "\n",
        "    decoder_concat_input = concatenate([context, decoder_out])\n",
        "\n",
        "    # decoder_concat_input = Concatenate(axis=-1, name='concat_layer')([decoder_out, attn_out])\n",
        "\n",
        "    dense = Dense(classes, activation='softmax', name='softmax_layer')\n",
        "    dense_time = TimeDistributed(dense, name='time_distributed_layer')\n",
        "    decoder_pred = dense_time(decoder_concat_input)\n",
        "\n",
        "    full_model = Model(inputs=[encoder_inputs, decoder_inputs], outputs=decoder_pred)\n",
        "    full_model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])\n",
        "    attention_model = Model(inputs=[encoder_inputs, decoder_inputs], outputs=attention)\n",
        "    attention_model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])\n",
        "\n",
        "    return full_model,attention_model"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "lNEMyVQ1NvA4",
        "colab_type": "text"
      },
      "source": [
        "#Training"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "UUWMYuWHN22-",
        "colab_type": "text"
      },
      "source": [
        "##Glove"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KyjWOPLOXvp4",
        "colab_type": "code",
        "outputId": "3a120e75-dcbb-48d2-b897-71a0b2eea106",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 975
        }
      },
      "source": [
        "xTrain,xVal,yTrain,yVal = getData(normalize=True,k=10000)\n",
        "full_model,attention_model = seq2seq(embed='glove')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "rating\n",
            "1     21619\n",
            "2      6931\n",
            "3      6513\n",
            "4      5012\n",
            "5      8013\n",
            "6      6343\n",
            "7      9456\n",
            "8     18890\n",
            "9     27531\n",
            "10    50989\n",
            "dtype: int64\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:20: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame\n",
            "\n",
            "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "rating\n",
            "1      7299\n",
            "2      2334\n",
            "3      2205\n",
            "4      1659\n",
            "5      2710\n",
            "6      2119\n",
            "7      3091\n",
            "8      6156\n",
            "9      9177\n",
            "10    17016\n",
            "dtype: int64\n",
            "label\n",
            "0     40075\n",
            "1     14356\n",
            "2    106866\n",
            "dtype: int64\n",
            "label\n",
            "0    13497\n",
            "1     4829\n",
            "2    35440\n",
            "dtype: int64\n",
            "label\n",
            "0    3302\n",
            "1    3363\n",
            "2    3335\n",
            "dtype: int64\n",
            "label\n",
            "0    331\n",
            "1    352\n",
            "2    317\n",
            "dtype: int64\n",
            "(10000, 100, 50) (1000, 100, 50)\n",
            "(?, 100, 128)\n",
            "(?, 64)\n",
            "(?, 64)\n",
            "(?, 1, 128)\n",
            "(?, 1, 100)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6a-LYVtJ8MF8",
        "colab_type": "code",
        "outputId": "241c3c6c-80ac-451f-9962-276117f84611",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 555
        }
      },
      "source": [
        "training = full_model.fit(x=xTrain, y = yTrain, epochs=15, batch_size=128,validation_split=0.1)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 9000 samples, validate on 1000 samples\n",
            "Epoch 1/15\n",
            "9000/9000 [==============================] - 17s 2ms/sample - loss: 1.0831 - acc: 0.3940 - val_loss: 1.0522 - val_acc: 0.4430\n",
            "Epoch 2/15\n",
            "9000/9000 [==============================] - 15s 2ms/sample - loss: 1.0287 - acc: 0.4710 - val_loss: 1.0022 - val_acc: 0.5060\n",
            "Epoch 3/15\n",
            "9000/9000 [==============================] - 15s 2ms/sample - loss: 0.9976 - acc: 0.4957 - val_loss: 0.9829 - val_acc: 0.5100\n",
            "Epoch 4/15\n",
            "9000/9000 [==============================] - 15s 2ms/sample - loss: 0.9677 - acc: 0.5167 - val_loss: 0.9677 - val_acc: 0.5160\n",
            "Epoch 5/15\n",
            "9000/9000 [==============================] - 15s 2ms/sample - loss: 0.9525 - acc: 0.5259 - val_loss: 0.9544 - val_acc: 0.5270\n",
            "Epoch 6/15\n",
            "9000/9000 [==============================] - 15s 2ms/sample - loss: 0.9205 - acc: 0.5546 - val_loss: 0.9620 - val_acc: 0.5240\n",
            "Epoch 7/15\n",
            "9000/9000 [==============================] - 15s 2ms/sample - loss: 0.8983 - acc: 0.5723 - val_loss: 0.9309 - val_acc: 0.5260\n",
            "Epoch 8/15\n",
            "9000/9000 [==============================] - 15s 2ms/sample - loss: 0.8713 - acc: 0.5876 - val_loss: 0.9648 - val_acc: 0.5310\n",
            "Epoch 9/15\n",
            "9000/9000 [==============================] - 15s 2ms/sample - loss: 0.8581 - acc: 0.5959 - val_loss: 0.9639 - val_acc: 0.5310\n",
            "Epoch 10/15\n",
            "9000/9000 [==============================] - 15s 2ms/sample - loss: 0.8234 - acc: 0.6186 - val_loss: 0.9571 - val_acc: 0.5460\n",
            "Epoch 11/15\n",
            "9000/9000 [==============================] - 15s 2ms/sample - loss: 0.8028 - acc: 0.6274 - val_loss: 0.9455 - val_acc: 0.5430\n",
            "Epoch 12/15\n",
            "9000/9000 [==============================] - 15s 2ms/sample - loss: 0.7710 - acc: 0.6483 - val_loss: 0.9523 - val_acc: 0.5540\n",
            "Epoch 13/15\n",
            "9000/9000 [==============================] - 15s 2ms/sample - loss: 0.7381 - acc: 0.6646 - val_loss: 1.0163 - val_acc: 0.5400\n",
            "Epoch 14/15\n",
            "9000/9000 [==============================] - 15s 2ms/sample - loss: 0.7176 - acc: 0.6760 - val_loss: 0.9811 - val_acc: 0.5350\n",
            "Epoch 15/15\n",
            "9000/9000 [==============================] - 15s 2ms/sample - loss: 0.6691 - acc: 0.7031 - val_loss: 1.0087 - val_acc: 0.5440\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "t3t7kBjWvSMA",
        "colab_type": "code",
        "outputId": "6af77455-b4f1-47fa-c2d6-092087fe76c8",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 191
        }
      },
      "source": [
        "from sklearn.metrics import classification_report\n",
        "import numpy as np\n",
        "\n",
        "Y_test = np.argmax(yVal, axis=2).reshape((-1)) # Convert one-hot to index\n",
        "y_pred = full_model.predict(xVal)\n",
        "y_pred = y_pred.argmax(axis=2).reshape((-1))\n",
        "print(classification_report(Y_test, y_pred))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.54      0.55      0.55       331\n",
            "           1       0.49      0.41      0.45       352\n",
            "           2       0.56      0.65      0.60       317\n",
            "\n",
            "    accuracy                           0.53      1000\n",
            "   macro avg       0.53      0.54      0.53      1000\n",
            "weighted avg       0.53      0.53      0.53      1000\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MdCe73C_wSmD",
        "colab_type": "code",
        "outputId": "1de012ee-4e4f-4515-8ef9-cf2875ed5161",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 337
        }
      },
      "source": [
        "from sklearn import metrics\n",
        "cm=metrics.confusion_matrix(Y_test,y_pred)\n",
        "\n",
        "import seaborn as sn\n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "df_cm = pd.DataFrame(cm, index = ['negative','neutral','positive'],\n",
        "                  columns = ['negative','neutral','postive'])\n",
        "plt.figure(figsize = (5,5))\n",
        "\n",
        "ax = sn.heatmap(df_cm, annot=True)\n",
        "\n",
        "bottom, top = ax.get_ylim()\n",
        "ax.set_ylim(bottom + 0.5, top - 0.5)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(3.0, 0.0)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 43
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAATQAAAEvCAYAAADLkD3HAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 360x360 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xFFQGeEpVVFk",
        "colab_type": "code",
        "outputId": "26cb1007-5158-441b-b7a9-c19aae2f0877",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 337
        }
      },
      "source": [
        "from sklearn import metrics\n",
        "cm=metrics.confusion_matrix(Y_test,y_pred)\n",
        "\n",
        "import seaborn as sn\n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "df_cm = pd.DataFrame(cm, index = ['negative','neutral','positive'],\n",
        "                  columns = ['negative','neutral','postive'])\n",
        "plt.figure(figsize = (5,5))\n",
        "\n",
        "ax = sn.heatmap(df_cm, annot=True)\n",
        "\n",
        "bottom, top = ax.get_ylim()\n",
        "ax.set_ylim(bottom + 0.5, top - 0.5)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(3.0, 0.0)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 45
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAATQAAAEvCAYAAADLkD3HAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 360x360 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TQZ5Q9feSJZ7",
        "colab_type": "code",
        "outputId": "48ec64d1-f4a0-4d6c-d797-57d018705feb",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 358
        }
      },
      "source": [
        "y_pred = attention_model.predict(xVal)\n",
        "y_pred_class = full_model.predict(xVal)\n",
        "\n",
        "req = np.argmax(y_pred,axis = 2).reshape((-1))\n",
        "classes = np.argmax(y_pred_class,axis=2).reshape(-1)\n",
        "imp_vecs = xVal\n",
        "pos_words = []\n",
        "neg_words = []\n",
        "neu_words = []\n",
        "for i,vecs in enumerate(imp_vecs):\n",
        "    if(classes[i] == 0):\n",
        "        neg_words.append({'word':v2w(vecs)})\n",
        "    if(classes[i] == 1):\n",
        "        neu_words.append({'word':v2w(vecs)})\n",
        "    if(classes[i] == 2):\n",
        "        pos_words.append({'word':v2w(vecs)})\n",
        "\n",
        "neg_words = pd.DataFrame(neg_words)\n",
        "neu_words = pd.DataFrame(neu_words)\n",
        "pos_words = pd.DataFrame(pos_words)\n",
        "\n",
        "print(neg_words['word'].value_counts())\n",
        "print(neu_words['word'].value_counts())\n",
        "print(pos_words['word'].value_counts())"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "error",
          "ename": "NameError",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-47-c2f1eee16472>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m     12\u001b[0m         \u001b[0mneg_words\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m{\u001b[0m\u001b[0;34m'word'\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0mv2w\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mvecs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m}\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     13\u001b[0m     \u001b[0;32mif\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mclasses\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 14\u001b[0;31m         \u001b[0mneu_words\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m{\u001b[0m\u001b[0;34m'word'\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0mv2w\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mvecs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m}\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     15\u001b[0m     \u001b[0;32mif\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mclasses\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m2\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     16\u001b[0m         \u001b[0mpos_words\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m{\u001b[0m\u001b[0;34m'word'\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0mv2w\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mvecs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m}\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m<ipython-input-38-c853a3bde3cf>\u001b[0m in \u001b[0;36mv2w\u001b[0;34m(vec)\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mv2w\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mvec\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 2\u001b[0;31m     \u001b[0;32mif\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mvec\u001b[0m\u001b[0;34m==\u001b[0m\u001b[0mavg_vec\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mall\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      3\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0;34m'UNK'\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m     \u001b[0;32mif\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mprod\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mvec\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0;34m'None'\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;31mNameError\u001b[0m: name 'avg_vec' is not defined"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "cn4w0tV-N4MQ",
        "colab_type": "text"
      },
      "source": [
        "##Word2Vec"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "outputId": "e27f1be1-fcfd-474e-fb73-30503f65e079",
        "id": "tzZodXJ5PLT_",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "xTrain,xVal,yTrain,yVal = getData(normalize=True,k=10000,embed='w2v')\n"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "rating\n",
            "1     21619\n",
            "2      6931\n",
            "3      6513\n",
            "4      5012\n",
            "5      8013\n",
            "6      6343\n",
            "7      9456\n",
            "8     18890\n",
            "9     27531\n",
            "10    50989\n",
            "dtype: int64\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:20: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame\n",
            "\n",
            "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "rating\n",
            "1      7299\n",
            "2      2334\n",
            "3      2205\n",
            "4      1659\n",
            "5      2710\n",
            "6      2119\n",
            "7      3091\n",
            "8      6156\n",
            "9      9177\n",
            "10    17016\n",
            "dtype: int64\n",
            "label\n",
            "0     40075\n",
            "1     14356\n",
            "2    106866\n",
            "dtype: int64\n",
            "label\n",
            "0    13497\n",
            "1     4829\n",
            "2    35440\n",
            "dtype: int64\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/smart_open/smart_open_lib.py:402: UserWarning: This function is deprecated, use smart_open.open instead. See the migration notes for details: https://github.com/RaRe-Technologies/smart_open/blob/master/README.rst#migrating-to-the-new-open-function\n",
            "  'See the migration notes for details: %s' % _MIGRATION_NOTES_URL\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "label\n",
            "0    3304\n",
            "1    3349\n",
            "2    3347\n",
            "dtype: int64\n",
            "label\n",
            "0    334\n",
            "1    351\n",
            "2    315\n",
            "dtype: int64\n",
            "(10000, 100, 300) (1000, 100, 300)\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow_core/python/ops/init_ops.py:97: calling GlorotUniform.__init__ (from tensorflow.python.ops.init_ops) with dtype is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Call initializer instance with the dtype argument instead of passing it to the constructor\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow_core/python/ops/init_ops.py:97: calling Orthogonal.__init__ (from tensorflow.python.ops.init_ops) with dtype is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Call initializer instance with the dtype argument instead of passing it to the constructor\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow_core/python/ops/init_ops.py:97: calling Zeros.__init__ (from tensorflow.python.ops.init_ops) with dtype is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Call initializer instance with the dtype argument instead of passing it to the constructor\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow_core/python/ops/resource_variable_ops.py:1630: calling BaseResourceVariable.__init__ (from tensorflow.python.ops.resource_variable_ops) with constraint is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "If using Keras pass *_constraint arguments to layers.\n",
            "(?, 100, 128)\n",
            "(?, 64)\n",
            "(?, 64)\n",
            "(?, 1, 128)\n",
            "(?, 1, 100)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ln3gFHwobftJ",
        "colab_type": "code",
        "outputId": "515f0e3d-a67a-4e9a-f8fe-052525297c5d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 104
        }
      },
      "source": [
        "full_model,attention_model = seq2seq(embed='w2v')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(?, 100, 128)\n",
            "(?, 64)\n",
            "(?, 64)\n",
            "(?, 1, 128)\n",
            "(?, 1, 100)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "outputId": "e23cc919-3302-4cc3-a64d-7d99ca44604c",
        "id": "_gWnl96-PLUM",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 627
        }
      },
      "source": [
        "training = full_model.fit(x=xTrain, y = yTrain, epochs=15, batch_size=128,validation_split=0.1)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow_core/python/ops/math_grad.py:1424: where (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use tf.where in 2.0, which has the same broadcast rule as np.where\n",
            "Train on 9000 samples, validate on 1000 samples\n",
            "Epoch 1/15\n",
            "9000/9000 [==============================] - 24s 3ms/sample - loss: 1.0641 - acc: 0.4230 - val_loss: 0.9801 - val_acc: 0.5390\n",
            "Epoch 2/15\n",
            "9000/9000 [==============================] - 22s 2ms/sample - loss: 0.9571 - acc: 0.5342 - val_loss: 0.9254 - val_acc: 0.5770\n",
            "Epoch 3/15\n",
            "9000/9000 [==============================] - 22s 2ms/sample - loss: 0.9138 - acc: 0.5637 - val_loss: 0.9248 - val_acc: 0.5570\n",
            "Epoch 4/15\n",
            "9000/9000 [==============================] - 22s 2ms/sample - loss: 0.8834 - acc: 0.5809 - val_loss: 0.8941 - val_acc: 0.5920\n",
            "Epoch 5/15\n",
            "9000/9000 [==============================] - 22s 2ms/sample - loss: 0.8542 - acc: 0.5989 - val_loss: 0.9024 - val_acc: 0.5650\n",
            "Epoch 6/15\n",
            "9000/9000 [==============================] - 22s 2ms/sample - loss: 0.8282 - acc: 0.6160 - val_loss: 0.9032 - val_acc: 0.5790\n",
            "Epoch 7/15\n",
            "9000/9000 [==============================] - 22s 2ms/sample - loss: 0.8084 - acc: 0.6262 - val_loss: 0.8848 - val_acc: 0.5980\n",
            "Epoch 8/15\n",
            "9000/9000 [==============================] - 21s 2ms/sample - loss: 0.7673 - acc: 0.6551 - val_loss: 0.8888 - val_acc: 0.5870\n",
            "Epoch 9/15\n",
            "9000/9000 [==============================] - 22s 2ms/sample - loss: 0.7326 - acc: 0.6748 - val_loss: 0.8742 - val_acc: 0.5800\n",
            "Epoch 10/15\n",
            "9000/9000 [==============================] - 23s 3ms/sample - loss: 0.6982 - acc: 0.6898 - val_loss: 0.9188 - val_acc: 0.5770\n",
            "Epoch 11/15\n",
            "9000/9000 [==============================] - 22s 2ms/sample - loss: 0.6672 - acc: 0.7057 - val_loss: 0.9482 - val_acc: 0.5890\n",
            "Epoch 12/15\n",
            "9000/9000 [==============================] - 22s 2ms/sample - loss: 0.6136 - acc: 0.7361 - val_loss: 0.9744 - val_acc: 0.5960\n",
            "Epoch 13/15\n",
            "9000/9000 [==============================] - 22s 2ms/sample - loss: 0.5724 - acc: 0.7562 - val_loss: 1.0608 - val_acc: 0.5590\n",
            "Epoch 14/15\n",
            "9000/9000 [==============================] - 22s 2ms/sample - loss: 0.5321 - acc: 0.7791 - val_loss: 1.0478 - val_acc: 0.5890\n",
            "Epoch 15/15\n",
            "9000/9000 [==============================] - 22s 2ms/sample - loss: 0.4839 - acc: 0.8017 - val_loss: 1.0830 - val_acc: 0.5910\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "outputId": "efcb4bed-d31a-4d7f-b567-0e168d5c2b03",
        "id": "kXHtNPGlPLUX",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 243
        }
      },
      "source": [
        "from sklearn.metrics import classification_report\n",
        "import numpy as np\n",
        "\n",
        "Y_test = np.argmax(yVal, axis=2).reshape((-1)) # Convert one-hot to index\n",
        "print(Y_test.shape)\n",
        "y_pred = full_model.predict(xVal)\n",
        "print(y_pred.shape)\n",
        "y_pred = y_pred.argmax(axis=2).reshape((-1))\n",
        "print(y_pred.shape)\n",
        "print(classification_report(Y_test, y_pred))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(1000,)\n",
            "(1000, 1, 3)\n",
            "(1000,)\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.63      0.59      0.61       334\n",
            "           1       0.52      0.58      0.55       351\n",
            "           2       0.64      0.62      0.63       315\n",
            "\n",
            "    accuracy                           0.59      1000\n",
            "   macro avg       0.60      0.59      0.59      1000\n",
            "weighted avg       0.60      0.59      0.59      1000\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "outputId": "76877f7d-189e-4380-836f-03f428ca586a",
        "id": "F7AoVUumPLUg",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 337
        }
      },
      "source": [
        "from sklearn import metrics\n",
        "cm=metrics.confusion_matrix(Y_test,y_pred)\n",
        "\n",
        "import seaborn as sn\n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "df_cm = pd.DataFrame(cm, index = ['negative','neutral','positive'],\n",
        "                  columns = ['negative','neutral','postive'])\n",
        "plt.figure(figsize = (5,5))\n",
        "\n",
        "ax = sn.heatmap(df_cm, annot=True)\n",
        "\n",
        "bottom, top = ax.get_ylim()\n",
        "ax.set_ylim(bottom + 0.5, top - 0.5)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(3.0, 0.0)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 30
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAATQAAAEvCAYAAADLkD3HAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 360x360 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IQZRCOB3aOQ6",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# y_pred = attention_model.predict(xVal)\n",
        "# y_pred_class = full_model.predict(xVal)\n",
        "\n",
        "# req = np.argmax(y_pred,axis = 2).reshape((-1))\n",
        "# classes = np.argmax(y_pred_class,axis=2).reshape(-1)\n",
        "# imp_vecs = xVal\n",
        "# pos_words = []\n",
        "# neg_words = []\n",
        "# neu_words = []\n",
        "# for i,vecs in enumerate(imp_vecs):\n",
        "#     if(classes[i] == 0):\n",
        "#         neg_words.append({'word':v2w(vecs,avg_vec,)})\n",
        "#     if(classes[i] == 1):\n",
        "#         neu_words.append({'word':v2w(vecs)})\n",
        "#     if(classes[i] == 2):\n",
        "#         pos_words.append({'word':v2w(vecs)})\n",
        "\n",
        "# neg_words = pd.DataFrame(neg_words)\n",
        "# neu_words = pd.DataFrame(neu_words)\n",
        "# pos_words = pd.DataFrame(pos_words)\n",
        "\n",
        "# print(neg_words['word'].value_counts())\n",
        "# print(neu_words['word'].value_counts())\n",
        "# print(pos_words['word'].value_counts())"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "2M-GcQyUN6XN",
        "colab_type": "text"
      },
      "source": [
        "##Elmo"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "outputId": "24255f9f-7f62-48cf-daa1-9766162f5375",
        "id": "plGYTUMGPRf-",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "xTrain,xVal,yTrain,yVal = getData(normalize=True,k=5000,embed='elmo')\n",
        "full_model,attention_model = seq2seq(embed='elmo')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "rating\n",
            "1     21619\n",
            "2      6931\n",
            "3      6513\n",
            "4      5012\n",
            "5      8013\n",
            "6      6343\n",
            "7      9456\n",
            "8     18890\n",
            "9     27531\n",
            "10    50989\n",
            "dtype: int64\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:20: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame\n",
            "\n",
            "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "rating\n",
            "1      7299\n",
            "2      2334\n",
            "3      2205\n",
            "4      1659\n",
            "5      2710\n",
            "6      2119\n",
            "7      3091\n",
            "8      6156\n",
            "9      9177\n",
            "10    17016\n",
            "dtype: int64\n",
            "label\n",
            "0     40075\n",
            "1     14356\n",
            "2    106866\n",
            "dtype: int64\n",
            "label\n",
            "0    13497\n",
            "1     4829\n",
            "2    35440\n",
            "dtype: int64\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "100%|██████████| 336/336 [00:00<00:00, 905710.89B/s]\n",
            "100%|██████████| 374434792/374434792 [00:11<00:00, 32457044.78B/s]\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "\n",
            "\n",
            ":: Lookup of 5000 ELMo representations. This takes a while ::\n",
            "100%[==================================================] 5000/5000 sentences\n",
            "\n",
            ":: Lookup of 1000 ELMo representations. This takes a while ::\n",
            "100%[==================================================] 1000/1000 sentenceslabel\n",
            "0    1681\n",
            "1    1641\n",
            "2    1678\n",
            "dtype: int64\n",
            "label\n",
            "0    323\n",
            "1    322\n",
            "2    355\n",
            "dtype: int64\n",
            "(5000, 100, 1024) (1000, 100, 1024)\n",
            "(?, 100, 128)\n",
            "(?, 64)\n",
            "(?, 64)\n",
            "(?, 1, 128)\n",
            "(?, 1, 100)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "outputId": "14ae923b-d92f-4dde-c38a-7419165a02fc",
        "id": "cfEbwc0VPRgM",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 555
        }
      },
      "source": [
        "training = full_model.fit(x=xTrain, y = yTrain, epochs=15, batch_size=128,validation_split=0.1)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 4500 samples, validate on 500 samples\n",
            "Epoch 1/15\n",
            "4500/4500 [==============================] - 14s 3ms/sample - loss: 1.0738 - acc: 0.4218 - val_loss: 1.0276 - val_acc: 0.4660\n",
            "Epoch 2/15\n",
            "4500/4500 [==============================] - 12s 3ms/sample - loss: 0.9543 - acc: 0.5422 - val_loss: 0.9799 - val_acc: 0.5140\n",
            "Epoch 3/15\n",
            "4500/4500 [==============================] - 12s 3ms/sample - loss: 0.8358 - acc: 0.6260 - val_loss: 0.9441 - val_acc: 0.5320\n",
            "Epoch 4/15\n",
            "4500/4500 [==============================] - 12s 3ms/sample - loss: 0.7400 - acc: 0.6756 - val_loss: 0.9717 - val_acc: 0.5200\n",
            "Epoch 5/15\n",
            "4500/4500 [==============================] - 12s 3ms/sample - loss: 0.6235 - acc: 0.7504 - val_loss: 1.0851 - val_acc: 0.5160\n",
            "Epoch 6/15\n",
            "4500/4500 [==============================] - 12s 3ms/sample - loss: 0.4913 - acc: 0.8149 - val_loss: 1.1109 - val_acc: 0.5200\n",
            "Epoch 7/15\n",
            "4500/4500 [==============================] - 12s 3ms/sample - loss: 0.3789 - acc: 0.8600 - val_loss: 1.2224 - val_acc: 0.5320\n",
            "Epoch 8/15\n",
            "4500/4500 [==============================] - 12s 3ms/sample - loss: 0.2334 - acc: 0.9300 - val_loss: 1.3681 - val_acc: 0.5420\n",
            "Epoch 9/15\n",
            "4500/4500 [==============================] - 12s 3ms/sample - loss: 0.1591 - acc: 0.9587 - val_loss: 1.5772 - val_acc: 0.5320\n",
            "Epoch 10/15\n",
            "4500/4500 [==============================] - 12s 3ms/sample - loss: 0.0863 - acc: 0.9833 - val_loss: 1.7532 - val_acc: 0.5320\n",
            "Epoch 11/15\n",
            "4500/4500 [==============================] - 11s 3ms/sample - loss: 0.0511 - acc: 0.9922 - val_loss: 1.8653 - val_acc: 0.5160\n",
            "Epoch 12/15\n",
            "4500/4500 [==============================] - 12s 3ms/sample - loss: 0.0248 - acc: 0.9984 - val_loss: 2.0603 - val_acc: 0.5140\n",
            "Epoch 13/15\n",
            "4500/4500 [==============================] - 12s 3ms/sample - loss: 0.0139 - acc: 0.9998 - val_loss: 2.2244 - val_acc: 0.5220\n",
            "Epoch 14/15\n",
            "4500/4500 [==============================] - 12s 3ms/sample - loss: 0.0085 - acc: 1.0000 - val_loss: 2.3187 - val_acc: 0.5240\n",
            "Epoch 15/15\n",
            "4500/4500 [==============================] - 12s 3ms/sample - loss: 0.0058 - acc: 1.0000 - val_loss: 2.3874 - val_acc: 0.5360\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "outputId": "3d908494-0355-4369-dce9-1775e8c3792a",
        "id": "SRnElbItPRga",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 191
        }
      },
      "source": [
        "from sklearn.metrics import classification_report\n",
        "import numpy as np\n",
        "\n",
        "Y_test = np.argmax(yVal, axis=2).reshape((-1)) # Convert one-hot to index\n",
        "y_pred = full_model.predict(xVal)\n",
        "y_pred = y_pred.argmax(axis=2).reshape((-1))\n",
        "print(classification_report(Y_test, y_pred))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.58      0.58      0.58       323\n",
            "           1       0.49      0.52      0.50       322\n",
            "           2       0.65      0.61      0.63       355\n",
            "\n",
            "    accuracy                           0.57      1000\n",
            "   macro avg       0.57      0.57      0.57      1000\n",
            "weighted avg       0.57      0.57      0.57      1000\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "outputId": "c3c050e9-8648-4015-84e1-c8d019fcd229",
        "id": "ruBfOOKoPRgi",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 337
        }
      },
      "source": [
        "from sklearn import metrics\n",
        "cm=metrics.confusion_matrix(Y_test,y_pred)\n",
        "\n",
        "import seaborn as sn\n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "df_cm = pd.DataFrame(cm, index = ['negative','neutral','positive'],\n",
        "                  columns = ['negative','neutral','postive'])\n",
        "plt.figure(figsize = (5,5))\n",
        "\n",
        "ax = sn.heatmap(df_cm, annot=True)\n",
        "\n",
        "bottom, top = ax.get_ylim()\n",
        "ax.set_ylim(bottom + 0.5, top - 0.5)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(3.0, 0.0)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 43
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAATQAAAEvCAYAAADLkD3HAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjEsIGh0\ndHA6Ly9tYXRwbG90bGliLm9yZy8QZhcZAAAgAElEQVR4nO3dd5wU9f3H8dfn7jiqdFBEBESQWBAE\nERUUGwqICJYgYoAYsYvdIJYomtgwikbl/IGKUROMGhUrIAZFFBuCAoYm0nuTfnef3x87kF2Eu2Vh\nb3fn3k8f87id2dn5fna8+/AtM98xd0dEJAyyUh2AiMi+ooQmIqGhhCYioaGEJiKhoYQmIqGhhCYi\noZGT7AI2/X2grguJw6UDp6U6hIwxcvGkVIeQEfK3LrREPrdtxZyE/mbL1DwkofL2JdXQRCQ0kl5D\nE5EMU1iQ6ggSpoQmIrG8MNURJEwJTURiFSqhiUhIuGpoIhIaqqGJSGhkcA1Nl22ISKzCgsSWYphZ\nPTMbZ2bTzOwHM+sfbL8gWC80s1Y7fWaAmc0ysx/N7MziylANTURiJa+Glg/c5O7fmNl+wNdmNhr4\nHugODI3e2cwOB3oARwAHAmPMrIm77zZ7KqGJSKwk9aG5+2JgcfB6vZlNB+q6+2gAs1/daNAV+Ie7\nbwHmmtksoDUwcXdlqMkpIjHcCxNazKyfmX0VtfTbXRlm1gBoAXxRRCh1gflR6wuCbbulGpqIxEqw\nhubueUBecfuZWSXgNeB6d1+XUGG7oYQmIrGSOMppZmWIJLOX3P31YnZfCNSLWj8o2LZbanKKSKzk\njXIaMAyY7u6PxhHJW0APMytrZg2BxkCRU62ohiYisZJXQzsRuASYamaTg223A2WBJ4BawDtmNtnd\nz3T3H8xsJDCNyAjp1UWNcIISmojsLHmjnJ8Cu5sz7Y3dfOZ+4P54y1BCE5FYulNARCT1VEMTkVi6\nOV1EwqKYfve0poQmIrEyuA9NCU1EYqnJKSKhoRqaiISGnvokIqGhGpqIhIb60EQkNFRDE5HQUA1N\nREJDCU1EwkJ3CohIeKiGJiKhoUEBEQkN1dBEJDQyuIamCR5FJDRUQxORWGpyikhoZHCTUwlNRGKp\nhiYioaGEJiKhoSZnct391leMn7mY6hXL8toVHX71/rpNW7n77a9YsHoDuTlZ3NOlFYfWrrJXZW7N\nL+CON79k+uLVVCmfy4PntaFu1YpMnLOUIWOnsq2gkDLZWdxwejNaN6y9V2Wli46XduGUHqfjDvNn\nzGPoLU/QpFVTet7eG7MstmzczDM3DWHpvCWpDjXlsrKy+OLz91i0cAldu/XmlPYn8uCDd5KbW4Zv\nvpnKZf1uoqAgQ28hyuAaWkZctnHO0fV5qmfb3b7/fxNmcNj+VXn18jO4r+uxPPTB5N3uu7OFazZw\n6YiPf7X9jck/UblcLm9f05FexzXh8bFTAahWPpfHe5zIv67owKCuxzLwzUl7/H3SUbX9q3Nm384M\nPPsWbuvQn6zsLI7v0pbf33cFf+v/GLd3upEJb47n3GsvSHWoaeG6a//AjBkzATAzhg97jIt7XUXz\nFqfx888L+N0lGXyevDCxJQ3EndDMrLyZHZbMYHanZf1aVC6fu9v35yxfR+uGtQBoWLMyi9ZuZOUv\nmwF4Z8o8Lh42lgvzRjPona8pKPS4yvz4x0V0Obo+AKcfXpdJc5fh7jStU43a+5UHoFGtymzZVsDW\n/Az9l3gn2dnZ5JbLJSs7i9zyZVm9dBXuTvlKke9bYb8KrFm6KsVRpl7dunXo1PE0hg9/BYAaNaqx\ndetWZs6cA8CYMePp3q1TKkPcO4WFiS1pIK6EZmZdgMnA+8F6czN7K5mB7Ykm+1dh7IyFAExduIrF\nazaydP0m5ixfxwfTFvB8n1MY2e8Mssx4d+rPcR1z2fpNHFA58oeck5VFpXJlWLNpa8w+Y6Yv5Dd1\nqpGbk71vv1AKrF66infy3uSJiXk89eVwNq3fwNRPvuPZ2/7Grc/fyROfP0vb7u156+nXUx1qyj06\n+B7+OOA+CoM/4hUrVpGTk0PLY5oB0L17Zw6qd2AqQ9w7GVxDi7cP7U9Aa+BjAHefbGYNd7ezmfUD\n+gE80bcjl57aYu+iLMbvT2zKQx9M5sK80TSuXYXDDqhKlhmTflrG9MWruXjYWAC2bCugesWyANww\n8jMWrtlAfkEhi9du5MK80QD0bN2Yc5s3KLbMWcvW8vhHU3m6Z7ukfa+SVLFyRVp2aE3/tlewcd0G\n+j91Cyd2O5ljz2rDQ30GMXvyTM6+/Fx63dmXZ297KtXhpkznTqezbNkKvvl2KiefdPyO7Rf3uorB\nj/yJsmVzGT1mPAUF6fEHnpA0qW0lIt6Ets3d15pZ9Lbdtt3cPQ/IA9j094HxtfH2QqWyZbj3nGO3\nl02nJ97joGoV+fbnFXRpVp/rTjvqV5/564UnAJE+tLve+pJhv2sf837t/cqzZN0m9q9cgfzCQn7Z\nvI2qQbN36bqN3PjqRAZ1PZZ61Ssl98uVkCPbHs2y+UtZv2odAF++/zmHtWxK/d80YPbkSF/RxLc/\n5bYRd6UyzJQ74YRWdDm7Ax3POpVy5cpSufJ+vPD8EHr3uY72p3YH4IzTT6Jx40NSHOleyOCEFm8f\n2g9m1hPINrPGZvYE8FkS49oj6zZvZVvwL+Lr386l5cE1qVS2DK0b1mb0jIWs2hDpT1u7aSuL1myI\n65gnN6nD29/NA2DMtIUc26A2Zsa6zVu59pUJ9D/1KFrUq5mcL5QCKxYtp3GLJuSWiyTtI05sxoJZ\n86mwXwUOaBhpPh3V7mgWzVqQyjBTbuAdD9DgkFYc2qQNF/e6inHjJtC7z3XUqlUDgNzcXG65+Wry\n8l5McaR7wT2xJQ3EW0O7FhgIbAFeBj4A7ktWUDv74+tf8NW85azZuIUOj73DlScfTn7wr8gFLRsx\nd8V67nzzS4xIR/2furSC4PU17Y/gipc+wR1ysowBHVtwYNWKxZbZrUVDBv57El2efI/K5XN5sPtx\nAPzzy9n8vPoXhn4yjaGfTAPgmYvbUb1iueR8+RIye/JMvnh3In9+ZzAFBYX89MMcPnr5Q1YtXsn1\nz9yKFxayYe0G8m55MtWhpqWbb7ySTp1PJysri6FDRzDu4wmpDilxGVxDM48js5rZMe7+TSIFlEST\nMwwuHTgt1SFkjJGLw3GpTLLlb11oxe/1a5teujOhv9nyFw9KqLx9Kd4a2mAzOwD4F/BPd/8+iTGJ\nSCqlyYhlIuLqQ3P3U4BTgOXAUDObamZ3JDUyEUmNsF+HBuDuS9x9CHAFkWvSSvdwl4iknbianGb2\nG+C3wHnASuCfwE1JjEtEUiVNRiwTEW8f2nAiSexMd1+UxHhEJNXSpPmYiLgSmrsfX/xeIhIKGZzQ\niuxDM7ORwc+pZjYlaplqZlNKJkQRKVFJupfTzIab2TIz+z5qW3Mz+9zMJpvZV2bWOthuZjbEzGYF\nOeeYeEIvrobWP/h5djwHE5HM53HOSJOA54EngRFR2x4C7nH398ysU7DeHugINA6W44Cng59FKrKG\n5u6Lg5dXufu86AW4as++i4hkhCRdtuHu44Gd559yoHLwugqwvY++KzDCIz4HqppZneLKiPeyjTN2\nsa1jnJ8VkUxSstMHXQ88bGbzgUeAAcH2usD8qP0WBNuKVFwf2pVmNhU4bKc+tLmA+tBEwqjQE1rM\nrF/QD7Z96RdHaVcCN7h7PeAGYNjehF5cH9rLwHvAX4A/Rm1f7+6aulQkjBIc5YyeNmwP9OZ/ffWv\nAv8XvF4I1Iva76BgW5GK60Nb6+4/uftFQb/ZJiJt3kpmdvAeBi4imaBkb31aBJwcvD4VmBm8fgv4\nXTDa2QZYG9Wnv1vx3inQBXgUOBBYBtQHpgNH7FnsIpL2knSngJm9QmQEs6aZLQDuBi4DHjezHGAz\nwUzXwLtAJ2AWsBHoG08Z8d4pcB/QBhjj7i3M7BSgV5yfFZFMkqQLa939ot281XIX+zpw9Z6WEe8o\n5zZ3XwlkmVmWu48DWu1pYSKSARIcFEgH8dbQ1phZJWA88JKZLQPim8taRDJL2OdDI3KR2yYiw6rv\nA7OBLskKSkRSKOw1NHePro29kKRYRCQNeAbfnB7vKOd6fv3YurXAV8BN7j5nXwcmIrKn4u1De4zI\nrQcvAwb0ABoB3xCZK619MoITkRRIk+ZjIuJNaOe4+9FR63lmNtndbzOz25MRmIikSCkYFNhoZhea\nWVawXEjkIjgo4gnqIpKBwj4oAFwMPA48RSSBfQ70MrPywDVJik1EUiHsgwJBp//uLtP4dN+FIyIp\nlya1rUTE1eQ0syZmNnb71Llm1kzP5RQJqZKdD22fircP7VkiE69tA3D3KURGOkUkbEpBH1oFd59k\nZtHb8pMQj4ikWOgvrAVWmFkjghFNMzsfKHZuIhHJQGlS20pEvAntaiIzUTY1s4XAXCIjnyISNqUg\noS0EngPGAdWBdUSmzr03SXGJSKqkSQd/IuJNaG8Ca4jc6rSomH1FJJOVghraQe5+VlIjEZG0kMQH\nDSddvJdtfGZmRyU1EhFJD6Xgso22QJ/geZxbiMy44e7eLGmRiUhqlILLNvSUdJHSIk1qW4mI917O\neckORETSRAYntHj70ERE0l68TU4RKSU8SQ8aLglKaCISK4ObnElPaBcPnJrsIkJhxHmpjiBzTH+5\nQapDCDclNBEJi0y+sFYJTURiKaGJSGhk7nW1SmgiEktNThEJDyU0EQkNNTlFJCzU5BSR8FANTUTC\nQjU0EQkP1dBEJCwy+BkpSmgishMlNBEJi0yuoWmCRxEJDSU0EYlVmOBSDDMbbmbLzOz7qG1/MrOF\nZjY5WDpFvTfAzGaZ2Y9mdmY8oavJKSIxktjkfB54Ehix0/a/uvsj0RvM7HCgB3AEcCAwxsyauHtB\nUQWohiYiMbwwsaXY47qPB1bFGUZX4B/uvsXd5wKzgNbFfUgJTURiJJrQzKyfmX0VtfSLs8hrzGxK\n0CStFmyrC8yP2mdBsK1ISmgiEsstocXd89y9VdSSF0dpTwONgObAYmDw3oSuPjQRiVGSl224+9Lt\nr83sWWBUsLoQqBe160HBtiKphiYiMbzQEloSYWZ1ola7AdtHQN8CephZWTNrCDQGJhV3PNXQRCRG\nsmpoZvYK0B6oaWYLgLuB9mbWHHDgJ+ByAHf/wcxGAtOAfODq4kY4QQlNRHbinlhtq/jj+kW72Dys\niP3vB+7fkzKU0EQkRibf+qSEJiIxEu0PSwdKaCISwzN3fkclNBGJpRqaiISGEpqIhIaanCISGplc\nQ9OdAiISGqqhiUiMZF1YWxKU0EQkhi6sFZHQKFQNTUTCQk1OEQmNTB7lVEITkRi6Dk1EQkM1NBEJ\nDQ0KiEhoaFAgzZ196Tmc3qMDuDNvxjyevOVxLr//Ko5ocyQb120A4ImbH+enaXNTHOneKfvb68g+\nvBX+y1o2PXztr94vc0o3co45ObKSlU3W/gex4a5LYOMviReanUPZnjeQXe9QfMM6No94GF+9jOwm\nzcnt/DvIyYH8fLa+/TwFs6YkXk4aqd/oYB4ceu+O9br1D+Tph/6PUa++x4NDB3FgvQNYNH8Jt/a7\nk/Vr16cw0sSoDy2NVd+/Op37dqH/aVezdctWbvrbrbTt0g6AEX9+jonvfpbiCPedbV+OZdunoyjb\n84Zdvz/uDbaNewOA7MOPpczJXeNOZlatNuUu6s+mpwbGbM857gzY9Asb/3w5Oc3bkXt2b7a8+HAk\nuQ27D1+3iqwDDqbc5few8Z6+e/cF08S82T/T4/Q+AGRlZfHB5H8z7r3/0PfaS5j0yVc89+Tf6XtN\nL/pe24sh9z2d2mATkMlNzlJxL2d2dha55XLJys6ibPmyrFoa78ObM0vhnB/wOBNUzjEnkf/t+P+t\nt2xP+esfofxNj1H2gqvA4vvVyDnyOLZ9+REA+VMmkNP46EgsC+fg6yLnuXDJz1iZXMgO37+frdu1\nYsFPC1m8YCntz2zH2yPfA+Dtke9xylknpTi6xLhbQks6CH1CW7V0FW/m/ZuhE4cx7MsX2Lh+A999\nMhmAnjf34tH3h9D3zkvJyQ3fH9tulcklp+kx5E+J1E6t9kHkNG/LpiG3sWnw9VBYSE7Lk+M6lFWp\nga9ZEVkpLMQ3b4CK+8Xsk93sBAoWzIaC/H36NdLBmeeexvv/HgNAjVrVWLFsJQArlq2kRq1qRX00\nbbkntqSDIv+KzWw9kcdL/eotwN29clKi2ocqVq5I6w7HcWXby9iwbgM3P3UbJ3Vrz0sPjWD1stXk\n5OZw5V+uodsV5/HqkH+mOtwSkXNEawrmTt/R3MxpcjRZBzWi/A2Rh1ZbmVz8l7UAlOs7AKu+P5ad\ng1WrRfmbHgNg2/i3yf9ybLFlZe1fj7Jn92bT0LuT9G1SJ6dMDid3aMsT9z+zy/c9Xf7K91AmNzmL\nTGjuvl9R7++OmfUD+gE0r96MhpXqJ3KYfaJZ2+Ysnb+UdavWAfDF+xNp2rIp49/4GID8rfl89OoY\nuvbrlrIYS1pOi3YxzU2A/K/GsfWdEb/ad/NzfwF234fma1diVWvia1dCVhZWriJsiHSEW5UalOt7\nO5tffgxfuSRJ3yZ12p7ahhlT/8uqFasBWLl8NTVr12DFspXUrF2DVSvWpDjCxKRL8zERe9TkNLPa\nZnbw9mV3+7l7nru3cvdWqUxmACsWLadJi8PILZcLwFEnHs2CWfOpVvt/zYHjOrTh5x/npSrEklWu\nAtmNjiT/+y92bMqfOYWcZidglapENlSohFWrFdfhCn6YRJljTwUgp9mJ5G8fySxXkXKX3cWWd0ZQ\n+NP0ffoV0sVZ3c7g/X+P3rH+nw8/pcuFHQHocmFHPv7gk1SFtlcK3RJa0kFcHUdmdg4wGDgQWAbU\nB6YDRyQvtH1j5uT/MvHdCTzyzmMUFhQw54c5fPjyB9z5wp+oXL0yZsbcaXMZevtTqQ51r5XtdTPZ\nhx6JVaxMhbuGs/WDVyArG4D8ie8DkHNUG/J//Ba2btnxOV86n63v/Z1yl98TGQwoyGfL60Px1cuL\nLXPbF6Mp1/NGKtw+FN+4ns0jHgagTNvOZNWoQ26H30KH3wKweejdO5qyma5chXIcd9Kx3HfLQzu2\nPffEizyYN4hze57N4gWRyzakZFk87Xwz+w44FRjj7i3M7BSgl7tfWtxnu9c/JzM7EkrYiPNSHUHm\naPfyylSHkBG+XTIhoWrT5wd2T+hvts2i11NeTYu3ybnN3VcCWWaW5e7jgFZJjEtEUiT0TU5gjZlV\nAsYDL5nZMmBD8sISkVQpDYMCXYGNwA3A+8BsoEuyghKR1ClMcEkHxdbQzCwbGOXupxCJ+4WkRyUi\nKeNkbg2t2ITm7gVmVmhmVdw9HENUIrJbhRk8jBdvH9ovwFQzG01U35m7X5eUqEQkZQrDXEMLvB4s\n0TI4j4vI7oS6yRmo6u6PR28ws/5JiEdEUixdOvgTEe8oZ+9dbOuzD+MQkTThWEJLOihuto2LgJ5A\nQzN7K+qt/YBwTiomUsplcg2tuCbnZ8BioCaRezm3Ww+EYz5lEYkR2oTm7vOAecDxJROOiKRaujQf\nExHvbBvREz3mAmWADZkwwaOI7JkMfixnfIMC7r6fu1cOElh54Dwg8+fbEZFfKcQSWopjZsPNbJmZ\nfR+17WEzm2FmU8zsDTOrGvXeADObZWY/mtmZ8cS+x88U8Ih/A3EVICKZxRNc4vA8cNZO20YDR7p7\nM+C/wAAAMzsc6EFkzsWzgKeC2zCLFG+Ts3vUahaRqYM2x/NZEREAdx9vZg122vZh1OrnwPnB667A\nP9x9CzDXzGYBrYGJRZUR74W10TNr5AM/BQWKSMikcJTz98D2JxXVJZLgtlsQbCtSXAnN3cPxhFgR\nKVahJTYqEP1wpECeu+fF+dmBRCpLLyVUeCDeJmcT4Glgf3c/0syaAee4+317U7iIpJ9Eb9IOkldc\nCSyamfUBzgZO8/89E2AhUC9qt4OCbUWKd1DgWSKdddsA3H0KkQ47EQmZkpzg0czOAm4lUkHaGPXW\nW0APMytrZg2BxsCk4o4Xbx9aBXefZLFV0fA9BltEknYdmpm9ArQHaprZAuBuIhWlssDoIL987u5X\nuPsPZjYSmEYk11zt7gXFlRFvQlthZo0IaqNmdj6RW6JEJGSSNR+au1+0i83Ditj/fuD+PSkj3oR2\nNZG2cVMzWwjMBS7ek4JEJDNk8kSH8Sa0hcBzwDigOrCOyJRC9yYpLhFJkUy+9SnehPYmsAb4BliU\nvHBEJNVCO9tGlIPcfedbFkQkhDK5yRnvZRufmdlRSY1ERNJCoSW2pIN4a2htgT5mNhfYAhiR+9Sb\nJS0yEUmJ0tDk7JjUKEQkbYQ+oQUz14pIKeBp0nxMRLw1NBEpJUJfQxOR0kMJTURCozRctiEikvZU\nQxORGOlyTVkilNBEJIb60EQkNJTQRCQ0MnlQQAlNRGKoD01EQkNNThEJDTU5i/DW4q+TXUQoXPHa\ncakOIWNM+v7FVIcQaoUZnNJUQxORGGpyikhoZG79TAlNRHaiGpqIhIYu2xCR0NCggIiERuamMyU0\nEdmJ+tBEJDQyucmpCR5FJDRUQxORGJlbP1NCE5GdqA9NREIjk/vQlNBEJEbmpjMlNBHZiZqcIhIa\nnsF1NCU0EYmhGpqIhIYGBUQkNDI3nSmhichOVEMTkdDI5D403cspIjE8wf/iYWb9zex7M/vBzK4P\ntlU3s9FmNjP4WS3R2JXQRCRGYYJLcczsSOAyoDVwNHC2mR0K/BEY6+6NgbHBekKU0EQkRhJraL8B\nvnD3je6eD/wH6A50BV4I9nkBODfR2JXQRCRGojU0M+tnZl9FLf12OvT3QDszq2FmFYBOQD1gf3df\nHOyzBNg/0dg1KCAiMQo9sVFOd88D8op4f7qZPQh8CGwAJgMFO+3jZpbwMKtqaCJSYtx9mLu3dPeT\ngNXAf4GlZlYHIPi5LNHjK6GJSAxPcImHmdUOfh5MpP/sZeAtoHewS2/gzURjV5NTRGIk+cLa18ys\nBrANuNrd15jZA8BIM7sUmAdcmOjBldBEJEYyZ9tw93a72LYSOG1fHF8JTURiZPKdAkpoIhJD93KK\nSGhogkcRCQ01OUUkNDzBC2vTgRKaiMRQH5qIhIaanCISGhoUEJHQUJNTREJDgwIiEhrqQ8sAWVlZ\nfPH5eyxauISu3XpzSvsTefDBO8nNLcM330zlsn43UVBQUPyBQuysS8+mfY/TcYcFM+aRd8uTNG55\nGD0H9iG7TA4/TZ3Ns7f+jcKCTP6Vh8VLl3P7oEdYuXo1hnF+145ccmHsJKmjPviIYS+9Cg4VKpTn\nzpuvoWnjQ/aq3K1btzJg0GCm/TiTqlUq88i9A6hbZ38+m/QNjz3zHNu25VOmTA43XX0px7Vsvldl\n7Y1M7kMrNdMHXXftH5gxYyYAZsbwYY9xca+raN7iNH7+eQG/u+SCFEeYWtX2r06Hvp258+xbGdDh\nerKyszi+azsuH3wdT14zmAEdrmfFwuW0O/+UVIe613Kys7nl2st466U8Xs77K/94fRSz586L2afu\ngQfw/JMP8caLT3NFn4u456EhcR9/4eKl9Lnm1l9tf33Uh1TerxLvjRzOJb89l0efGg5AtaqVefLB\nP/HGi09z/x03MeDeR/buC+6lQjyhJR2UioRWt24dOnU8jeHDXwGgRo1qbN26lZkz5wAwZsx4unfr\nlMoQ00J2dja55XLJys4it3xZtmzcQv62fJbMjcyO/P0n33FsxzYpjnLv1apZncMPOxSAihUrcEj9\neixdvjJmnxZHHU6VyvsB0OyIpixdtmLHe29/8BE9/tCf83pfzT0PDYm7Zv/RJxPp2ul0ADq0b8cX\nX0/G3flNk0OpXasGAIc2rM/mLVvYunXrXn/P0iiuhGYRvczsrmD9YDNrndzQ9p1HB9/DHwfcR2Fh\npKm0YsUqcnJyaHlMMwC6d+/MQfUOTGWIKbd66SrezXuTxycO5ckvh7Fx/Ua+GDWB7OxsGh7VCIDW\nnY6nRp2aKY5031q4eCnTZ86m2RGH7Xaf10d9QNs2rQCY/dPPvD/2P7z4zGBee+FvZGVlMerDcXGV\ntWz5Sg6oHTl/OTnZVKpYgTVr18XsM/rjTzn8sEPJzc1N8BvtPXdPaEkH8fahPUWkr/BU4F5gPfAa\ncGyS4tpnOnc6nWXLVvDNt1M5+aTjd2y/uNdVDH7kT5Qtm8voMeMpyPB+ob1VoXJFjunQmhvaXsnG\ndRu49qmbObHbSTx57WB63dWXnNwyTP1kcsb3n0XbuHETNwy8j9uuu5xKFSvucp9JX3/H66M+5MWn\nI83AL76azLQZs+hxaX8AtmzZQvVqVQG4bsC9LFy0lG3521i8dDnn9b4agF4XdqVb5w7FxjNrzjwe\nfWo4eX+9f198vYSlS/MxEfEmtOPc/Rgz+xbA3Veb2W7/CQme9tIPwLKrkJW161+WknDCCa3ocnYH\nOp51KuXKlaVy5f144fkh9O5zHe1P7Q7AGaefROO97PDNdEe2bcby+UtZvypSY/jq/S9o3LIpE94Y\nz6AL7ojs0+5o6jQMR012W34+1w+8j84dTuGM9ifucp8fZ83lrgce45nBg6hapTIQqb2c0/F0briy\n76/2H/KXu4BIrW/g/YN5/smHYt6vXasGS5at4IDatcjPL+CXDRt3HHfJsuX0v30Qf77zZg4+KLXn\nuDQMCmwzs2yCqcPNrBZFjO66e567t3L3VqlMZgAD73iABoe04tAmbbi411WMGzeB3n2uo1bQZ5Gb\nm8stN19NXt6LKY0z1VYuWsGhLZqQWy7y79QRJx7FwlkLqFyjCgA5uTl0ubIbY1/6IJVh7hPuzl1/\neYxD6tejd4/uu9xn8ZJlXH/7IP5y1y00OPigHdvbtGrO6I8/ZeXqNQCsXbeeRUuWxlXuKW3b8Oa7\nYwD48ONPOK7l0ZgZ69b/wlW33M31V/TlmGZH7OW323uF7gkt6SDeGtoQ4A2gtpndD5wP3JG0qErA\nzTdeSafOp5OVlcXQoSMY9/GEVIeUUrMnz2TSuxO5751HKCgoZN4Pcxj38odccHNPmp/Wiiwzxvz9\nA6Z99n2qQ91r3075gbffH5oU6VsAAAlmSURBVEvjRg12NAv7X96bxUuXA/Dbbp15+rmXWbtuPfc9\n8jcgMmAycvgQGjWsz7WX/Y5+1w+k0Aspk5PDwBuv4sADin+UZPezz2TAoIfpeOHvqVJ5Px6+J/KA\n8Fdee5v5CxbxzHMv88xzLwOQ99j91AiasiUtPVJTYizezjwza0pk3m8j8tj26fF8Lie3biafnxLT\no85xqQ4hYzz3dWova8gUZWoeYol87sS6pyb0Nzth4UcJlbcvxVVDM7MhwD/c/W9JjkdEUiyTBwXi\n7UP7GrjDzGab2SNm1iqZQYlI6mTyZRtxJTR3f8HdOxG5TONH4EEzm5nUyEQkJTL5ToE9vZfzUKAp\nUB+Iqw9NRDJLJl+2EW8f2kNAN2A28E9gkLuvSWZgIpIa6dJ8TES8NbTZwPHuvqLYPUUko6VL8zER\nRSY0M2vq7jOAL4GDzezg6Pfd/ZtkBiciJS/MNbQbidzCNHgX7zmReztFJERCW0Nz937By47uvjn6\nPTMrl7SoRCRlMnlQIN7r0D6Lc5uIZLjQ3stpZgcAdYHyZtaCyG1PAJWBCkmOTURkjxTXh3Ym0Ac4\nCHg0avt64PYkxSQiKZTJTc7i+tBeAF4ws/Pc/bUSiklEUihdmo+JKK7J2cvd/w40MLMbd37f3R/d\nxcdEJIOFtoYGbJ+dsVKyAxGR9BDaGpq7Dw1+3lMy4YhIqmVyDS3epz49ZGaVzayMmY01s+Vm1ivZ\nwYlIycvkyzbivQ6tg7uvA84GfiIy68YtyQpKRFLHE/wvHcR7c/r2/ToDr7r7WrOUz7YrIkngnrmP\nKoy3hjbKzGYALYGxwVOfNhfzGRHJQMmc4NHMqprZv8xshplNN7Pjzay6mY02s5nBz2qJxh7vjLV/\nBE4AWrn7NmAD0DXRQkUkfSV5Cu7HgffdvSlwNJGJYv9I5MFLjYGxwXpC4p3gsQzQCzgpaGr+B3gm\n0UJFJH0la7YNM6sCnETk7iPcfSuw1cy6Au2D3V4APgZuS6SMePvQngbKAE8F65cE2/6QSKEikr6S\nOB9aQ2A58JyZHU3k4Uv9gf3dfXGwzxKg+Iec7ka8Ce1Ydz86av0jM/su0UJFJH0legmGmfUjMn/i\ndnnunhe1ngMcA1zr7l+Y2ePs1Lx0dzezhDNqvAmtwMwaufvsIPBDgIJECxWR9JXoJRhB8sorYpcF\nwAJ3/yJY/xeRhLbUzOq4+2IzqwMsSygA4k9otwDjzGxOsN4A6JtooSKSvpLV5HT3JWY238wOc/cf\ngdOAacHSG3gg+PlmomXEm9AmAEODANYAHwATEy1URNJXkqfgvhZ4ycxygTlEKkZZwEgzuxSYB1yY\n6MHjTWgjgHXAoGC9J/AicEGiBYtIekrmQ1LcfTLQahdvnbYvjh9vQjvS3Q+PWh9nZtP2RQAiIvtK\nvHcKfGNmbbavmNlxwFfJCUlEUimTb06Pt4bWEvjMzH4O1g8GfjSzqURGWpslJToRKXFhfi7ndmcl\nNQoRSRuhfS7ndu4+L9mBiEh6KA01NBEpJdKlPywRSmgiEiNdJmtMhBKaiMRQDU1EQkN9aCISGmpy\nikhoqIYmIqGhhCYioZG56Qwsk7Nxosys304zacpu6FzFR+cpPcR7c3rY9Ct+FwnoXMVH5ykNlNaE\nJiIhpIQmIqFRWhOa+jrip3MVH52nNFAqBwVEJJxKaw1NREKo1Cc0M6tqZldFrR9oZv9KZUzpxswa\nmFnPBD/7y76OJx2Z2blmdnjU+r1mdnoqYyqNSn1CA6oCOxKauy9y9/NTGE86akDkSV+/Yma6ODvi\nXGBHQnP3u9x9TArjKZXSPqEFtYPpZvasmf1gZh+aWXkza2Rm75vZ12b2iZk1DfZvZGafm9lUM7tv\new3BzCqZ2Vgz+yZ4r2tQxANAIzObbGYPB+V9H3zmczM7IiqWj82slZlVNLPhZjbJzL6NOlZaSeDc\nPW9m50d9fnvt6gGgXXCObjCzPmb2lpl9BIwt4txmhOA8zTCzl4Lz9S8zq2BmpwX/f6cG/7/LBvs/\nYGbTzGyKmT1iZicA5wAPB+eo0fZzaWZnmdmrUWW1N7NRwesOZjYxOG+vmlml1JyBEHH3tF6I1A7y\ngebB+kigFzAWaBxsOw74KHg9CrgoeH0F8EvwOgeoHLyuCcwCLDj+9zuV933w+gbgnuB1HeDH4PWf\ngV7B66rAf4GKqT5X++DcPQ+cH/X57eeuPTAqansfYAFQvahzG32MdF6C8+TAicH6cOAOYD7QJNg2\nArgeqAH8GPX9qu7m3D0PnB+cm5+3/34ATwf/D2oC46O23wbclepzkelL2tfQAnM98oBSgK+J/AKe\nALxqZpOJPNW9TvD+8cD2fxFfjjqGAX82synAGKAusH8x5Y4k8ksJkac5b+9b6wD8MSj7Y6AckSdh\npaM9OXd7YrS7rwpeJ3Ju0818d58QvP47kQffznX3/wbbXgBOAtYCm4FhZtYd2FjUQd09H3gf6BI0\nzzsDbwJtiDRRJwT/H3oD9fftVyp9MqX/Y0vU6wIifyxr3L35HhzjYqAW0NLdt5nZT0QS0W65+0Iz\nW2lmzYDfEqnxQeQP+Dx3/3EPyk+VPTl3+QTdEGaWBeQWcdwNUa/3+NymoZ2vX1pDpDYWu5N7vpm1\nJpLwzgeuAU4t5tj/CPZbBXzl7uvNzIj8o3DRXkcuO2RKDW1n64C5ZnYBgEUcHbz3OXBe8LpH1Geq\nAMuCP7hT+N+/huuB/Yoo65/ArUAVd58SbPsAuDb4pcTMWuztFypBRZ27n4g8gxUifUJlgtfFnaPd\nndtMcrCZHR+87knkQdoNzOzQYNslwH+Cfq4q7v4ukS6J7eeuqHP0H+AY4DIiyQ0iv6cnbj9+0C/b\nZF9+odIoUxMaRGoFl5rZd8APwPaO6OuBG4Pmz6FEmggALwGtLPJw5N8BMwDcfSWRav/3ZvbwLsr5\nF5HEODJq2yAif+xTzOyHYD2T7O7cPQucHGw/nv/VwqYABWb2nZndsIvj7fLcZpgfgavNbDpQDfgr\n0JdI03wqUAg8QyRpjQp+vz4Fbgw+/w/glmAQoVH0gd29gEjfbsfgJ+6+nEhf5CvBsSYCTZP6DUuB\n0N0pYGYVgE3u7mbWg8gAQUaNuknJMrMGRAY9jkxxKLKXMqUPbU+0BJ4MmoNrgN+nOB4RKSGhq6GJ\nSOmVyX1oIiIxlNBEJDSU0EQkNJTQRCQ0lNBEJDSU0EQkNP4fHxTejXbYwuQAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 360x360 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3MZdTH3NaQnw",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# y_pred = attention_model.predict(xVal)\n",
        "# y_pred_class = full_model.predict(xVal)\n",
        "\n",
        "# req = np.argmax(y_pred,axis = 2).reshape((-1))\n",
        "# classes = np.argmax(y_pred_class,axis=2).reshape(-1)\n",
        "# imp_vecs = xVal\n",
        "# pos_words = []\n",
        "# neg_words = []\n",
        "# neu_words = []\n",
        "# for i,vecs in enumerate(imp_vecs):\n",
        "#     if(classes[i] == 0):\n",
        "#         neg_words.append({'word':v2w(vecs)})\n",
        "#     if(classes[i] == 1):\n",
        "#         neu_words.append({'word':v2w(vecs)})\n",
        "#     if(classes[i] == 2):\n",
        "#         pos_words.append({'word':v2w(vecs)})\n",
        "\n",
        "# neg_words = pd.DataFrame(neg_words)\n",
        "# neu_words = pd.DataFrame(neu_words)\n",
        "# pos_words = pd.DataFrame(pos_words)\n",
        "\n",
        "# print(neg_words['word'].value_counts())\n",
        "# print(neu_words['word'].value_counts())\n",
        "# print(pos_words['word'].value_counts())"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "d6Y6W5EHgAJX",
        "colab_type": "code",
        "outputId": "16192a6c-9d15-48be-ec7f-e234213c4b5c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "xTrain,xVal,yTrain,yVal = getData(normalize=True,k=5000,embed='elmo',classes=2)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "rating\n",
            "1     21619\n",
            "2      6931\n",
            "3      6513\n",
            "4      5012\n",
            "5      8013\n",
            "6      6343\n",
            "7      9456\n",
            "8     18890\n",
            "9     27531\n",
            "10    50989\n",
            "dtype: int64\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:20: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame\n",
            "\n",
            "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "rating\n",
            "1      7299\n",
            "2      2334\n",
            "3      2205\n",
            "4      1659\n",
            "5      2710\n",
            "6      2119\n",
            "7      3091\n",
            "8      6156\n",
            "9      9177\n",
            "10    17016\n",
            "dtype: int64\n",
            "label\n",
            "0     48088\n",
            "1    113209\n",
            "dtype: int64\n",
            "label\n",
            "0    16207\n",
            "1    37559\n",
            "dtype: int64\n",
            "\n",
            "\n",
            ":: Lookup of 5000 ELMo representations. This takes a while ::\n",
            "100%[==================================================] 5000/5000 sentences\n",
            "\n",
            ":: Lookup of 1000 ELMo representations. This takes a while ::\n",
            "100%[==================================================] 1000/1000 sentenceslabel\n",
            "0    2538\n",
            "1    2462\n",
            "dtype: int64\n",
            "label\n",
            "0    527\n",
            "1    473\n",
            "dtype: int64\n",
            "(5000, 100, 1024) (1000, 100, 1024)\n",
            "(?, 100, 128)\n",
            "(?, 64)\n",
            "(?, 64)\n",
            "(?, 1, 128)\n",
            "(?, 1, 100)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Ah4bYPs0hHN2",
        "colab_type": "code",
        "outputId": "37d73f84-833f-4fea-c9c6-38a4f78186c8",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 104
        }
      },
      "source": [
        "full_model,attention_model = seq2seq(embed='elmo',classes=2)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(?, 100, 128)\n",
            "(?, 64)\n",
            "(?, 64)\n",
            "(?, 1, 128)\n",
            "(?, 1, 100)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "D3NzGtaegFC8",
        "colab_type": "code",
        "outputId": "114f279a-3226-470c-870d-863b3b5f9546",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 555
        }
      },
      "source": [
        "training = full_model.fit(x=xTrain, y = yTrain, epochs=15, batch_size=128,validation_split=0.1)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 4500 samples, validate on 500 samples\n",
            "Epoch 1/15\n",
            "4500/4500 [==============================] - 14s 3ms/sample - loss: 0.6650 - acc: 0.5913 - val_loss: 0.5873 - val_acc: 0.6840\n",
            "Epoch 2/15\n",
            "4500/4500 [==============================] - 12s 3ms/sample - loss: 0.5402 - acc: 0.7316 - val_loss: 0.5973 - val_acc: 0.6800\n",
            "Epoch 3/15\n",
            "4500/4500 [==============================] - 12s 3ms/sample - loss: 0.4628 - acc: 0.7791 - val_loss: 0.4963 - val_acc: 0.7480\n",
            "Epoch 4/15\n",
            "4500/4500 [==============================] - 12s 3ms/sample - loss: 0.3618 - acc: 0.8376 - val_loss: 0.5183 - val_acc: 0.7480\n",
            "Epoch 5/15\n",
            "4500/4500 [==============================] - 12s 3ms/sample - loss: 0.2553 - acc: 0.8996 - val_loss: 0.5492 - val_acc: 0.7580\n",
            "Epoch 6/15\n",
            "4500/4500 [==============================] - 12s 3ms/sample - loss: 0.1552 - acc: 0.9424 - val_loss: 0.7053 - val_acc: 0.7480\n",
            "Epoch 7/15\n",
            "4500/4500 [==============================] - 12s 3ms/sample - loss: 0.0748 - acc: 0.9789 - val_loss: 0.8649 - val_acc: 0.7440\n",
            "Epoch 8/15\n",
            "4500/4500 [==============================] - 13s 3ms/sample - loss: 0.0462 - acc: 0.9878 - val_loss: 0.9546 - val_acc: 0.7460\n",
            "Epoch 9/15\n",
            "4500/4500 [==============================] - 12s 3ms/sample - loss: 0.0627 - acc: 0.9791 - val_loss: 0.9036 - val_acc: 0.7460\n",
            "Epoch 10/15\n",
            "4500/4500 [==============================] - 12s 3ms/sample - loss: 0.0399 - acc: 0.9889 - val_loss: 0.9439 - val_acc: 0.7260\n",
            "Epoch 11/15\n",
            "4500/4500 [==============================] - 12s 3ms/sample - loss: 0.0197 - acc: 0.9964 - val_loss: 1.2325 - val_acc: 0.7480\n",
            "Epoch 12/15\n",
            "4500/4500 [==============================] - 12s 3ms/sample - loss: 0.0329 - acc: 0.9896 - val_loss: 1.1956 - val_acc: 0.7440\n",
            "Epoch 13/15\n",
            "4500/4500 [==============================] - 12s 3ms/sample - loss: 0.0459 - acc: 0.9847 - val_loss: 1.2896 - val_acc: 0.7400\n",
            "Epoch 14/15\n",
            "4500/4500 [==============================] - 13s 3ms/sample - loss: 0.0263 - acc: 0.9924 - val_loss: 1.2073 - val_acc: 0.7260\n",
            "Epoch 15/15\n",
            "4500/4500 [==============================] - 12s 3ms/sample - loss: 0.0147 - acc: 0.9973 - val_loss: 1.2594 - val_acc: 0.7520\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xZ3ckVoqhBGO",
        "colab_type": "code",
        "outputId": "f8882262-1a8f-4518-d04d-453331e253d4",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 173
        }
      },
      "source": [
        "from sklearn.metrics import classification_report\n",
        "import numpy as np\n",
        "\n",
        "Y_test = np.argmax(yVal, axis=2).reshape((-1)) # Convert one-hot to index\n",
        "y_pred = full_model.predict(xVal)\n",
        "y_pred = y_pred.argmax(axis=2).reshape((-1))\n",
        "print(classification_report(Y_test, y_pred))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.78      0.77      0.78       527\n",
            "           1       0.75      0.76      0.75       473\n",
            "\n",
            "    accuracy                           0.77      1000\n",
            "   macro avg       0.77      0.77      0.77      1000\n",
            "weighted avg       0.77      0.77      0.77      1000\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TGJx4GBihMmI",
        "colab_type": "code",
        "outputId": "8e0eb749-b031-47b7-ecfa-44ce15902f1d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 337
        }
      },
      "source": [
        "from sklearn import metrics\n",
        "cm=metrics.confusion_matrix(Y_test,y_pred)\n",
        "\n",
        "import seaborn as sn\n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "df_cm = pd.DataFrame(cm, index = ['negative','positive'],\n",
        "                  columns = ['negative','postive'])\n",
        "plt.figure(figsize = (5,5))\n",
        "\n",
        "ax = sn.heatmap(df_cm, annot=True)\n",
        "\n",
        "bottom, top = ax.get_ylim()\n",
        "ax.set_ylim(bottom + 0.5, top - 0.5)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(2.0, 0.0)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 51
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAATQAAAEvCAYAAADLkD3HAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nZqnhgGZmqfH/5Dm5V6qpYhIAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 360x360 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "v6abhK91hPpg",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "t"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ktscwE43nzD0",
        "colab_type": "text"
      },
      "source": [
        "#Training"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "DHpt6Oa5w-BJ",
        "colab_type": "text"
      },
      "source": [
        "##General"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TUEXL75Hn0lH",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "xTrain,xVal,yTrain,yVal = getData(k=5000)\n",
        "full_model,attention_model = seq2seq()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VUfui_E1yhYh",
        "colab_type": "code",
        "outputId": "2ae99aea-7364-45a3-a902-b72196ca7197",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 454
        }
      },
      "source": [
        "training=full_model.fit(x=xTrain, y=yTrain,\n",
        "  epochs=10,\n",
        "  validation_split=0.1,\n",
        "  shuffle=True\n",
        ")"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow_core/python/ops/math_grad.py:1424: where (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use tf.where in 2.0, which has the same broadcast rule as np.where\n",
            "Train on 9000 samples, validate on 1000 samples\n",
            "Epoch 1/10\n",
            "9000/9000 [==============================] - 88s 10ms/sample - loss: 0.7957 - acc: 0.6771 - val_loss: 0.7836 - val_acc: 0.6900\n",
            "Epoch 2/10\n",
            "9000/9000 [==============================] - 85s 9ms/sample - loss: 0.7208 - acc: 0.7059 - val_loss: 0.7310 - val_acc: 0.7060\n",
            "Epoch 3/10\n",
            "9000/9000 [==============================] - 84s 9ms/sample - loss: 0.6834 - acc: 0.7306 - val_loss: 0.7074 - val_acc: 0.7050\n",
            "Epoch 4/10\n",
            "9000/9000 [==============================] - 84s 9ms/sample - loss: 0.6461 - acc: 0.7482 - val_loss: 0.6953 - val_acc: 0.7150\n",
            "Epoch 5/10\n",
            "9000/9000 [==============================] - 86s 10ms/sample - loss: 0.6170 - acc: 0.7588 - val_loss: 0.6935 - val_acc: 0.7170\n",
            "Epoch 6/10\n",
            "9000/9000 [==============================] - 85s 9ms/sample - loss: 0.5759 - acc: 0.7758 - val_loss: 0.7063 - val_acc: 0.7190\n",
            "Epoch 7/10\n",
            "9000/9000 [==============================] - 85s 9ms/sample - loss: 0.5418 - acc: 0.7882 - val_loss: 0.6867 - val_acc: 0.7280\n",
            "Epoch 8/10\n",
            "9000/9000 [==============================] - 85s 9ms/sample - loss: 0.4955 - acc: 0.8076 - val_loss: 0.7197 - val_acc: 0.7160\n",
            "Epoch 9/10\n",
            "9000/9000 [==============================] - 86s 10ms/sample - loss: 0.4390 - acc: 0.8322 - val_loss: 0.7815 - val_acc: 0.7120\n",
            "Epoch 10/10\n",
            "9000/9000 [==============================] - 85s 9ms/sample - loss: 0.3926 - acc: 0.8519 - val_loss: 0.8017 - val_acc: 0.7150\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6pfSyGEH2pAN",
        "colab_type": "code",
        "outputId": "37874ab0-bbd4-4027-d697-2dcd6ee5d706",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "type(training)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "tensorflow.python.keras.callbacks.History"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 22
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nzjS1ngDPaIe",
        "colab_type": "code",
        "outputId": "212726eb-bfbe-4f6f-e15c-453d8d62189d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 52
        }
      },
      "source": [
        "full_model.evaluate(xVal,\n",
        "  y=yVal)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "10000/10000 [==============================] - 21s 2ms/sample - loss: 0.8051 - acc: 0.7235\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[0.805129394197464, 0.7235]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 23
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2Jnd7P4aujAL",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "y_pred = attention_model.predict(xVal)\n",
        "y_pred_class = full_model.predict(xVal)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Sew9ome5wMpy",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "req = np.argmax(y_pred,axis = 2).reshape((-1))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zUIKU8_986qy",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "classes = np.argmax(y_pred_class,axis=2).reshape(-1)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JJh6ViUeK10c",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "imp_vecs = xVal"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KLxBTarVK1rq",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def v2w(vec):\n",
        "    if((vec==avg_vec).all()):\n",
        "        return 'UNK'\n",
        "    if(np.prod(vec) == 0):\n",
        "        return 'None'\n",
        "    for k in w2v_dict:\n",
        "        if((w2v_dict[k]==vec).all()):\n",
        "            return k\n",
        "    return 'OOV'"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ADResJqCK1oh",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "pos_words = []\n",
        "neg_words = []\n",
        "neu_words = []\n",
        "for i,vecs in enumerate(imp_vecs):\n",
        "    if(classes[i] == 0):\n",
        "        neg_words.append({'word':v2w(vecs)})\n",
        "    if(classes[i] == 1):\n",
        "        neu_words.append({'word':v2w(vecs)})\n",
        "    if(classes[i] == 2):\n",
        "        pos_words.append({'word':v2w(vecs)})"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1_MFDc43K1lN",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "neg_words = pd.DataFrame(neg_words)\n",
        "neu_words = pd.DataFrame(neu_words)\n",
        "pos_words = pd.DataFrame(pos_words)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "B2mAy_MBbIWi",
        "colab_type": "code",
        "outputId": "cfec64de-e67b-4fcd-939d-de9803f98b70",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 642
        }
      },
      "source": [
        "print(neg_words['word'].value_counts())\n",
        "print(neu_words['word'].value_counts())\n",
        "print(pos_words['word'].value_counts())"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "bleeding       98\n",
            "pain           80\n",
            "i              66\n",
            "my             65\n",
            "feel           61\n",
            "               ..\n",
            "usually         1\n",
            "alternative     1\n",
            "cleanse         1\n",
            "jerk            1\n",
            "clot            1\n",
            "Name: word, Length: 717, dtype: int64\n",
            "feel        12\n",
            "my           9\n",
            "started      8\n",
            "bleeding     6\n",
            "like         6\n",
            "            ..\n",
            "worked       1\n",
            "paxil        1\n",
            "taste        1\n",
            "helps        1\n",
            "pure         1\n",
            "Name: word, Length: 146, dtype: int64\n",
            "love          374\n",
            "life          367\n",
            "my            211\n",
            "works         209\n",
            "great         182\n",
            "             ... \n",
            "continued       1\n",
            "seltzer         1\n",
            "two             1\n",
            "eyed            1\n",
            "obsessions      1\n",
            "Name: word, Length: 1375, dtype: int64\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KiT_eXEJE2ii",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "y_pred_class = full_model.predict({\n",
        "    'ex':xVal,\n",
        "    'dx':_GO\n",
        "  })\n",
        "\n",
        "y_test = np.argmax(yVal,axis=2)\n",
        "y_pred_class = np.argmax(y_pred_class,axis=2)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8w-57_qTFCSX",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "y_test=y_test.reshape((-1))\n",
        "y_pred_class=y_pred_class.reshape((-1))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OiXqRuV-bISz",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from sklearn import metrics\n",
        "\n",
        "cm=metrics.confusion_matrix(y_test[:k],y_pred_class)\n",
        "\n",
        "import seaborn as sn\n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "df_cm = pd.DataFrame(cm, index = ['negative','neutral','positive'],\n",
        "                  columns = ['negative','neutral','postive'])\n",
        "plt.figure(figsize = (5,5))\n",
        "\n",
        "ax = sn.heatmap(df_cm, annot=True)\n",
        "\n",
        "bottom, top = ax.get_ylim()\n",
        "ax.set_ylim(bottom + 0.5, top - 0.5)\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "To_eGLQeXVFq",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import tensorflow as tf\n",
        "import keras\n",
        "\n",
        "config = tf.ConfigProto( device_count = {'GPU': 1} ) \n",
        "sess = tf.Session(config=config) \n",
        "keras.backend.set_session(sess)\n",
        "\n",
        "# from tensorflow.python.client import device_lib\n",
        "\n",
        "# print(device_lib.list_local_devices())"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "bT9aU9StJusa",
        "colab_type": "text"
      },
      "source": [
        "##Normalized Data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "inaUdN39KKDF",
        "colab": {}
      },
      "source": [
        "# sequence_length = 100\n",
        "# num_classes = 3"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "NRHvAzyLKKDh",
        "colab": {}
      },
      "source": [
        "# data = df['vectors'].to_numpy() #,np.array(df['r_label'])\n",
        "# data = pad_sequences(df['vectors'], maxlen=sequence_length,dtype='float32')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "VhY03pMQKKDt",
        "colab": {}
      },
      "source": [
        "# # labels = to_categorical(labels,num_classes=3)\n",
        "# # for l in labels\n",
        "\n",
        "# labels_e = {'_GO':np.array([0,0,0]),'neg':np.array([1,0,0]),'nut':np.array([0,1,0]),'pos':np.array([0,0,1])}\n",
        "# def foo2(x):\n",
        "#   if(x < 5):\n",
        "#     return [1,0,0]\n",
        "#   elif(x<7):\n",
        "#     return [0,1,0]\n",
        "#   return [0,0,1]\n",
        "\n",
        "# labels = df.rating.apply(foo2)\n",
        "# labels = np.array(list(labels))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yB9ndKxyKJ-7",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "xTrain,xVal,yTrain,yVal = getData(k=10000)\n",
        "full_model,attention_model = seq2seq()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CqbrJcU0K2EX",
        "colab_type": "code",
        "outputId": "fc73f873-7b36-4d80-a144-712d6b01e908",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "xVal.shape"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(8400, 100, 50)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 112
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "P-nA917JKutJ",
        "colab_type": "code",
        "outputId": "835284aa-f09a-4cb7-9f6b-540e7ed61ec4",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 382
        }
      },
      "source": [
        "training=full_model.fit(x=xTrain,\n",
        "  y=yTrain,\n",
        "  epochs=10,\n",
        "  validation_split=0.1,\n",
        "  shuffle=True\n",
        ")"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 9000 samples, validate on 1000 samples\n",
            "Epoch 1/10\n",
            "9000/9000 [==============================] - 85s 9ms/sample - loss: 0.9115 - acc: 0.6133 - val_loss: 0.8558 - val_acc: 0.6010\n",
            "Epoch 2/10\n",
            "9000/9000 [==============================] - 85s 9ms/sample - loss: 0.7337 - acc: 0.6841 - val_loss: 0.8703 - val_acc: 0.6100\n",
            "Epoch 3/10\n",
            "9000/9000 [==============================] - 85s 9ms/sample - loss: 0.6469 - acc: 0.7301 - val_loss: 0.9072 - val_acc: 0.6150\n",
            "Epoch 4/10\n",
            "9000/9000 [==============================] - 86s 10ms/sample - loss: 0.5682 - acc: 0.7684 - val_loss: 0.9604 - val_acc: 0.5880\n",
            "Epoch 5/10\n",
            "9000/9000 [==============================] - 85s 9ms/sample - loss: 0.4907 - acc: 0.8013 - val_loss: 1.0351 - val_acc: 0.6010\n",
            "Epoch 6/10\n",
            "9000/9000 [==============================] - 85s 9ms/sample - loss: 0.4119 - acc: 0.8460 - val_loss: 1.1464 - val_acc: 0.5870\n",
            "Epoch 7/10\n",
            "9000/9000 [==============================] - 85s 9ms/sample - loss: 0.3400 - acc: 0.8764 - val_loss: 1.2578 - val_acc: 0.5850\n",
            "Epoch 8/10\n",
            "9000/9000 [==============================] - 86s 10ms/sample - loss: 0.2736 - acc: 0.9063 - val_loss: 1.3801 - val_acc: 0.5890\n",
            "Epoch 9/10\n",
            "9000/9000 [==============================] - 85s 9ms/sample - loss: 0.2252 - acc: 0.9247 - val_loss: 1.4690 - val_acc: 0.5830\n",
            "Epoch 10/10\n",
            "9000/9000 [==============================] - 85s 9ms/sample - loss: 0.1686 - acc: 0.9483 - val_loss: 1.6313 - val_acc: 0.5830\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "outputId": "1dffd77a-c79b-4f0b-bb1c-94a2e374adba",
        "id": "_wvedApuLGXc",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 52
        }
      },
      "source": [
        "full_model.evaluate(xVal,\n",
        "  y=yVal)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "5000/5000 [==============================] - 10s 2ms/sample - loss: 1.2979 - acc: 0.5596\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[1.29794847946167, 0.5596]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 93
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "i64hvTo1LGX7",
        "colab": {}
      },
      "source": [
        "y_pred = attention_model.predict(xVal)\n",
        "y_pred_class = full_model.predict(xVal)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "hJKiBcueLGYH",
        "colab": {}
      },
      "source": [
        "req = np.argmax(y_pred,axis = 2).reshape((-1))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "xhfmTreFLGYQ",
        "colab": {}
      },
      "source": [
        "classes = np.argmax(y_pred_class,axis=2).reshape(-1)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "P5_7EvJELGYZ",
        "colab": {}
      },
      "source": [
        "imp_vecs = xVal"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "1U3ZC0idLGYi",
        "colab": {}
      },
      "source": [
        "def v2w(vec):\n",
        "    if((vec==avg_vec).all()):\n",
        "        return 'UNK'\n",
        "    if(np.prod(vec) == 0):\n",
        "        return 'None'\n",
        "    for k in w2v_dict:\n",
        "        if((w2v_dict[k]==vec).all()):\n",
        "            return k\n",
        "    return 'OOV'"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "hBcd-_tyLGYp",
        "colab": {}
      },
      "source": [
        "pos_words = []\n",
        "neg_words = []\n",
        "neu_words = []\n",
        "for i,vecs in enumerate(imp_vecs):\n",
        "    if(classes[i] == 0):\n",
        "        neg_words.append({'word':v2w(vecs)})\n",
        "    if(classes[i] == 1):\n",
        "        neu_words.append({'word':v2w(vecs)})\n",
        "    if(classes[i] == 2):\n",
        "        pos_words.append({'word':v2w(vecs)})"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "T5XGsV7ILGYv",
        "colab": {}
      },
      "source": [
        "neg_words = pd.DataFrame(neg_words)\n",
        "neu_words = pd.DataFrame(neu_words)\n",
        "pos_words = pd.DataFrame(pos_words)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "outputId": "66f53d8e-fd1f-4ebd-9b5a-2ed630baa2bc",
        "id": "xTXcXismLGY2",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 642
        }
      },
      "source": [
        "print(neg_words['word'].value_counts())\n",
        "print(neu_words['word'].value_counts())\n",
        "print(pos_words['word'].value_counts())"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "i               63\n",
            "bad             56\n",
            "life            55\n",
            "horrible        47\n",
            "my              35\n",
            "                ..\n",
            "knees            1\n",
            "gut              1\n",
            "dramatically     1\n",
            "medicated        1\n",
            "true             1\n",
            "Name: word, Length: 504, dtype: int64\n",
            "i          88\n",
            "eat        39\n",
            "started    34\n",
            "UNK        34\n",
            "bad        34\n",
            "           ..\n",
            "falling     1\n",
            "just        1\n",
            "dove        1\n",
            "rough       1\n",
            "sweaty      1\n",
            "Name: word, Length: 537, dtype: int64\n",
            "life          145\n",
            "love           89\n",
            "years          87\n",
            "great          69\n",
            "i              36\n",
            "             ... \n",
            "uplifting       1\n",
            "bile            1\n",
            "particular      1\n",
            "weight          1\n",
            "hope            1\n",
            "Name: word, Length: 439, dtype: int64\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "bhJ_HBoLLGY-",
        "colab": {}
      },
      "source": [
        "y_pred_class = full_model.predict(xVal)\n",
        "\n",
        "y_pred_class = np.argmax(y_pred_class,axis=2)\n",
        "y_pred_class=y_pred_class.reshape((-1))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "Cp_MK2vWLGZE",
        "colab": {}
      },
      "source": [
        "y_test = np.argmax(yVal,axis=2)\n",
        "y_test=y_test.reshape((-1))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "outputId": "7023d75a-bfd2-46fa-f5b2-6a94db9705e9",
        "id": "UMA2TR_4LGZM",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 337
        }
      },
      "source": [
        "from sklearn import metrics\n",
        "\n",
        "cm=metrics.confusion_matrix(y_test[:k],y_pred_class)\n",
        "\n",
        "import seaborn as sn\n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "df_cm = pd.DataFrame(cm, index = ['negative','neutral','positive'],\n",
        "                  columns = ['negative','neutral','postive'])\n",
        "plt.figure(figsize = (5,5))\n",
        "\n",
        "ax = sn.heatmap(df_cm, annot=True)\n",
        "\n",
        "bottom, top = ax.get_ylim()\n",
        "ax.set_ylim(bottom + 0.5, top - 0.5)\n"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(3.0, 0.0)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 110
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAToAAAEvCAYAAADVWQ10AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nSjmeBjqllONpoFNKOZ4GOqWU4/0fw3NNH6YfuiUAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 360x360 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "raVipwd0ljBO",
        "colab_type": "text"
      },
      "source": [
        "## Binary Classification"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Y8XpvJjPllT_",
        "colab_type": "code",
        "outputId": "99ee5eb3-101e-4b96-f2d5-8faddc4daaed",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 191
        }
      },
      "source": [
        "xTrain,xVal,yTrain,yVal,_GO,_GO_test = getData(classes=2,k=1000)\n",
        "full_model,attention_model = seq2seq(classes=2)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:17: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame\n",
            "\n",
            "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "(1000, 100, 50) (53766, 100, 50)\n",
            "(?, 100, 128)\n",
            "(?, 64)\n",
            "(?, 64)\n",
            "(?, 1, 128)\n",
            "(?, 1, 100)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "outputId": "1c826fd2-6f45-4c45-ac9e-15d0476ed576",
        "id": "ba7TMc7WxvF8",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 382
        }
      },
      "source": [
        "training=full_model.fit(x={\n",
        "    'ex':xTrain,\n",
        "    'dx':_GO\n",
        "  },\n",
        "  y=yTrain,\n",
        "  epochs=10,\n",
        "  validation_split=0.1,\n",
        "  shuffle=True\n",
        ")"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 900 samples, validate on 100 samples\n",
            "Epoch 1/10\n",
            "900/900 [==============================] - 11s 12ms/sample - loss: 0.6122 - acc: 0.6878 - val_loss: 0.5769 - val_acc: 0.7500\n",
            "Epoch 2/10\n",
            "900/900 [==============================] - 9s 10ms/sample - loss: 0.5922 - acc: 0.6878 - val_loss: 0.5130 - val_acc: 0.7600\n",
            "Epoch 3/10\n",
            "900/900 [==============================] - 10s 11ms/sample - loss: 0.5759 - acc: 0.7011 - val_loss: 0.4897 - val_acc: 0.7500\n",
            "Epoch 4/10\n",
            "900/900 [==============================] - 9s 10ms/sample - loss: 0.5532 - acc: 0.7267 - val_loss: 0.5027 - val_acc: 0.7800\n",
            "Epoch 5/10\n",
            "900/900 [==============================] - 9s 10ms/sample - loss: 0.5262 - acc: 0.7356 - val_loss: 0.4738 - val_acc: 0.7600\n",
            "Epoch 6/10\n",
            "900/900 [==============================] - 9s 10ms/sample - loss: 0.5122 - acc: 0.7544 - val_loss: 0.4425 - val_acc: 0.7800\n",
            "Epoch 7/10\n",
            "900/900 [==============================] - 9s 10ms/sample - loss: 0.4787 - acc: 0.7700 - val_loss: 0.4271 - val_acc: 0.7600\n",
            "Epoch 8/10\n",
            "900/900 [==============================] - 9s 10ms/sample - loss: 0.4558 - acc: 0.7867 - val_loss: 0.4364 - val_acc: 0.7500\n",
            "Epoch 9/10\n",
            "900/900 [==============================] - 9s 10ms/sample - loss: 0.4182 - acc: 0.8089 - val_loss: 0.4622 - val_acc: 0.7500\n",
            "Epoch 10/10\n",
            "900/900 [==============================] - 9s 10ms/sample - loss: 0.3986 - acc: 0.8167 - val_loss: 0.4359 - val_acc: 0.7800\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "outputId": "93201609-4a4e-4b20-9c2b-4e2e33bb8d9c",
        "id": "1MenI49-xvGc",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 52
        }
      },
      "source": [
        "full_model.evaluate({\n",
        "    'ex':xVal,\n",
        "    'dx':_GO_test\n",
        "  },\n",
        "  y=yVal)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "53766/53766 [==============================] - 121s 2ms/sample - loss: 0.6211 - acc: 0.7204\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[0.6211253218944507, 0.7204367]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 197
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "E7NdThdLxvGm",
        "colab": {}
      },
      "source": [
        "y_pred = attention_model.predict({\n",
        "    'ex':xVal,\n",
        "    'dx':_GO_test\n",
        "  })\n",
        "y_pred_class = full_model.predict({\n",
        "    'ex':xVal,\n",
        "    'dx':_GO_test\n",
        "  })"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "oOSbi8ewxvGv",
        "colab": {}
      },
      "source": [
        "req = np.argmax(y_pred,axis = 2).reshape((-1))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "TutG_2PnxvG2",
        "colab": {}
      },
      "source": [
        "classes = np.argmax(y_pred_class,axis=2).reshape(-1)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "gOamS0eOxvG-",
        "colab": {}
      },
      "source": [
        "imp_vecs = xVal"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "Df2ggHPOxvHE",
        "colab": {}
      },
      "source": [
        "def v2w(vec):\n",
        "    if((vec==avg_vec).all()):\n",
        "        return 'UNK'\n",
        "    if(vec.sum() == 0):\n",
        "        return 'None'\n",
        "    for k in w2v_dict:\n",
        "        if((w2v_dict[k]==vec).all()):\n",
        "            return k\n",
        "    return 'OOV'"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "Ge1ZQiksxvHK",
        "outputId": "37ee2839-97ae-4111-9743-aa2943b420d2",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 323
        }
      },
      "source": [
        "pos_words = []\n",
        "neg_words = []\n",
        "neu_words = []\n",
        "for i,vecs in enumerate(imp_vecs):\n",
        "    if(classes[i] == 0):\n",
        "        neg_words.append({'word':v2w(vecs)})\n",
        "    if(classes[i] == 1):\n",
        "        pos_words.append({'word':v2w(vecs)})"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "error",
          "ename": "KeyboardInterrupt",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-214-3bde68125324>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      6\u001b[0m         \u001b[0mneg_words\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m{\u001b[0m\u001b[0;34m'word'\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0mv2w\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mvecs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m}\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      7\u001b[0m     \u001b[0;32mif\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mclasses\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 8\u001b[0;31m         \u001b[0mpos_words\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m{\u001b[0m\u001b[0;34m'word'\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0mv2w\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mvecs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m}\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
            "\u001b[0;32m<ipython-input-212-fdcf6cb320c1>\u001b[0m in \u001b[0;36mv2w\u001b[0;34m(vec)\u001b[0m\n\u001b[1;32m      5\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0;34m'None'\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      6\u001b[0m     \u001b[0;32mfor\u001b[0m \u001b[0mk\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mw2v_dict\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 7\u001b[0;31m         \u001b[0;32mif\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mw2v_dict\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mk\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m==\u001b[0m\u001b[0mvec\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mall\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      8\u001b[0m             \u001b[0;32mreturn\u001b[0m \u001b[0mk\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      9\u001b[0m     \u001b[0;32mreturn\u001b[0m \u001b[0;34m'OOV'\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "Sc7aKGJ8xvHQ",
        "colab": {}
      },
      "source": [
        "neg_words = pd.DataFrame(neg_words)\n",
        "pos_words = pd.DataFrame(pos_words)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "c2tDLpyuxvHX",
        "colab": {}
      },
      "source": [
        "print(neg_words['word'].value_counts())\n",
        "print(pos_words['word'].value_counts())"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "rXMiLH6kxvHe",
        "colab": {}
      },
      "source": [
        "y_pred_class = full_model.predict({\n",
        "    'ex':xVal,\n",
        "    'dx':_GO_test\n",
        "  })\n",
        "\n",
        "y_pred_class = np.argmax(y_pred_class,axis=2)\n",
        "y_pred_class=y_pred_class.reshape((-1))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "2wz3mpl4xvHn",
        "colab": {}
      },
      "source": [
        "y_test = np.argmax(yVal,axis=2)\n",
        "y_test=y_test.reshape((-1))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "outputId": "31a0be30-5564-4584-bf90-782584172cbe",
        "id": "GRn4_otexvHt",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 337
        }
      },
      "source": [
        "from sklearn import metrics\n",
        "\n",
        "cm=metrics.confusion_matrix(y_test,y_pred_class)\n",
        "\n",
        "import seaborn as sn\n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "df_cm = pd.DataFrame(cm, index = ['negative','positive'],\n",
        "                  columns = ['negative','postive'])\n",
        "plt.figure(figsize = (5,5))\n",
        "\n",
        "ax = sn.heatmap(df_cm, annot=True)\n",
        "\n",
        "bottom, top = ax.get_ylim()\n",
        "ax.set_ylim(bottom + 0.5, top - 0.5)\n"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(2.0, 0.0)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 218
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAUAAAAEvCAYAAAA0MhglAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAVBEYksBUERiSwFQRGJLAVBEYuv/AZ0MgfWHRIuMAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 360x360 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "E3jkDEP5xvH5",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YudSzKFl-c19",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vofciefc-ai6",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}