l10c01_nlp_lstms_with_reviews_subwords_dataset.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "punL79CN7Ox6"
      },
      "source": [
        "##### Copyright 2020 The TensorFlow Authors."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "cellView": "form",
        "id": "_ckMIh7O7s6D"
      },
      "outputs": [],
      "source": [
        "#@title Licensed under the Apache License, Version 2.0 (the \"License\");\n",
        "# you may not use this file except in compliance with the License.\n",
        "# You may obtain a copy of the License at\n",
        "#\n",
        "# https://www.apache.org/licenses/LICENSE-2.0\n",
        "#\n",
        "# Unless required by applicable law or agreed to in writing, software\n",
        "# distributed under the License is distributed on an \"AS IS\" BASIS,\n",
        "# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n",
        "# See the License for the specific language governing permissions and\n",
        "# limitations under the License."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "hAclqSm3OOml"
      },
      "source": [
        "# Using LSTMs with the subwords dataset\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "S5Uhzt6vVIB2"
      },
      "source": [
        "<table class=\"tfo-notebook-buttons\" align=\"left\">\n",
        "  <td>\n",
        "    <a target=\"_blank\" href=\"https://colab.research.google.com/github/tensorflow/examples/blob/master/courses/udacity_intro_to_tensorflow_for_deep_learning/l10c01_nlp_lstms_with_reviews_subwords_dataset.ipynb\"><img src=\"https://www.tensorflow.org/images/colab_logo_32px.png\" />Run in Google Colab</a>\n",
        "  </td>\n",
        "  <td>\n",
        "    <a target=\"_blank\" href=\"https://github.com/tensorflow/examples/blob/master/courses/udacity_intro_to_tensorflow_for_deep_learning/l10c01_nlp_lstms_with_reviews_subwords_dataset.ipynb\"><img src=\"https://www.tensorflow.org/images/GitHub-Mark-32px.png\" />View source on GitHub</a>\n",
        "  </td>\n",
        "</table>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "KTVx8__oGR9J"
      },
      "source": [
        "In this colab, you'll compare the results of using a model with an Embedding layer and then adding bidirectional LSTM layers.\n",
        "\n",
        "You'll work with the dataset of subwords for the combined Yelp and Amazon reviews.\n",
        "\n",
        "You'll use your models to predict the sentiment of new reviews."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {
        "id": "L62G7LTwNzoD"
      },
      "outputs": [],
      "source": [
        "import tensorflow as tf\n",
        "\n",
        "from tensorflow.keras.preprocessing.sequence import pad_sequences"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "hLcl0QHvDjTV"
      },
      "source": [
        "# Get the dataset\n",
        "\n",
        "Start by getting the dataset containing Amazon and Yelp reviews, with their related sentiment (1 for positive, 0 for negative). This dataset was originally extracted from [here](https://www.kaggle.com/marklvl/sentiment-labelled-sentences-data-set).\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "metadata": {
        "id": "nCOtiRJZbxCH",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "bddc1fc3-446e-4fbc-c02e-0c9afefe1b46"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "--2021-12-10 03:31:21--  https://drive.google.com/uc?id=13ySLC_ue6Umt9RJYSeM2t-V0kCv-4C-P\n",
            "Resolving drive.google.com (drive.google.com)... 108.177.15.113, 108.177.15.102, 108.177.15.101, ...\n",
            "Connecting to drive.google.com (drive.google.com)|108.177.15.113|:443... connected.\n",
            "HTTP request sent, awaiting response... 302 Moved Temporarily\n",
            "Location: https://doc-08-ak-docs.googleusercontent.com/docs/securesc/ha0ro937gcuc7l7deffksulhg5h7mbp1/0pq9murkfdcgi6s3nf6i144reoh4gqsv/1639107075000/11118900490791463723/*/13ySLC_ue6Umt9RJYSeM2t-V0kCv-4C-P [following]\n",
            "Warning: wildcards not supported in HTTP.\n",
            "--2021-12-10 03:31:22--  https://doc-08-ak-docs.googleusercontent.com/docs/securesc/ha0ro937gcuc7l7deffksulhg5h7mbp1/0pq9murkfdcgi6s3nf6i144reoh4gqsv/1639107075000/11118900490791463723/*/13ySLC_ue6Umt9RJYSeM2t-V0kCv-4C-P\n",
            "Resolving doc-08-ak-docs.googleusercontent.com (doc-08-ak-docs.googleusercontent.com)... 66.102.1.132, 2a00:1450:400c:c06::84\n",
            "Connecting to doc-08-ak-docs.googleusercontent.com (doc-08-ak-docs.googleusercontent.com)|66.102.1.132|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 127831 (125K) [text/csv]\n",
            "Saving to: ‘/tmp/sentiment.csv’\n",
            "\n",
            "/tmp/sentiment.csv  100%[===================>] 124.83K  --.-KB/s    in 0.003s  \n",
            "\n",
            "2021-12-10 03:31:22 (43.0 MB/s) - ‘/tmp/sentiment.csv’ saved [127831/127831]\n",
            "\n"
          ]
        }
      ],
      "source": [
        "!wget --no-check-certificate \\\n",
        "    https://drive.google.com/uc?id=13ySLC_ue6Umt9RJYSeM2t-V0kCv-4C-P -O /tmp/sentiment.csv"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {
        "id": "XuqER_KMD-xS"
      },
      "outputs": [],
      "source": [
        "import pandas as pd\n",
        "\n",
        "dataset = pd.read_csv('/tmp/sentiment.csv')\n",
        "\n",
        "# Extract out sentences and labels\n",
        "sentences = dataset['text'].tolist()\n",
        "labels = dataset['sentiment'].tolist()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {
        "id": "Tbsx1T2CXPNO",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "3de513b3-aa91-491a-8cf1-bbd8da655e32"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "So there is no way for me to plug it in here in the US unless I go by a converter.\n",
            "0\n",
            "\n",
            "\n",
            "Good case Excellent value.\n",
            "1\n",
            "\n",
            "\n"
          ]
        }
      ],
      "source": [
        "# Print some example sentences and labels\n",
        "for x in range(2):\n",
        "  print(sentences[x])\n",
        "  print(labels[x])\n",
        "  print(\"\\n\")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "33AthPiALFZK"
      },
      "source": [
        "#Create a subwords dataset\n",
        "\n",
        "We will use the Amazon and Yelp reviews dataset with tensorflow_datasets's SubwordTextEncoder functionality. \n",
        "\n",
        "SubwordTextEncoder.build_from_corpus() will create a tokenizer for us. You could also use this functionality to get subwords from a much larger corpus of text as well, but we'll just use our existing dataset here.\n",
        "\n",
        "We'll create a subword vocab_size of only the 1,000 most common subwords, as well as cutting off each subword to be at most 5 characters.\n",
        "\n",
        "Check out the related documentation for the the subword text encoder [here](https://www.tensorflow.org/datasets/api_docs/python/tfds/features/text/SubwordTextEncoder#build_from_corpus)."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "metadata": {
        "id": "6NaicNCcLYyf",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "9f174460-8ce2-4c3c-d7db-2d83424e651f"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Vocab size is  999\n"
          ]
        }
      ],
      "source": [
        "import tensorflow_datasets as tfds\n",
        "\n",
        "vocab_size = 1000\n",
        "tokenizer = tfds.deprecated.text.SubwordTextEncoder.build_from_corpus(sentences, vocab_size, max_subword_length=5)\n",
        "\n",
        "# How big is the vocab size?\n",
        "print(\"Vocab size is \", tokenizer.vocab_size)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 7,
      "metadata": {
        "id": "xvRVoeIVLevh",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "8d1359b9-d56f-4c12-98e9-af030ccf298d"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "I have to jiggle the plug to get it to line up right to get decent volume.\n",
            "[4, 31, 6, 849, 162, 450, 12, 1, 600, 438, 775, 6, 175, 14, 6, 55, 213, 159, 474, 775, 6, 175, 614, 380, 295, 148, 72, 789]\n"
          ]
        }
      ],
      "source": [
        "# Check that the tokenizer works appropriately\n",
        "num = 5\n",
        "print(sentences[num])\n",
        "encoded = tokenizer.encode(sentences[num])\n",
        "print(encoded)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 8,
      "metadata": {
        "id": "G_vacTCifklV",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "f02fa1a3-5f4b-4132-a597-b43bdb2d30f1"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "I \n",
            "have \n",
            "to \n",
            "j\n",
            "ig\n",
            "gl\n",
            "e \n",
            "the \n",
            "pl\n",
            "ug\n",
            " \n",
            "to \n",
            "get \n",
            "it \n",
            "to \n",
            "li\n",
            "ne \n",
            "up \n",
            "right\n",
            " \n",
            "to \n",
            "get \n",
            "dec\n",
            "ent \n",
            "vo\n",
            "lu\n",
            "me\n",
            ".\n"
          ]
        }
      ],
      "source": [
        "# Separately print out each subword, decoded\n",
        "for i in encoded:\n",
        "  print(tokenizer.decode([i]))"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "cT528cptLupl"
      },
      "source": [
        "## Replace sentence data with encoded subwords\n",
        "\n",
        "Now, we'll create the sequences to be used for training by actually encoding each of the individual sentences. This is equivalent to `text_to_sequences` with the `Tokenizer` we used in earlier exercises."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 9,
      "metadata": {
        "id": "lkseMhxjL09F"
      },
      "outputs": [],
      "source": [
        "for i, sentence in enumerate(sentences):\n",
        "  sentences[i] = tokenizer.encode(sentence)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 10,
      "metadata": {
        "id": "y21yRuzmL43U",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "429bccec-7719-402e-9b62-391b7673f38b"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[4, 31, 6, 849, 162, 450, 12, 1, 600, 438, 775, 6, 175, 14, 6, 55, 213, 159, 474, 775, 6, 175, 614, 380, 295, 148, 72, 789]\n"
          ]
        }
      ],
      "source": [
        "# Check the sentences are appropriately replaced\n",
        "print(sentences[5])"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8HrcPHESMBMs"
      },
      "source": [
        "## Final pre-processing\n",
        "\n",
        "Before training, we still need to pad the sequences, as well as split into training and test sets."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 11,
      "metadata": {
        "id": "50-hTsogLSL-"
      },
      "outputs": [],
      "source": [
        "import numpy as np\n",
        "\n",
        "max_length = 50\n",
        "trunc_type='post'\n",
        "padding_type='post'\n",
        "\n",
        "# Pad all sequences\n",
        "sequences_padded = pad_sequences(sentences, maxlen=max_length, \n",
        "                                 padding=padding_type, truncating=trunc_type)\n",
        "\n",
        "# Separate out the sentences and labels into training and test sets\n",
        "training_size = int(len(sentences) * 0.8)\n",
        "\n",
        "training_sequences = sequences_padded[0:training_size]\n",
        "testing_sequences = sequences_padded[training_size:]\n",
        "training_labels = labels[0:training_size]\n",
        "testing_labels = labels[training_size:]\n",
        "\n",
        "# Make labels into numpy arrays for use with the network later\n",
        "training_labels_final = np.array(training_labels)\n",
        "testing_labels_final = np.array(testing_labels)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "PahZm7YEQ8EI"
      },
      "source": [
        "# Create the model using an Embedding"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 12,
      "metadata": {
        "id": "c_nyQeI0RCCv",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "09af1f37-b766-469c-b6ff-5e9ea581492c"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " embedding (Embedding)       (None, 50, 16)            16000     \n",
            "                                                                 \n",
            " global_average_pooling1d (G  (None, 16)               0         \n",
            " lobalAveragePooling1D)                                          \n",
            "                                                                 \n",
            " dense (Dense)               (None, 6)                 102       \n",
            "                                                                 \n",
            " dense_1 (Dense)             (None, 1)                 7         \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 16,109\n",
            "Trainable params: 16,109\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "embedding_dim = 16\n",
        "\n",
        "model = tf.keras.Sequential([\n",
        "    tf.keras.layers.Embedding(vocab_size, embedding_dim, input_length=max_length),\n",
        "    tf.keras.layers.GlobalAveragePooling1D(), \n",
        "    tf.keras.layers.Dense(6, activation='relu'),\n",
        "    tf.keras.layers.Dense(1, activation='sigmoid')\n",
        "])\n",
        "\n",
        "model.summary()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3WRXrx8BRO2L"
      },
      "source": [
        "# Train the model"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 13,
      "metadata": {
        "id": "oBKyVYvxRQ_9",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "2f512a02-111c-4830-cca6-c1a7a47a553c"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/30\n",
            "50/50 [==============================] - 3s 11ms/step - loss: 0.6916 - accuracy: 0.5248 - val_loss: 0.6974 - val_accuracy: 0.4236\n",
            "Epoch 2/30\n",
            "50/50 [==============================] - 0s 7ms/step - loss: 0.6877 - accuracy: 0.5443 - val_loss: 0.6966 - val_accuracy: 0.4586\n",
            "Epoch 3/30\n",
            "50/50 [==============================] - 0s 7ms/step - loss: 0.6822 - accuracy: 0.5712 - val_loss: 0.6906 - val_accuracy: 0.5113\n",
            "Epoch 4/30\n",
            "50/50 [==============================] - 0s 7ms/step - loss: 0.6704 - accuracy: 0.6303 - val_loss: 0.6768 - val_accuracy: 0.5664\n",
            "Epoch 5/30\n",
            "50/50 [==============================] - 0s 7ms/step - loss: 0.6524 - accuracy: 0.6999 - val_loss: 0.6670 - val_accuracy: 0.5865\n",
            "Epoch 6/30\n",
            "50/50 [==============================] - 0s 6ms/step - loss: 0.6277 - accuracy: 0.7313 - val_loss: 0.6477 - val_accuracy: 0.6591\n",
            "Epoch 7/30\n",
            "50/50 [==============================] - 0s 7ms/step - loss: 0.5952 - accuracy: 0.7891 - val_loss: 0.6122 - val_accuracy: 0.7719\n",
            "Epoch 8/30\n",
            "50/50 [==============================] - 0s 7ms/step - loss: 0.5582 - accuracy: 0.8192 - val_loss: 0.5961 - val_accuracy: 0.7243\n",
            "Epoch 9/30\n",
            "50/50 [==============================] - 0s 7ms/step - loss: 0.5218 - accuracy: 0.8236 - val_loss: 0.5819 - val_accuracy: 0.7018\n",
            "Epoch 10/30\n",
            "50/50 [==============================] - 0s 7ms/step - loss: 0.4837 - accuracy: 0.8443 - val_loss: 0.5652 - val_accuracy: 0.7218\n",
            "Epoch 11/30\n",
            "50/50 [==============================] - 0s 7ms/step - loss: 0.4488 - accuracy: 0.8537 - val_loss: 0.5428 - val_accuracy: 0.7419\n",
            "Epoch 12/30\n",
            "50/50 [==============================] - 0s 7ms/step - loss: 0.4163 - accuracy: 0.8650 - val_loss: 0.5214 - val_accuracy: 0.7444\n",
            "Epoch 13/30\n",
            "50/50 [==============================] - 0s 7ms/step - loss: 0.3877 - accuracy: 0.8701 - val_loss: 0.5159 - val_accuracy: 0.7469\n",
            "Epoch 14/30\n",
            "50/50 [==============================] - 0s 7ms/step - loss: 0.3627 - accuracy: 0.8745 - val_loss: 0.5131 - val_accuracy: 0.7469\n",
            "Epoch 15/30\n",
            "50/50 [==============================] - 0s 7ms/step - loss: 0.3400 - accuracy: 0.8801 - val_loss: 0.4990 - val_accuracy: 0.7544\n",
            "Epoch 16/30\n",
            "50/50 [==============================] - 0s 7ms/step - loss: 0.3223 - accuracy: 0.8883 - val_loss: 0.5128 - val_accuracy: 0.7519\n",
            "Epoch 17/30\n",
            "50/50 [==============================] - 0s 7ms/step - loss: 0.3048 - accuracy: 0.8964 - val_loss: 0.5252 - val_accuracy: 0.7343\n",
            "Epoch 18/30\n",
            "50/50 [==============================] - 0s 6ms/step - loss: 0.2891 - accuracy: 0.9058 - val_loss: 0.5039 - val_accuracy: 0.7594\n",
            "Epoch 19/30\n",
            "50/50 [==============================] - 1s 12ms/step - loss: 0.2749 - accuracy: 0.9033 - val_loss: 0.5043 - val_accuracy: 0.7594\n",
            "Epoch 20/30\n",
            "50/50 [==============================] - 0s 7ms/step - loss: 0.2619 - accuracy: 0.9058 - val_loss: 0.5038 - val_accuracy: 0.7544\n",
            "Epoch 21/30\n",
            "50/50 [==============================] - 0s 7ms/step - loss: 0.2509 - accuracy: 0.9102 - val_loss: 0.5073 - val_accuracy: 0.7519\n",
            "Epoch 22/30\n",
            "50/50 [==============================] - 0s 7ms/step - loss: 0.2429 - accuracy: 0.9159 - val_loss: 0.5219 - val_accuracy: 0.7494\n",
            "Epoch 23/30\n",
            "50/50 [==============================] - 0s 6ms/step - loss: 0.2299 - accuracy: 0.9190 - val_loss: 0.5241 - val_accuracy: 0.7519\n",
            "Epoch 24/30\n",
            "50/50 [==============================] - 0s 8ms/step - loss: 0.2222 - accuracy: 0.9284 - val_loss: 0.5340 - val_accuracy: 0.7494\n",
            "Epoch 25/30\n",
            "50/50 [==============================] - 0s 7ms/step - loss: 0.2123 - accuracy: 0.9291 - val_loss: 0.5399 - val_accuracy: 0.7469\n",
            "Epoch 26/30\n",
            "50/50 [==============================] - 0s 7ms/step - loss: 0.2060 - accuracy: 0.9291 - val_loss: 0.5547 - val_accuracy: 0.7419\n",
            "Epoch 27/30\n",
            "50/50 [==============================] - 0s 7ms/step - loss: 0.1973 - accuracy: 0.9347 - val_loss: 0.5537 - val_accuracy: 0.7444\n",
            "Epoch 28/30\n",
            "50/50 [==============================] - 0s 7ms/step - loss: 0.1902 - accuracy: 0.9397 - val_loss: 0.5493 - val_accuracy: 0.7594\n",
            "Epoch 29/30\n",
            "50/50 [==============================] - 0s 7ms/step - loss: 0.1854 - accuracy: 0.9410 - val_loss: 0.5832 - val_accuracy: 0.7419\n",
            "Epoch 30/30\n",
            "50/50 [==============================] - 0s 7ms/step - loss: 0.1772 - accuracy: 0.9422 - val_loss: 0.5747 - val_accuracy: 0.7469\n"
          ]
        }
      ],
      "source": [
        "num_epochs = 30\n",
        "model.compile(loss='binary_crossentropy',optimizer='adam',metrics=['accuracy'])\n",
        "history = model.fit(training_sequences, training_labels_final, epochs=num_epochs, validation_data=(testing_sequences, testing_labels_final))\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "HhLPbUl2AZ0y"
      },
      "source": [
        "# Plot the accuracy and loss"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 14,
      "metadata": {
        "id": "jzBM1PpJAYfD",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 541
        },
        "outputId": "2260fa39-1162-42e0-eea6-27ba4c95a828"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ],
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "\n",
        "def plot_graphs(history, string):\n",
        "  plt.plot(history.history[string])\n",
        "  plt.plot(history.history['val_'+string])\n",
        "  plt.xlabel(\"Epochs\")\n",
        "  plt.ylabel(string)\n",
        "  plt.legend([string, 'val_'+string])\n",
        "  plt.show()\n",
        "  \n",
        "plot_graphs(history, \"accuracy\")\n",
        "plot_graphs(history, \"loss\")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Fwr5inBiWffb"
      },
      "source": [
        "# Define a function to predict the sentiment of reviews\n",
        "\n",
        "We'll be creating models with some differences and will use each model to predict the sentiment of some new reviews.\n",
        "\n",
        "To save time, create a function that will take in a model and some new reviews, and print out the sentiment of each reviews.\n",
        "\n",
        "The higher the sentiment value is to 1, the more positive the review is."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 15,
      "metadata": {
        "id": "aPNOYiiaha2y"
      },
      "outputs": [],
      "source": [
        "# Define a function to take a series of reviews\n",
        "# and predict whether each one is a positive or negative review\n",
        "\n",
        "# max_length = 100 # previously defined\n",
        "\n",
        "def predict_review(model, new_sentences, maxlen=max_length, show_padded_sequence=True ):\n",
        "  # Keep the original sentences so that we can keep using them later\n",
        "  # Create an array to hold the encoded sequences\n",
        "  new_sequences = []\n",
        "\n",
        "  # Convert the new reviews to sequences\n",
        "  for i, frvw in enumerate(new_sentences):\n",
        "    new_sequences.append(tokenizer.encode(frvw))\n",
        "\n",
        "  trunc_type='post' \n",
        "  padding_type='post'\n",
        "\n",
        "  # Pad all sequences for the new reviews\n",
        "  new_reviews_padded = pad_sequences(new_sequences, maxlen=max_length, \n",
        "                                 padding=padding_type, truncating=trunc_type)             \n",
        "\n",
        "  classes = model.predict(new_reviews_padded)\n",
        "\n",
        "  # The closer the class is to 1, the more positive the review is\n",
        "  for x in range(len(new_sentences)):\n",
        "    \n",
        "    # We can see the padded sequence if desired\n",
        "    # Print the sequence\n",
        "    if (show_padded_sequence):\n",
        "      print(new_reviews_padded[x])\n",
        "    # Print the review as text\n",
        "    print(new_sentences[x])\n",
        "    # Print its predicted class\n",
        "    print(classes[x])\n",
        "    print(\"\\n\")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 16,
      "metadata": {
        "id": "Qg-maex27KPW",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "9aa014a8-d4a4-4cc4-ba86-1baf3359e1a4"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[  4 281  16  25   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            "I love this phone\n",
            "[0.85955244]\n",
            "\n",
            "\n",
            "[812 227 864 100 775   9 525 843   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            "Everything was cold\n",
            "[0.08771736]\n",
            "\n",
            "\n",
            "[812 227 864 100 775   9 109   8 333 731  24  61   4 171  59  77   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            "Everything was hot exactly as I wanted\n",
            "[0.2779013]\n",
            "\n",
            "\n",
            "[812 227 864 100 775   9 157 359 853   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            "Everything was green\n",
            "[0.07703219]\n",
            "\n",
            "\n",
            "[  1 109 228 540 237 635 241 423 340  89 298   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            "the host seated us immediately\n",
            "[0.39272216]\n",
            "\n",
            "\n",
            "[154 242  47 635 341  12 569 547 147 319 775 125  85   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            "they gave us free chocolate cake\n",
            "[0.69143677]\n",
            "\n",
            "\n",
            "[158 190 853 782   8 607 775 210 232 146 775 470  67 305 101  15   1 328\n",
            " 296  26  19   1 661 641 195   0   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            "we couldn't hear each other talk because of the shouting in the kitchen\n",
            "[0.01040554]\n",
            "\n",
            "\n"
          ]
        }
      ],
      "source": [
        "# Use the model to predict some reviews   \n",
        "fake_reviews = [\"I love this phone\", \n",
        "                \"Everything was cold\",\n",
        "                \"Everything was hot exactly as I wanted\", \n",
        "                \"Everything was green\", \n",
        "                \"the host seated us immediately\",\n",
        "                \"they gave us free chocolate cake\", \n",
        "                \"we couldn't hear each other talk because of the shouting in the kitchen\"\n",
        "              ]\n",
        "\n",
        "predict_review(model, fake_reviews)\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ycJKbMq3K4iy"
      },
      "source": [
        "# Define a function to train and show the results of models with different layers\n",
        "\n",
        "In the rest of this colab, we will define models, and then see the results. \n",
        "\n",
        "Define a function that will take the model, compile it, train it, graph the accuracy and loss, and then predict some results."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 17,
      "metadata": {
        "id": "PevUcINXK3gn"
      },
      "outputs": [],
      "source": [
        "def fit_model_now (model, sentences) :\n",
        "  model.compile(loss='binary_crossentropy',optimizer='adam',metrics=['accuracy'])\n",
        "  model.summary()\n",
        "  history = model.fit(training_sequences, training_labels_final, epochs=num_epochs, \n",
        "                      validation_data=(testing_sequences, testing_labels_final))\n",
        "  return history\n",
        "\n",
        "def plot_results (history):\n",
        "  plot_graphs(history, \"accuracy\")\n",
        "  plot_graphs(history, \"loss\")\n",
        "\n",
        "def fit_model_and_show_results (model, sentences):\n",
        "  history = fit_model_now(model, sentences)\n",
        "  plot_results(history)\n",
        "  predict_review(model, sentences)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "U13JBiJUG1oq"
      },
      "source": [
        "# Add a bidirectional LSTM\n",
        "\n",
        "Create a new model that uses a bidirectional LSTM.\n",
        "\n",
        "Then use the function we have already defined to compile the model, train it, graph the accuracy and loss, then predict some results."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 18,
      "metadata": {
        "id": "scTUsFPAG4zP",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "e6cdf7db-ba6d-4358-e285-c74ee4c470a7"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential_1\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " embedding_1 (Embedding)     (None, 50, 16)            16000     \n",
            "                                                                 \n",
            " bidirectional (Bidirectiona  (None, 32)               4224      \n",
            " l)                                                              \n",
            "                                                                 \n",
            " dense_2 (Dense)             (None, 6)                 198       \n",
            "                                                                 \n",
            " dense_3 (Dense)             (None, 1)                 7         \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 20,429\n",
            "Trainable params: 20,429\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "Epoch 1/30\n",
            "50/50 [==============================] - 7s 42ms/step - loss: 0.6905 - accuracy: 0.5223 - val_loss: 0.7001 - val_accuracy: 0.4110\n",
            "Epoch 2/30\n",
            "50/50 [==============================] - 1s 24ms/step - loss: 0.6686 - accuracy: 0.5637 - val_loss: 0.6788 - val_accuracy: 0.5038\n",
            "Epoch 3/30\n",
            "50/50 [==============================] - 1s 24ms/step - loss: 0.5925 - accuracy: 0.7050 - val_loss: 0.6307 - val_accuracy: 0.6416\n",
            "Epoch 4/30\n",
            "50/50 [==============================] - 1s 24ms/step - loss: 0.5038 - accuracy: 0.8154 - val_loss: 0.6451 - val_accuracy: 0.6992\n",
            "Epoch 5/30\n",
            "50/50 [==============================] - 1s 24ms/step - loss: 0.4305 - accuracy: 0.8763 - val_loss: 0.6380 - val_accuracy: 0.7343\n",
            "Epoch 6/30\n",
            "50/50 [==============================] - 1s 24ms/step - loss: 0.3936 - accuracy: 0.9008 - val_loss: 0.7494 - val_accuracy: 0.7168\n",
            "Epoch 7/30\n",
            "50/50 [==============================] - 1s 24ms/step - loss: 0.3600 - accuracy: 0.9240 - val_loss: 0.6266 - val_accuracy: 0.7519\n",
            "Epoch 8/30\n",
            "50/50 [==============================] - 1s 24ms/step - loss: 0.2686 - accuracy: 0.9297 - val_loss: 0.5995 - val_accuracy: 0.7719\n",
            "Epoch 9/30\n",
            "50/50 [==============================] - 1s 24ms/step - loss: 0.1743 - accuracy: 0.9479 - val_loss: 0.8727 - val_accuracy: 0.7444\n",
            "Epoch 10/30\n",
            "50/50 [==============================] - 1s 24ms/step - loss: 0.1445 - accuracy: 0.9598 - val_loss: 0.7069 - val_accuracy: 0.7469\n",
            "Epoch 11/30\n",
            "50/50 [==============================] - 1s 24ms/step - loss: 0.1279 - accuracy: 0.9655 - val_loss: 1.1417 - val_accuracy: 0.7343\n",
            "Epoch 12/30\n",
            "50/50 [==============================] - 1s 24ms/step - loss: 0.1141 - accuracy: 0.9718 - val_loss: 1.1546 - val_accuracy: 0.7218\n",
            "Epoch 13/30\n",
            "50/50 [==============================] - 1s 24ms/step - loss: 0.1066 - accuracy: 0.9711 - val_loss: 0.9987 - val_accuracy: 0.7544\n",
            "Epoch 14/30\n",
            "50/50 [==============================] - 1s 25ms/step - loss: 0.0841 - accuracy: 0.9818 - val_loss: 1.0426 - val_accuracy: 0.7544\n",
            "Epoch 15/30\n",
            "50/50 [==============================] - 1s 23ms/step - loss: 0.0842 - accuracy: 0.9812 - val_loss: 1.1747 - val_accuracy: 0.7419\n",
            "Epoch 16/30\n",
            "50/50 [==============================] - 1s 25ms/step - loss: 0.0765 - accuracy: 0.9837 - val_loss: 1.0816 - val_accuracy: 0.7619\n",
            "Epoch 17/30\n",
            "50/50 [==============================] - 1s 24ms/step - loss: 0.0760 - accuracy: 0.9831 - val_loss: 1.0380 - val_accuracy: 0.7519\n",
            "Epoch 18/30\n",
            "50/50 [==============================] - 1s 24ms/step - loss: 0.0746 - accuracy: 0.9843 - val_loss: 1.2014 - val_accuracy: 0.7519\n",
            "Epoch 19/30\n",
            "50/50 [==============================] - 1s 24ms/step - loss: 0.0723 - accuracy: 0.9843 - val_loss: 1.2514 - val_accuracy: 0.7393\n",
            "Epoch 20/30\n",
            "50/50 [==============================] - 1s 24ms/step - loss: 0.0710 - accuracy: 0.9843 - val_loss: 1.3006 - val_accuracy: 0.7419\n",
            "Epoch 21/30\n",
            "50/50 [==============================] - 1s 30ms/step - loss: 0.0670 - accuracy: 0.9856 - val_loss: 1.2888 - val_accuracy: 0.7393\n",
            "Epoch 22/30\n",
            "50/50 [==============================] - 1s 24ms/step - loss: 0.0648 - accuracy: 0.9862 - val_loss: 1.2721 - val_accuracy: 0.7444\n",
            "Epoch 23/30\n",
            "50/50 [==============================] - 1s 24ms/step - loss: 0.0644 - accuracy: 0.9862 - val_loss: 1.2920 - val_accuracy: 0.7519\n",
            "Epoch 24/30\n",
            "50/50 [==============================] - 1s 24ms/step - loss: 0.0621 - accuracy: 0.9868 - val_loss: 1.3400 - val_accuracy: 0.7519\n",
            "Epoch 25/30\n",
            "50/50 [==============================] - 1s 24ms/step - loss: 0.0612 - accuracy: 0.9868 - val_loss: 1.5156 - val_accuracy: 0.7293\n",
            "Epoch 26/30\n",
            "50/50 [==============================] - 1s 24ms/step - loss: 0.0994 - accuracy: 0.9743 - val_loss: 0.8640 - val_accuracy: 0.7494\n",
            "Epoch 27/30\n",
            "50/50 [==============================] - 1s 24ms/step - loss: 0.0926 - accuracy: 0.9736 - val_loss: 1.5582 - val_accuracy: 0.7193\n",
            "Epoch 28/30\n",
            "50/50 [==============================] - 1s 24ms/step - loss: 0.0805 - accuracy: 0.9812 - val_loss: 1.0933 - val_accuracy: 0.7444\n",
            "Epoch 29/30\n",
            "50/50 [==============================] - 1s 25ms/step - loss: 0.0644 - accuracy: 0.9856 - val_loss: 1.3499 - val_accuracy: 0.7243\n",
            "Epoch 30/30\n",
            "50/50 [==============================] - 1s 24ms/step - loss: 0.0592 - accuracy: 0.9874 - val_loss: 1.3442 - val_accuracy: 0.7419\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[  4 281  16  25   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            "I love this phone\n",
            "[0.99946624]\n",
            "\n",
            "\n",
            "[812 227 864 100 775   9 525 843   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            "Everything was cold\n",
            "[0.03113783]\n",
            "\n",
            "\n",
            "[812 227 864 100 775   9 109   8 333 731  24  61   4 171  59  77   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            "Everything was hot exactly as I wanted\n",
            "[0.95132273]\n",
            "\n",
            "\n",
            "[812 227 864 100 775   9 157 359 853   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            "Everything was green\n",
            "[0.04094881]\n",
            "\n",
            "\n",
            "[  1 109 228 540 237 635 241 423 340  89 298   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            "the host seated us immediately\n",
            "[0.0292745]\n",
            "\n",
            "\n",
            "[154 242  47 635 341  12 569 547 147 319 775 125  85   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            "they gave us free chocolate cake\n",
            "[0.99960846]\n",
            "\n",
            "\n",
            "[158 190 853 782   8 607 775 210 232 146 775 470  67 305 101  15   1 328\n",
            " 296  26  19   1 661 641 195   0   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            "we couldn't hear each other talk because of the shouting in the kitchen\n",
            "[0.02581133]\n",
            "\n",
            "\n"
          ]
        }
      ],
      "source": [
        "# Define the model\n",
        "model_bidi_lstm = tf.keras.Sequential([\n",
        "    tf.keras.layers.Embedding(vocab_size, embedding_dim, input_length=max_length),\n",
        "    tf.keras.layers.Bidirectional(tf.keras.layers.LSTM(embedding_dim)), \n",
        "    tf.keras.layers.Dense(6, activation='relu'), \n",
        "    tf.keras.layers.Dense(1, activation='sigmoid')\n",
        "])\n",
        "\n",
        "# Compile and train the model and then show the predictions for our extra sentences\n",
        "fit_model_and_show_results(model_bidi_lstm, fake_reviews)\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "QsxKPbCnPJTj"
      },
      "source": [
        "# Use multiple bidirectional layers\n",
        "\n",
        "Now let's see if we get any improvements from adding another Bidirectional LSTM layer to the model.\n",
        "\n",
        "Notice that the first Bidirectionl LSTM layer returns a sequence."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 19,
      "metadata": {
        "id": "3N6Zul47PMED",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "655c686e-39b4-4df5-f62e-33e23252121a"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential_2\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " embedding_2 (Embedding)     (None, 50, 16)            16000     \n",
            "                                                                 \n",
            " bidirectional_1 (Bidirectio  (None, 50, 32)           4224      \n",
            " nal)                                                            \n",
            "                                                                 \n",
            " bidirectional_2 (Bidirectio  (None, 32)               6272      \n",
            " nal)                                                            \n",
            "                                                                 \n",
            " dense_4 (Dense)             (None, 6)                 198       \n",
            "                                                                 \n",
            " dense_5 (Dense)             (None, 1)                 7         \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 26,701\n",
            "Trainable params: 26,701\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "Epoch 1/30\n",
            "50/50 [==============================] - 11s 89ms/step - loss: 0.6932 - accuracy: 0.5009 - val_loss: 0.6939 - val_accuracy: 0.4110\n",
            "Epoch 2/30\n",
            "50/50 [==============================] - 3s 50ms/step - loss: 0.6929 - accuracy: 0.5223 - val_loss: 0.6950 - val_accuracy: 0.4110\n",
            "Epoch 3/30\n",
            "50/50 [==============================] - 2s 49ms/step - loss: 0.6927 - accuracy: 0.5223 - val_loss: 0.6958 - val_accuracy: 0.4110\n",
            "Epoch 4/30\n",
            "50/50 [==============================] - 3s 50ms/step - loss: 0.6926 - accuracy: 0.5223 - val_loss: 0.6967 - val_accuracy: 0.4110\n",
            "Epoch 5/30\n",
            "50/50 [==============================] - 2s 50ms/step - loss: 0.6925 - accuracy: 0.5223 - val_loss: 0.6976 - val_accuracy: 0.4110\n",
            "Epoch 6/30\n",
            "50/50 [==============================] - 2s 50ms/step - loss: 0.6924 - accuracy: 0.5223 - val_loss: 0.6980 - val_accuracy: 0.4110\n",
            "Epoch 7/30\n",
            "50/50 [==============================] - 3s 51ms/step - loss: 0.6923 - accuracy: 0.5223 - val_loss: 0.6983 - val_accuracy: 0.4110\n",
            "Epoch 8/30\n",
            "50/50 [==============================] - 2s 49ms/step - loss: 0.6792 - accuracy: 0.5813 - val_loss: 0.6293 - val_accuracy: 0.6617\n",
            "Epoch 9/30\n",
            "50/50 [==============================] - 3s 51ms/step - loss: 0.5196 - accuracy: 0.7891 - val_loss: 0.5537 - val_accuracy: 0.7243\n",
            "Epoch 10/30\n",
            "50/50 [==============================] - 2s 50ms/step - loss: 0.3700 - accuracy: 0.8638 - val_loss: 0.6385 - val_accuracy: 0.7193\n",
            "Epoch 11/30\n",
            "50/50 [==============================] - 2s 49ms/step - loss: 0.3071 - accuracy: 0.8920 - val_loss: 0.6163 - val_accuracy: 0.7368\n",
            "Epoch 12/30\n",
            "50/50 [==============================] - 3s 51ms/step - loss: 0.2528 - accuracy: 0.9159 - val_loss: 0.6695 - val_accuracy: 0.7419\n",
            "Epoch 13/30\n",
            "50/50 [==============================] - 3s 50ms/step - loss: 0.2279 - accuracy: 0.9297 - val_loss: 0.7072 - val_accuracy: 0.7243\n",
            "Epoch 14/30\n",
            "50/50 [==============================] - 3s 51ms/step - loss: 0.1915 - accuracy: 0.9473 - val_loss: 0.6852 - val_accuracy: 0.7669\n",
            "Epoch 15/30\n",
            "50/50 [==============================] - 3s 51ms/step - loss: 0.1711 - accuracy: 0.9542 - val_loss: 0.7705 - val_accuracy: 0.7419\n",
            "Epoch 16/30\n",
            "50/50 [==============================] - 2s 50ms/step - loss: 0.1785 - accuracy: 0.9479 - val_loss: 0.7934 - val_accuracy: 0.7193\n",
            "Epoch 17/30\n",
            "50/50 [==============================] - 2s 50ms/step - loss: 0.1434 - accuracy: 0.9630 - val_loss: 0.8137 - val_accuracy: 0.7419\n",
            "Epoch 18/30\n",
            "50/50 [==============================] - 2s 50ms/step - loss: 0.1212 - accuracy: 0.9724 - val_loss: 0.8746 - val_accuracy: 0.7318\n",
            "Epoch 19/30\n",
            "50/50 [==============================] - 3s 51ms/step - loss: 0.1192 - accuracy: 0.9692 - val_loss: 0.8881 - val_accuracy: 0.7293\n",
            "Epoch 20/30\n",
            "50/50 [==============================] - 3s 50ms/step - loss: 0.1091 - accuracy: 0.9736 - val_loss: 0.9203 - val_accuracy: 0.7393\n",
            "Epoch 21/30\n",
            "50/50 [==============================] - 2s 49ms/step - loss: 0.1141 - accuracy: 0.9730 - val_loss: 0.9262 - val_accuracy: 0.7343\n",
            "Epoch 22/30\n",
            "50/50 [==============================] - 3s 50ms/step - loss: 0.1043 - accuracy: 0.9768 - val_loss: 0.9247 - val_accuracy: 0.7469\n",
            "Epoch 23/30\n",
            "50/50 [==============================] - 2s 50ms/step - loss: 0.1232 - accuracy: 0.9692 - val_loss: 0.8741 - val_accuracy: 0.7494\n",
            "Epoch 24/30\n",
            "50/50 [==============================] - 3s 51ms/step - loss: 0.1035 - accuracy: 0.9780 - val_loss: 0.9464 - val_accuracy: 0.7343\n",
            "Epoch 25/30\n",
            "50/50 [==============================] - 2s 50ms/step - loss: 0.0900 - accuracy: 0.9799 - val_loss: 0.9482 - val_accuracy: 0.7393\n",
            "Epoch 26/30\n",
            "50/50 [==============================] - 2s 50ms/step - loss: 0.0811 - accuracy: 0.9837 - val_loss: 0.9925 - val_accuracy: 0.7444\n",
            "Epoch 27/30\n",
            "50/50 [==============================] - 3s 50ms/step - loss: 0.0752 - accuracy: 0.9856 - val_loss: 0.9859 - val_accuracy: 0.7544\n",
            "Epoch 28/30\n",
            "50/50 [==============================] - 3s 50ms/step - loss: 0.0908 - accuracy: 0.9793 - val_loss: 0.9950 - val_accuracy: 0.7444\n",
            "Epoch 29/30\n",
            "50/50 [==============================] - 3s 50ms/step - loss: 0.0765 - accuracy: 0.9849 - val_loss: 1.0543 - val_accuracy: 0.7343\n",
            "Epoch 30/30\n",
            "50/50 [==============================] - 3s 51ms/step - loss: 0.0768 - accuracy: 0.9831 - val_loss: 1.0214 - val_accuracy: 0.7393\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[  4 281  16  25   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            "I love this phone\n",
            "[0.98478323]\n",
            "\n",
            "\n",
            "[812 227 864 100 775   9 525 843   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            "Everything was cold\n",
            "[0.02741526]\n",
            "\n",
            "\n",
            "[812 227 864 100 775   9 109   8 333 731  24  61   4 171  59  77   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            "Everything was hot exactly as I wanted\n",
            "[0.9722758]\n",
            "\n",
            "\n",
            "[812 227 864 100 775   9 157 359 853   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            "Everything was green\n",
            "[0.01218022]\n",
            "\n",
            "\n",
            "[  1 109 228 540 237 635 241 423 340  89 298   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            "the host seated us immediately\n",
            "[0.98215955]\n",
            "\n",
            "\n",
            "[154 242  47 635 341  12 569 547 147 319 775 125  85   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            "they gave us free chocolate cake\n",
            "[0.98337674]\n",
            "\n",
            "\n",
            "[158 190 853 782   8 607 775 210 232 146 775 470  67 305 101  15   1 328\n",
            " 296  26  19   1 661 641 195   0   0   0   0   0   0   0   0   0   0   0\n",
            "   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            "we couldn't hear each other talk because of the shouting in the kitchen\n",
            "[0.01280827]\n",
            "\n",
            "\n"
          ]
        }
      ],
      "source": [
        "model_multiple_bidi_lstm = tf.keras.Sequential([\n",
        "    tf.keras.layers.Embedding(vocab_size, embedding_dim, input_length=max_length),\n",
        "    tf.keras.layers.Bidirectional(tf.keras.layers.LSTM(embedding_dim, \n",
        "                                                       return_sequences=True)), \n",
        "    tf.keras.layers.Bidirectional(tf.keras.layers.LSTM(embedding_dim)),\n",
        "    tf.keras.layers.Dense(6, activation='relu'),\n",
        "    tf.keras.layers.Dense(1, activation='sigmoid')\n",
        "])\n",
        "\n",
        "fit_model_and_show_results(model_multiple_bidi_lstm, fake_reviews)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ABVYYPwba8Hx"
      },
      "source": [
        "# Compare predictions for all the models\n",
        "\n",
        "It can be hard to see which model gives a better prediction for different reviews when you examine each model separately. So for comparison purposes, here we define some more reviews and print out the predictions that each of the three models gives for each review:\n",
        "\n",
        "*   Embeddings and a Global Average Pooling layer\n",
        "*   Embeddings and a Bidirectional LSTM layer\n",
        "*   Embeddings and two Bidirectional LSTM layers\n",
        "\n",
        "The results are not always what you might expect. The input dataset is fairly small, it has less than 2000 reviews. Some of the reviews are fairly short, and some of the short ones are fairly repetitive which reduces their impact on improving the  model, such as these two reviews:\n",
        "\n",
        "*   Bad Quality.\n",
        "*   Low Quality.\n",
        "\n",
        "Feel free to add more reviews of your own, or change the reviews. The results will depend on the combination of words in the reviews, and how well they match to reviews in the training set. \n",
        "\n",
        "How do the different models handle things like \"wasn't good\" which contains a positive word (good) but is a poor review?\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 20,
      "metadata": {
        "id": "6XebrXt0jtOy"
      },
      "outputs": [],
      "source": [
        "my_reviews =[\"lovely\", \"dreadful\", \"stay away\",\n",
        "             \"everything was hot exactly as I wanted\",\n",
        "             \"everything was not exactly as I wanted\",\n",
        "             \"they gave us free chocolate cake\",\n",
        "             \"I've never eaten anything so spicy in my life, my throat burned for hours\",\n",
        "             \"for a phone that is as expensive as this one I expect it to be much easier to use than this thing is\",\n",
        "             \"we left there very full for a low price so I'd say you just can't go wrong at this place\",\n",
        "             \"that place does not have quality meals and it isn't a good place to go for dinner\",\n",
        "             ]"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 21,
      "metadata": {
        "id": "tRWGjkJLkY2y",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "c44c970c-8784-4cb9-e6a9-5bff36957bc6"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "===================================\n",
            " Embeddings only:\n",
            " ===================================\n",
            "lovely\n",
            "[0.90628594]\n",
            "\n",
            "\n",
            "dreadful\n",
            "[0.26473883]\n",
            "\n",
            "\n",
            "stay away\n",
            "[0.58034515]\n",
            "\n",
            "\n",
            "everything was hot exactly as I wanted\n",
            "[0.71621734]\n",
            "\n",
            "\n",
            "everything was not exactly as I wanted\n",
            "[0.47994143]\n",
            "\n",
            "\n",
            "they gave us free chocolate cake\n",
            "[0.69143677]\n",
            "\n",
            "\n",
            "I've never eaten anything so spicy in my life, my throat burned for hours\n",
            "[0.01850466]\n",
            "\n",
            "\n",
            "for a phone that is as expensive as this one I expect it to be much easier to use than this thing is\n",
            "[0.52018315]\n",
            "\n",
            "\n",
            "we left there very full for a low price so I'd say you just can't go wrong at this place\n",
            "[0.7585648]\n",
            "\n",
            "\n",
            "that place does not have quality meals and it isn't a good place to go for dinner\n",
            "[0.82910997]\n",
            "\n",
            "\n"
          ]
        }
      ],
      "source": [
        "print(\"===================================\\n\",\"Embeddings only:\\n\", \"===================================\",)\n",
        "predict_review(model, my_reviews, show_padded_sequence=False)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 22,
      "metadata": {
        "id": "G2FJR3IVBt30",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "9d2f7e56-e943-4e4a-d1bb-b492eddb67de"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "===================================\n",
            " With a single bidirectional LSTM:\n",
            " ===================================\n",
            "lovely\n",
            "[0.9946215]\n",
            "\n",
            "\n",
            "dreadful\n",
            "[0.03350855]\n",
            "\n",
            "\n",
            "stay away\n",
            "[0.04112142]\n",
            "\n",
            "\n",
            "everything was hot exactly as I wanted\n",
            "[0.99827707]\n",
            "\n",
            "\n",
            "everything was not exactly as I wanted\n",
            "[0.1785508]\n",
            "\n",
            "\n",
            "they gave us free chocolate cake\n",
            "[0.99960846]\n",
            "\n",
            "\n",
            "I've never eaten anything so spicy in my life, my throat burned for hours\n",
            "[0.6534659]\n",
            "\n",
            "\n",
            "for a phone that is as expensive as this one I expect it to be much easier to use than this thing is\n",
            "[0.03051597]\n",
            "\n",
            "\n",
            "we left there very full for a low price so I'd say you just can't go wrong at this place\n",
            "[0.99772125]\n",
            "\n",
            "\n",
            "that place does not have quality meals and it isn't a good place to go for dinner\n",
            "[0.03098973]\n",
            "\n",
            "\n"
          ]
        }
      ],
      "source": [
        "print(\"===================================\\n\", \"With a single bidirectional LSTM:\\n\", \"===================================\")\n",
        "predict_review(model_bidi_lstm, my_reviews, show_padded_sequence=False)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 23,
      "metadata": {
        "id": "81v1r3Y2BwvC",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "d1266cc4-643e-4970-9b19-e984db5a7efa"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "===================================\n",
            " With two bidirectional LSTMs:\n",
            " ===================================\n",
            "lovely\n",
            "[0.98244196]\n",
            "\n",
            "\n",
            "dreadful\n",
            "[0.00842986]\n",
            "\n",
            "\n",
            "stay away\n",
            "[0.9820576]\n",
            "\n",
            "\n",
            "everything was hot exactly as I wanted\n",
            "[0.9829973]\n",
            "\n",
            "\n",
            "everything was not exactly as I wanted\n",
            "[0.9774137]\n",
            "\n",
            "\n",
            "they gave us free chocolate cake\n",
            "[0.98337674]\n",
            "\n",
            "\n",
            "I've never eaten anything so spicy in my life, my throat burned for hours\n",
            "[0.02585457]\n",
            "\n",
            "\n",
            "for a phone that is as expensive as this one I expect it to be much easier to use than this thing is\n",
            "[0.98213667]\n",
            "\n",
            "\n",
            "we left there very full for a low price so I'd say you just can't go wrong at this place\n",
            "[0.7370272]\n",
            "\n",
            "\n",
            "that place does not have quality meals and it isn't a good place to go for dinner\n",
            "[0.9771752]\n",
            "\n",
            "\n"
          ]
        }
      ],
      "source": [
        "print(\"===================================\\n\",\"With two bidirectional LSTMs:\\n\", \"===================================\")\n",
        "predict_review(model_multiple_bidi_lstm, my_reviews, show_padded_sequence=False)"
      ]
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "name": "l10c01_nlp_lstms_with_reviews_subwords_dataset.ipynb",
      "toc_visible": true,
      "provenance": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}