Notebooks/07_Word_Embeddings.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "cb2fecfa",
   "metadata": {
    "papermill": {
     "duration": 0.015556,
     "end_time": "2023-02-20T21:20:58.974769",
     "exception": false,
     "start_time": "2023-02-20T21:20:58.959213",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "<center>\n",
    "<img src='..\\assets\\1\\lang-pic.jpg' width=600>\n",
    "</center>\n",
    "    \n",
    "# 1. Introduction\n",
    "\n",
    "## <p style=\"font-family:JetBrains Mono; font-weight:normal; letter-spacing: 1px; color:#207d06; font-size:100%; text-align:left;padding: 0px; border-bottom: 3px solid #207d06;\">1.1 NLP series</p>\n",
    "\n",
    "This is the **seventh in a series of notebooks** covering the **fundamentals of Natural Language Processing (NLP)**. I find that the best way to learn is by teaching others, hence why I am sharing my journey learning this field from scratch. I hope these notebooks can be helpful to you too. \n",
    "\n",
    "NLP series:\n",
    "\n",
    "1. [Tokenization](./01_Tokenization.ipynb)\n",
    "2. [Preprocessing](./02_Pre_Processing.ipynb)\n",
    "3. [Bag of Words and Similarity](./03_BOW_Similarity.ipynb)\n",
    "4. [TF-IDF and Document Search](./04_TFIDF_DocSearch.ipynb)\n",
    "5. [Naive Bayes Text Classification](./05_NaiveBayes_TextClf.ipynb)\n",
    "6. [LDA Topic Modelling](./06_LDA_TopicModelling.ipynb)\n",
    "7. Word Embeddings\n",
    "<a target=\"_blank\" href=\"https://colab.research.google.com/github/JUSTSUJAY/NLP_One_Shot/blob/main/Notebooks/07_Word_Embeddings.ipynb\">\n",
    "  <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/>\n",
    "</a>\n",
    "[![Kaggle](https://kaggle.com/static/images/open-in-kaggle.svg)](https://kaggle.com/kernels/welcome?src=https://github.com/JUSTSUJAY/NLP_One_Shot/blob/main/Notebooks/07_Word_Embeddings.ipynb)\n",
    "\n",
    "## <p style=\"font-family:JetBrains Mono; font-weight:normal; letter-spacing: 1px; color:#207d06; font-size:100%; text-align:left;padding: 0px; border-bottom: 3px solid #207d06;\">1.2 Outline</p>\n",
    "\n",
    "So far we've seen how to **vectorize documents** of text either through Bag-of-Words or TF-IDF. While these approaches work well for simple NLP tasks like classification, they have the drawback that they don't capture any relationships between words. \n",
    "\n",
    "In this notebook, we will see how to **vectorize individual words** via **static embeddings** in order to **capture word meaning**. For example, this will let us model that \"brother\" and \"sister\" are more similar in meaning than \"tree\" and \"car\". Note that in a later notebook we will also cover **contextual embeddings** where the embedding can change depending on the context."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d2c5107d",
   "metadata": {
    "papermill": {
     "duration": 0.014166,
     "end_time": "2023-02-20T21:20:59.003004",
     "exception": false,
     "start_time": "2023-02-20T21:20:58.988838",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# 2. Word Embeddings\n",
    "\n",
    "## <p style=\"font-family:JetBrains Mono; font-weight:normal; letter-spacing: 1px; color:#207d06; font-size:100%; text-align:left;padding: 0px; border-bottom: 3px solid #207d06;\">2.1 One-hot encoding</p>\n",
    "\n",
    "The simplest way to vectorize a set of words, is to use **one-hot encoding**. This maps each word into a vector with length equal to the size of the vocabulary. The vector is completely filled with 0's except for a single entry, which has a 1 correspoding to the index of the word in the vocabulary. \n",
    "\n",
    "<br>\n",
    "<center>\n",
    "<img src=\"../assets/7/onehot.png\" width=600>\n",
    "</center>\n",
    "<br>\n",
    "\n",
    "This is a pretty **terrible way** to vectorize words, not only because it is very **memory inefficient** but also because there is no relationship between words. \n",
    "\n",
    "In particular, if the voculary contains 10,000 words, then each vector has length equal to 10,000. Furthermore, any two distinct vectors will always have a **dot product equal to 0**, corresponding to no similarity.\n",
    "\n",
    "<br>\n",
    "\n",
    "## <p style=\"font-family:JetBrains Mono; font-weight:normal; letter-spacing: 1px; color:#207d06; font-size:100%; text-align:left;padding: 0px; border-bottom: 3px solid #207d06;\">2.2 Capturing meaning</p>\n",
    "\n",
    "A better way then would be to represent words as **shorter and denser vectors** that capture some meaning between words. And this is what an embedding aim to do. \n",
    "\n",
    "An **embedding** is simply a **representation of an object** (e.g. a word, movie, graph, etc) as a **vector of real numbers**. It *embeds* an object into a **high-dimensional vector space**. \n",
    "\n",
    "<br>\n",
    "<center>\n",
    "<img src=\"../assets/7/unitcube.png\" width=300>\n",
    "</center>\n",
    "<br>\n",
    "\n",
    "For example, let's say you have a collection of videos games. We can represent each game by measuring a number of its **attributes** like `[<fantasy>,<strategy>,<multiplayer>,<action>,<adventure>]`. So a game like 'Minecraft' could be represented by `[0.1,0.6,0.4,0.5,0.9]` (although the numbers don't have to be between 0 and 1) and if we wanted to find similar game, we could return the game with the **highest similar score**. \n",
    "\n",
    "Notice how there are many **different ways to embed the same object**. Moreover, the features we hand-selected may not be the best ones to represent these objects. But the main idea is to find a representation where **similar objects** (i.e. are semantically similar) are **close together**. We can do this with words too, but how do we find their best representations?\n",
    "\n",
    "<br>\n",
    "<center>\n",
    "<img src=\"../assets/7/embed.jpg\" width=700>\n",
    "</center>\n",
    "<br>\n",
    "\n",
    "## <p style=\"font-family:JetBrains Mono; font-weight:normal; letter-spacing: 1px; color:#207d06; font-size:100%; text-align:left;padding: 0px; border-bottom: 3px solid #207d06;\">2.3 Distributional Hypothesis</p>\n",
    "\n",
    "To train a word embedding, we first need to ask ourselves what makes two words **semantically similar**? One popular answer to this is the **distributional hypothesis**, which says that \"words which appear in **similar contexts** (i.e. share similar surrounding words) have **similar meanings**\". \n",
    "\n",
    "For example, consider the sentence \"My family enjoys eating bacalhau at Christmas\". You probably have no idea what 'bacalhau' is but just from the **context** we can **infer** that it is some kind of food. (In fact, the word refers to Portuguese salted cod.)\n",
    "\n",
    "That are many algorithms that use this idea to **train word embeddings**. We are going to focus on the main one, namely **Word2vec**, which was developed in 2013 at Google and was a breakthrough in NLP at the time. ([link to paper](https://arxiv.org/abs/1301.3781))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ec606189",
   "metadata": {
    "papermill": {
     "duration": 0.013414,
     "end_time": "2023-02-20T21:20:59.036544",
     "exception": false,
     "start_time": "2023-02-20T21:20:59.023130",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# 3. Word2vec\n",
    "\n",
    "## <p style=\"font-family:JetBrains Mono; font-weight:normal; letter-spacing: 1px; color:#207d06; font-size:100%; text-align:left;padding: 0px; border-bottom: 3px solid #207d06;\">3.1 Word addition</p>\n",
    "\n",
    "Word2vec is a way of **learning** word embeddings by **training shallow neural networks** on either the continuous bag-of-words or the skip-gram task. We will discuss both of these shortly. The **dimension** of the resulting dense vectors is usually between 50-1000 and a common value is 300.  \n",
    "\n",
    "<br>\n",
    "<center>\n",
    "<img src=\"../assets/7/word2vec.png\" width=500>\n",
    "</center>\n",
    "<br>\n",
    "\n",
    "The amazing thing to come out of Word2vec is not only that similar words are close together, but that we can perform **addition and subtraction** on these word vectors. For example, **“king” - “man” + “woman” = \"queen\"**. That is, if we take the vectors for king, man, woman and add/subtract them in this way, we will end up with a vector close to the one corresponding to queen. This means that these vectors can **capture very precisely** abstract concepts (like gender and royalty) without any input from us.\n",
    "\n",
    "Note that the values in each vector don't necessarily correspond to anything like fantasy, strategy, etc like in the video game example from before. The algorithm learns the best representation even if the **numbers don't correspond to anything tangible**. The most important thing though is that similar words are close together. \n",
    "\n",
    "<br>\n",
    "\n",
    "## <p style=\"font-family:JetBrains Mono; font-weight:normal; letter-spacing: 1px; color:#207d06; font-size:100%; text-align:left;padding: 0px; border-bottom: 3px solid #207d06;\">3.2 Training Word2vec</p>\n",
    "\n",
    "Word2vec is a collection of **two algorithms** that each learn word vectors **indirectly** through a word prediction task. One option is **Continuous Bag-of-Words** (CBOW) and the other is **skip-gram**. \n",
    "\n",
    "<br>\n",
    "<center>\n",
    "<img src=\"../assets/7/cbow.png\" width=600>\n",
    "</center>\n",
    "<br>\n",
    "\n",
    "In CBOW, the task is to **predict the missing word** from its surrounding words whereas in skip-gram the task is to **predict the surrounging words** given a single word. They are both quite similar so we will just focus on skip-gram (with an optimized adjustment called **negative sampling**) in this notebook.\n",
    "\n",
    "<br>\n",
    "\n",
    "## <p style=\"font-family:JetBrains Mono; font-weight:normal; letter-spacing: 1px; color:#207d06; font-size:100%; text-align:left;padding: 0px; border-bottom: 3px solid #207d06;\">3.3 Skip-Gram with Negative Sampling</p>\n",
    "\n",
    "The goal of **Skip-Gram with Negative Sampling** (SGNS) is to maximise the similarity between words that appear in **similar contexts** and conversely minize the similar between words that do not appear in similar contexts. Note that depending on the corpus you train on you will get different results. \n",
    "\n",
    "Consider the following sentence: \"We ate pizza at the Italian restaurant\". In SGNS we designate a word, for example \"pizza\", to be the **input word**. The words around the input word become the **context words**. From this, we generate **input/context pairs**: (pizza, We), (pizza,ate), (pizza,at), (pizza,the), (pizza,Italian) and (pizza,restaurant). These pairs are considered **positive examples**.\n",
    "\n",
    "Note that in practice there is a **context window** that limits context words to only e.g. 2 either side the input word. This is treated as a hyperparamenter that can be tuned. Furthermore, generating positive samples can be done in a **sliding window fashion**, where we move by one word to the right each time. \n",
    "\n",
    "<br>\n",
    "<center>\n",
    "<img src=\"../assets/7/skipgram.png\" width=600>\n",
    "</center>\n",
    "<br>\n",
    "\n",
    "We also generate **negative samples** by selecting words (called **negative context words**) at random from the corpus, e.g. (pizza,hero), (pizza,skip), (pizza,restoration), (pizza,crew), etc. It is ok if context words also appear as negative context words - the chance of this happening is small anyway.\n",
    "\n",
    "Doing this repeatedly for the whole corpus generates a **training set** of positive and negative examples. What is amazing is that **no labelling was needed** at any point. (This is an example of **self-supervised learning**.)\n",
    "\n",
    "This training set gives rise to a **binary prediction task**, which is to predict if two words are either likely or unlikely to appear close to each other in text. We will see that learning to do this task well will lead us to our desired word embeddings. \n",
    "\n",
    "We start by creating **two matrices initialised randomly** both with **dimensions |V| x d**, where |V| is the size of the vocabulary and d is the size of the word vectors (e.g. d=300). The first matrix will hold the **input word embeddings** and the second will hold the **context word embeddings**. The word \"pizza\" (and any other word) will have a vector in each matrix but the idea is that when \"pizza\" is an input word, the input matrix will get updated whereas when it is a context word, the context matrix will get updated. \n",
    "\n",
    "At each training iteration we pick **one positive sample** $(w, c_{pos})$ and **k negative samples** $(w,c_{neg(i)})$ (k is a hyperparameter, k=3 is common). For example, let (pizza,restaurant) be the positive sample. The important thing is that we take vector for \"pizza\" from the input matrix and the vector for \"restaurant\" from the context matrix. Similarly we do the same for the negative samples. \n",
    "\n",
    "The **probability** that words in a positive (resp. negative) example appear close to each other (resp. don't appear close to each other) is calculated as follows\n",
    "\n",
    "<br>\n",
    "\n",
    "$$\n",
    "\\large\n",
    "\\mathbb{P}(+|w,c_{pos}) = \\sigma (w \\cdot c_{pos}), \\qquad \\quad \\mathbb{P}(-|w,c_{neg}) = 1 - \\mathbb{P}(+|w,c_{neg}) = \\sigma (- w \\cdot c_{neg})\n",
    "$$\n",
    "\n",
    "<br>\n",
    "\n",
    "The **dot product** is used to measure similarity (like in cosine similarity) and the **sigmoid function** is there to convert the numbers into probabilities. We want to **maximize** both quantities so we use the loss function (working in log space makes derivates easier to calculate)\n",
    "\n",
    "<br>\n",
    "\n",
    "$$\n",
    "\\large\n",
    "\\begin{align*}\n",
    "L &= - [\\log \\mathbb{P}(+|w,c_{pos}) + \\sum_{i}^{k} \\log \\mathbb{P}(-|w,c_{neg(i)})] \\\\\n",
    "&= - [\\log \\sigma (w \\cdot c_{pos}) + \\sum_{i}^{k} \\log \\sigma (- w \\cdot c_{neg(i)})] \\\\\n",
    "\\end{align*}\n",
    "$$\n",
    "\n",
    "<br>\n",
    "\n",
    "**Minimizing** this loss function will maximise the probabilities above, which have been constructed to force similar words to be close to each other and disimilar words to be far away from each other. We can use any **optimization algorithm** we want to do this, for example stochastic gradient descent or the Adam optimizer. During training, the two matrices (corresponding input and context) get updated at each iteration. Once training is completed, we will have arrived at the word embeddings we wanted. Usually just the input matrix embedding is used but you could also average the two matrices together.\n",
    "\n",
    "<br>\n",
    "\n",
    "## <p style=\"font-family:JetBrains Mono; font-weight:normal; letter-spacing: 1px; color:#207d06; font-size:100%; text-align:left;padding: 0px; border-bottom: 3px solid #207d06;\">3.4 Training tricks</p>\n",
    "\n",
    "There are a few tricks used in practice to **maximise the performance** of the skip-gram model and thus **improve the quality** of the embeddings. First, the following formula is commonly used when sampling **negative context words** in negative examples. This is basically the **relative frequency** of each word but with a modification that puts **more weight on rare words** and less weight on common words. The authors suggest $\\alpha=0.75$ is a good choice for this hyperparameter. \n",
    "\n",
    "<br>\n",
    "\n",
    "$$\n",
    "\\large\n",
    "\\mathbb{P}_{\\alpha}(w) = \\frac{\\text{count}(w)^{\\alpha}}{\\sum_i \\text{count}(w_i)^{\\alpha}}\n",
    "$$\n",
    "\n",
    "<br>\n",
    "\n",
    "The **window size** when sampling **positive examples** is also a hyperparameter as we said earlier. Researchers have found that **narrower windows** are better at capturing **functional similarity** better whereas wider windows are better at capturing **relatedness**. \n",
    "\n",
    "The **number of negative samples** selected at each iteration is controlled by the hyperparameter **k**. For large datasets, k is recommended to be between **2-5** whereas for smaller datasets a better choice is between **5-20**. \n",
    "\n",
    "Finally, in some implementations of Word2vec each word vector (input/context) gets passed through a **layer** of a **connected neural network** before the dot product is taken. This can improve the performance of the model but conceptually the process of learning the word embeddings is the same. \n",
    "\n",
    "<br>\n",
    "\n",
    "## <p style=\"font-family:JetBrains Mono; font-weight:normal; letter-spacing: 1px; color:#207d06; font-size:100%; text-align:left;padding: 0px; border-bottom: 3px solid #207d06;\">3.5 Using word vectors</p>\n",
    "\n",
    "Training accurate word embeddings from scratch takes a **lot of computational resources**. If we want to use word vectors in our models then we have a few options depending on the requirements of our task.\n",
    "\n",
    "1. Use **pre-trained word vectors** as model inputs and keep them **constant during training**. \n",
    "2. Use **pre-trained word vectors** as model inputs and allow the model to **tune** them during training. \n",
    "3. **Train word embeddings from scratch** at the same time as training the model. \n",
    "\n",
    "Option 1 will be the **fastest** and will usually lead to very good results. Option 2 can produce **marginally better results** if you are willing to take longer to train your model. Option 3 can be useful if you have **lots of training data** and want to train a model for a very **specific task** that pretrained embeddings don't perform well on. We will explore these different options with code in the next section. "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3563cfa4",
   "metadata": {
    "papermill": {
     "duration": 0.016019,
     "end_time": "2023-02-20T21:20:59.066491",
     "exception": false,
     "start_time": "2023-02-20T21:20:59.050472",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# 4. Application\n",
    "\n",
    "## <p style=\"font-family:JetBrains Mono; font-weight:normal; letter-spacing: 1px; color:#207d06; font-size:100%; text-align:left;padding: 0px; border-bottom: 3px solid #207d06;\">4.1 Word vectors</p>\n",
    "\n",
    "We're going to be using **pre-trained word vectors** via the **Gensim** library, which we came across last time. These particular vectors are known as the **Google News vectors** as they were trained on a 3 billion word Google News corpus in 2015. In total, there are **3 million, 300-dimension vectors**. \n",
    "<br>\n",
    "\n",
    "* GoogleNews-vectors-negative300 - [Link](https://www.kaggle.com/datasets/leadbest/googlenewsvectorsnegative300)\n",
    "* Spotify App Reviews - [Link](https://www.kaggle.com/datasets/mfaaris/spotify-app-reviews-2022)\n",
    "\n",
    "**Import libraries**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "a6b8df44",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:20:59.098509Z",
     "iopub.status.busy": "2023-02-20T21:20:59.097364Z",
     "iopub.status.idle": "2023-02-20T21:21:12.930384Z",
     "shell.execute_reply": "2023-02-20T21:21:12.929120Z"
    },
    "papermill": {
     "duration": 13.852444,
     "end_time": "2023-02-20T21:21:12.933796",
     "exception": false,
     "start_time": "2023-02-20T21:20:59.081352",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "pd.set_option('max_colwidth', 600)\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "import seaborn as sns\n",
    "sns.set(style='darkgrid', font_scale=1.6)\n",
    "import spacy\n",
    "import random\n",
    "import collections\n",
    "import re\n",
    "import emoji\n",
    "\n",
    "from sklearn.decomposition import PCA\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import accuracy_score, confusion_matrix\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "from tensorflow.keras import layers\n",
    "import gensim"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a28c4b8f",
   "metadata": {
    "papermill": {
     "duration": 0.013457,
     "end_time": "2023-02-20T21:21:12.961165",
     "exception": false,
     "start_time": "2023-02-20T21:21:12.947708",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "**Load word vectors**\n",
    "\n",
    "The word vectors are stored in a kaggle dataset. We use the gensim library to load them."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "e20f44fc",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:21:12.992517Z",
     "iopub.status.busy": "2023-02-20T21:21:12.991821Z",
     "iopub.status.idle": "2023-02-20T21:21:18.177764Z",
     "shell.execute_reply": "2023-02-20T21:21:18.176329Z"
    },
    "papermill": {
     "duration": 5.205157,
     "end_time": "2023-02-20T21:21:18.180484",
     "exception": false,
     "start_time": "2023-02-20T21:21:12.975327",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 2.21 s, sys: 452 ms, total: 2.67 s\n",
      "Wall time: 5.18 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "# Path to dataset\n",
    "word2vec_path = \"../input/googlenewsvectorsnegative300/GoogleNews-vectors-negative300.bin\"\n",
    "\n",
    "# Load 200,000 most common words\n",
    "word2vec_model = gensim.models.KeyedVectors.load_word2vec_format(word2vec_path, binary=True, limit=200000) "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6a29c0ea",
   "metadata": {
    "papermill": {
     "duration": 0.013274,
     "end_time": "2023-02-20T21:21:18.207768",
     "exception": false,
     "start_time": "2023-02-20T21:21:18.194494",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "We can **retrieve** a single word vector by passing the token as the key."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "b2688883",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:21:18.236914Z",
     "iopub.status.busy": "2023-02-20T21:21:18.236420Z",
     "iopub.status.idle": "2023-02-20T21:21:18.253016Z",
     "shell.execute_reply": "2023-02-20T21:21:18.250567Z"
    },
    "papermill": {
     "duration": 0.034944,
     "end_time": "2023-02-20T21:21:18.256127",
     "exception": false,
     "start_time": "2023-02-20T21:21:18.221183",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Vector dimension: (300,)\n",
      "[ 0.0090332   0.12158203  0.3984375   0.21386719 -0.01672363  0.06396484\n",
      "  0.3125     -0.23535156 -0.10009766 -0.09326172  0.20507812 -0.03222656\n",
      " -0.21191406  0.00958252 -0.27734375  0.09619141 -0.1875     -0.04052734\n",
      "  0.2265625  -0.23632812 -0.04858398  0.03295898 -0.08056641  0.12255859\n",
      "  0.05859375 -0.09130859 -0.02685547  0.02587891 -0.11376953 -0.21972656\n",
      " -0.0378418  -0.15722656 -0.05981445 -0.01635742  0.11425781 -0.15234375\n",
      " -0.01086426  0.29296875  0.22265625 -0.00245667 -0.16308594 -0.11669922\n",
      " -0.20800781  0.31835938 -0.08740234  0.13378906  0.15429688  0.11816406\n",
      " -0.07470703  0.09326172 -0.32421875  0.01239014 -0.07226562 -0.10986328\n",
      " -0.07177734  0.16113281  0.05541992 -0.3984375  -0.04980469 -0.25195312\n",
      " -0.05249023 -0.05786133  0.16699219 -0.26953125 -0.04394531 -0.14257812\n",
      "  0.03662109  0.34765625 -0.10888672 -0.12011719 -0.19335938 -0.21972656\n",
      "  0.16015625 -0.06542969 -0.22070312  0.1484375   0.23144531  0.11669922\n",
      " -0.15429688  0.04174805 -0.08837891  0.21972656  0.08056641 -0.10888672\n",
      "  0.07080078 -0.11279297  0.00628662  0.37890625 -0.21582031  0.11035156\n",
      "  0.27148438 -0.29296875 -0.12988281 -0.10595703  0.15820312  0.21484375\n",
      "  0.27929688  0.04150391  0.22949219 -0.02038574  0.02307129  0.12353516\n",
      " -0.14355469 -0.1328125   0.03930664  0.01324463 -0.01965332 -0.10107422\n",
      "  0.27539062 -0.13867188 -0.10986328  0.06982422  0.14648438  0.203125\n",
      "  0.04736328  0.18066406 -0.10546875  0.20410156  0.22460938  0.09912109\n",
      "  0.08251953 -0.01672363 -0.09619141  0.1171875  -0.29101562 -0.0324707\n",
      " -0.05981445 -0.01818848 -0.09472656  0.20800781 -0.04663086 -0.30078125\n",
      " -0.05957031 -0.18554688 -0.06494141  0.22265625  0.21191406  0.01867676\n",
      " -0.140625   -0.04150391  0.15039062 -0.12158203 -0.10498047 -0.1875\n",
      " -0.29101562  0.0078125   0.25585938 -0.08642578  0.17578125 -0.03271484\n",
      "  0.19628906 -0.16796875 -0.1796875   0.05834961 -0.2578125  -0.23925781\n",
      "  0.15234375 -0.14355469 -0.01226807 -0.23925781 -0.24511719 -0.17675781\n",
      " -0.18261719  0.3828125   0.34765625  0.07226562  0.14941406 -0.16015625\n",
      "  0.02490234 -0.12207031  0.13476562  0.14941406 -0.08447266  0.10107422\n",
      "  0.13867188  0.01940918 -0.11328125 -0.08642578 -0.07470703  0.02758789\n",
      " -0.13183594 -0.03662109  0.06445312 -0.22167969  0.19726562 -0.09375\n",
      "  0.17871094 -0.01647949  0.05761719  0.09619141  0.04614258  0.00799561\n",
      "  0.06982422  0.05566406 -0.15820312  0.20214844  0.25       -0.11865234\n",
      "  0.05541992  0.00570679 -0.33984375 -0.09960938  0.02026367 -0.04785156\n",
      " -0.28515625 -0.28320312  0.24511719 -0.14648438 -0.32226562 -0.12158203\n",
      " -0.12988281  0.11328125  0.09716797  0.00958252  0.04541016  0.24121094\n",
      "  0.04077148  0.08789062  0.33203125 -0.02148438 -0.01507568 -0.17773438\n",
      "  0.26953125  0.10546875  0.13867188  0.07470703  0.13964844  0.01446533\n",
      " -0.06445312  0.16113281  0.22558594  0.09472656  0.07714844 -0.15039062\n",
      "  0.21679688  0.04443359  0.16992188  0.12597656  0.04321289 -0.08642578\n",
      "  0.30859375  0.1015625  -0.00656128  0.08496094 -0.17382812 -0.05834961\n",
      "  0.08740234  0.00732422 -0.22070312  0.13183594 -0.27734375 -0.33203125\n",
      "  0.11279297  0.0559082   0.17871094 -0.22851562  0.03369141  0.0267334\n",
      " -0.18554688  0.03320312  0.05053711 -0.109375   -0.26367188  0.07421875\n",
      "  0.07617188  0.26171875  0.20117188 -0.18066406 -0.10449219 -0.03076172\n",
      "  0.09033203 -0.15234375 -0.11816406 -0.10253906 -0.0703125  -0.13476562\n",
      " -0.08789062 -0.23144531  0.23828125  0.05761719  0.17382812  0.06445312\n",
      "  0.1953125   0.18945312 -0.05737305 -0.07910156  0.15332031 -0.17089844\n",
      "  0.11035156  0.34570312 -0.23144531  0.03955078 -0.3671875   0.03344727\n",
      "  0.06054688 -0.01989746  0.18164062 -0.171875    0.2890625   0.07470703]\n"
     ]
    }
   ],
   "source": [
    "# Get word vector\n",
    "science = word2vec_model['science']\n",
    "\n",
    "# Print shape and vector\n",
    "print('Vector dimension:', science.shape)\n",
    "print(science)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "15e8b6ef",
   "metadata": {
    "papermill": {
     "duration": 0.013778,
     "end_time": "2023-02-20T21:21:18.283929",
     "exception": false,
     "start_time": "2023-02-20T21:21:18.270151",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "And we can measure the similarity between any two words by using the `.similarity` method."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "aeee2a58",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:21:18.313219Z",
     "iopub.status.busy": "2023-02-20T21:21:18.312817Z",
     "iopub.status.idle": "2023-02-20T21:21:18.322242Z",
     "shell.execute_reply": "2023-02-20T21:21:18.320169Z"
    },
    "papermill": {
     "duration": 0.02753,
     "end_time": "2023-02-20T21:21:18.324973",
     "exception": false,
     "start_time": "2023-02-20T21:21:18.297443",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.630091\n",
      "0.23719835\n",
      "0.0036594481\n"
     ]
    }
   ],
   "source": [
    "print(word2vec_model.similarity('science', 'mathematics'))\n",
    "print(word2vec_model.similarity('science', 'space'))\n",
    "print(word2vec_model.similarity('science', 'beard'))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3777fd65",
   "metadata": {
    "papermill": {
     "duration": 0.013188,
     "end_time": "2023-02-20T21:21:18.351918",
     "exception": false,
     "start_time": "2023-02-20T21:21:18.338730",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "We can measure the similarity between sentences by using the `.n_similarity` method. This expects a list of words and works by **averaging** the word vectors in each sentence and computing the cosine similarity between the resulting two vectors. This doesn't take word order into account but works quite well."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "d6ce0383",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:21:18.381900Z",
     "iopub.status.busy": "2023-02-20T21:21:18.381344Z",
     "iopub.status.idle": "2023-02-20T21:21:18.395052Z",
     "shell.execute_reply": "2023-02-20T21:21:18.393668Z"
    },
    "papermill": {
     "duration": 0.031766,
     "end_time": "2023-02-20T21:21:18.397874",
     "exception": false,
     "start_time": "2023-02-20T21:21:18.366108",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6724018"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "word2vec_model.n_similarity(\"Man goes running\".split(),\"Woman goes swimming\".split())"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a3b58f2c",
   "metadata": {
    "papermill": {
     "duration": 0.014018,
     "end_time": "2023-02-20T21:21:18.426483",
     "exception": false,
     "start_time": "2023-02-20T21:21:18.412465",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "Gensim also has a `.most_similar` method, which returns the **closest** word vectors to the input."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "f0d87b0f",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:21:18.460469Z",
     "iopub.status.busy": "2023-02-20T21:21:18.460042Z",
     "iopub.status.idle": "2023-02-20T21:21:18.630664Z",
     "shell.execute_reply": "2023-02-20T21:21:18.628506Z"
    },
    "papermill": {
     "duration": 0.194201,
     "end_time": "2023-02-20T21:21:18.636054",
     "exception": false,
     "start_time": "2023-02-20T21:21:18.441853",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('custard', 0.5662760734558105),\n",
       " ('Jell_O', 0.5641272068023682),\n",
       " ('chocolate_pudding', 0.5506998300552368),\n",
       " ('marshmallow', 0.546853244304657),\n",
       " ('marmalade', 0.5442028641700745),\n",
       " ('jellies', 0.5395851731300354),\n",
       " ('Jello', 0.5389018654823303),\n",
       " ('gummy', 0.5371953248977661),\n",
       " ('gelatinous', 0.5302338004112244),\n",
       " ('mousse', 0.5300459861755371)]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "word2vec_model.most_similar(positive=['jelly'], topn=10)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9217bec2",
   "metadata": {
    "papermill": {
     "duration": 0.035645,
     "end_time": "2023-02-20T21:21:18.708408",
     "exception": false,
     "start_time": "2023-02-20T21:21:18.672763",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "We can visualize the word vectors in **2 dimensions by using PCA**, which is **dimensionality reduction technique**. This finds the subspace that **maximizes the variance** of the data, in other words it finds the 2-D plane where the data points are as spread out as possible. This allows us to get a better idea of the structure and patterns in the word vectors."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "245174df",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:21:18.752852Z",
     "iopub.status.busy": "2023-02-20T21:21:18.752416Z",
     "iopub.status.idle": "2023-02-20T21:21:19.121401Z",
     "shell.execute_reply": "2023-02-20T21:21:19.119965Z"
    },
    "papermill": {
     "duration": 0.388192,
     "end_time": "2023-02-20T21:21:19.124268",
     "exception": false,
     "start_time": "2023-02-20T21:21:18.736076",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Vectors to transform\n",
    "words = [\"football\",\"tennis\",\"rugby\",\"river\",\"ocean\",\"water\",\"brownie\",\"jelly\",\"cake\"] #['swim', 'swimming', 'cat', 'dog', 'feline', 'road', 'car', 'bus']\n",
    "word_vectors = np.array([word2vec_model[w] for w in words])\n",
    "\n",
    "# PCA\n",
    "pca = PCA(n_components=2)\n",
    "X_pca = pca.fit_transform(word_vectors)\n",
    "\n",
    "# Plot transformed vectors\n",
    "plt.figure(figsize=(12,6))\n",
    "plt.scatter(X_pca[:,0], X_pca[:,1], edgecolors='k', c='g', s=128)\n",
    "for word, (x,y) in zip(words, X_pca):\n",
    "    plt.text(x+0.01,y+0.01, word)\n",
    "plt.title(\"PCA plot of word vectors\", y=1.02)\n",
    "plt.xlabel(\"PC1\")\n",
    "plt.ylabel(\"PC2\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b78a2eea",
   "metadata": {
    "papermill": {
     "duration": 0.015706,
     "end_time": "2023-02-20T21:21:19.155611",
     "exception": false,
     "start_time": "2023-02-20T21:21:19.139905",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "Like we discussed previously, Word2vec can solve **analogies**. In gensim, we just need to specify the **positive terms** and the **negative terms** and it will retrieve the closest vector to the result.\n",
    "\n",
    "For example, \"France\"-\"Paris\"+\"Tokyo\" should give us \"Japan\" as Tokyo is the capital of Japan and indeed it does."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "235fe168",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:21:19.188588Z",
     "iopub.status.busy": "2023-02-20T21:21:19.187493Z",
     "iopub.status.idle": "2023-02-20T21:21:19.230464Z",
     "shell.execute_reply": "2023-02-20T21:21:19.228501Z"
    },
    "papermill": {
     "duration": 0.064548,
     "end_time": "2023-02-20T21:21:19.235500",
     "exception": false,
     "start_time": "2023-02-20T21:21:19.170952",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('Japan', 0.8167770504951477),\n",
       " ('Japanese', 0.6480902433395386),\n",
       " ('South_Korea', 0.6141558885574341)]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "word2vec_model.most_similar(positive=[\"France\", \"Tokyo\"], negative=[\"Paris\"], topn=3)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5636a636",
   "metadata": {
    "papermill": {
     "duration": 0.037347,
     "end_time": "2023-02-20T21:21:19.311575",
     "exception": false,
     "start_time": "2023-02-20T21:21:19.274228",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## <p style=\"font-family:JetBrains Mono; font-weight:normal; letter-spacing: 1px; color:#207d06; font-size:100%; text-align:left;padding: 0px; border-bottom: 3px solid #207d06;\">4.2 Modelling with pre-trained word embeddings</p>\n",
    "\n",
    "To showcase how word embeddings can be used in neural networks, we are going to perform sentiment analysis on a **spotify app reviews dataset**. Each review has a corresponding number of **stars** awarded **between 1 and 5**, which is what we are going to try to predict."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "eb4120a9",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:21:19.354738Z",
     "iopub.status.busy": "2023-02-20T21:21:19.354219Z",
     "iopub.status.idle": "2023-02-20T21:21:19.853772Z",
     "shell.execute_reply": "2023-02-20T21:21:19.852015Z"
    },
    "papermill": {
     "duration": 0.519464,
     "end_time": "2023-02-20T21:21:19.856805",
     "exception": false,
     "start_time": "2023-02-20T21:21:19.337341",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(61594, 2)\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Review</th>\n",
       "      <th>Rating</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Great music service, the audio is high quality and the app is easy to use. Also very quick and friendly support.</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Please ignore previous negative rating. This app is super great. I give it five stars+</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>This pop-up \"Get the best Spotify experience on Android 12\" is too annoying. Please let's get rid of this.</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Really buggy and terrible to use as of recently</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Dear Spotify why do I get songs that I didn't put on my playlist??? And why do we have shuffle play?</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                                                                             Review  \\\n",
       "0  Great music service, the audio is high quality and the app is easy to use. Also very quick and friendly support.   \n",
       "1                            Please ignore previous negative rating. This app is super great. I give it five stars+   \n",
       "2        This pop-up \"Get the best Spotify experience on Android 12\" is too annoying. Please let's get rid of this.   \n",
       "3                                                                   Really buggy and terrible to use as of recently   \n",
       "4              Dear Spotify why do I get songs that I didn't put on my playlist??? And why do we have shuffle play?   \n",
       "\n",
       "   Rating  \n",
       "0       5  \n",
       "1       5  \n",
       "2       4  \n",
       "3       1  \n",
       "4       1  "
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Load data\n",
    "data = pd.read_csv(\"/kaggle/input/spotify-app-reviews-2022/reviews.csv\")\n",
    "data = data[[\"Review\",\"Rating\"]]\n",
    "\n",
    "# Print shape and preview\n",
    "print(data.shape)\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5942e4f6",
   "metadata": {
    "papermill": {
     "duration": 0.014371,
     "end_time": "2023-02-20T21:21:19.886321",
     "exception": false,
     "start_time": "2023-02-20T21:21:19.871950",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "Currently some of the reviews are very **long** (the maximum is over 3000 characters!). Later in our keras model we will truncate these so they are **not longer than 250 characters**, which will speed up training time."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "3784a322",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:21:19.918962Z",
     "iopub.status.busy": "2023-02-20T21:21:19.918006Z",
     "iopub.status.idle": "2023-02-20T21:21:21.123990Z",
     "shell.execute_reply": "2023-02-20T21:21:21.121997Z"
    },
    "papermill": {
     "duration": 1.226081,
     "end_time": "2023-02-20T21:21:21.127080",
     "exception": false,
     "start_time": "2023-02-20T21:21:19.900999",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAokAAAEaCAYAAAB9zc1RAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8qNh9FAAAACXBIWXMAAAsTAAALEwEAmpwYAAA61ElEQVR4nO3de1yUZd4/8M8MMIOiBuoIKv5o5SAqB5UwklpPSYK6mYoaJXko9dk1NU9lvZ4edzN1k7S0R63II7AKafqI4O6juK15Ig+LkKOyuWyhGydBQQ7DHH5/8Mwttzen0RmHmfm8/0mu+XLPdX29Xy+/3dfhlhkMBgOIiIiIiBqRW7sDRERERNT+sEgkIiIiIgkWiUREREQkwSKRiIiIiCRYJBIRERGRBItEIiIiIpJgkUhEREREEs7W7oC9Ki+/B73eMkdQduvWCWVlVRa5tq1hLsSYDzHmQ4z5EGM+7mMuxBwpH3K5DB4ebk1+xiLRQvR6g8WKROP1qQFzIcZ8iDEfYsyHGPNxH3MhxnxwupmIiIiImsAikYiIiIgkWCQSERERkQSLRCIiIiKSYJFIRERERBIsEomIiIhIgkWiHZHJrN0DIiIishcsEu2ETAZ8dUTNQpGIiIjMgkWiHdHU66zdBSIiIrITLBKJiIiISIJFIhERERFJsEgkIiIiIgkWiUREREQkwSKRiIiIiCRYJBIRERGRhLM1v/yHH37A4cOHcfbsWRQWFqK+vh6+vr6IjY3F9OnTIWt06N+oUaNw8+ZNyTUUCgVyc3Ml7devX0dCQgIuXLgAg8GAwYMHY/ny5QgMDJTEVlRUYMOGDTh+/DgqKyvh5+eHN954A9HR0eYdMBEREZGNsGqRmJiYiNOnT2PMmDGYNm0aNBoNMjMzsWrVKly5cgUffPCBKN7b2xsLFy4UtTk5OUmuW1BQgLi4OLi6umL+/PlQKpVITk5GXFwcUlNT4efnJ8RqNBrMmjUL+fn5iI+Px5NPPonMzEwsXrwYdXV1mDhxokXGbkkyGWAwWLsXREREZMusWiS++uqrWLduHZRKpdA2Y8YMvPbaa0hNTUV8fDz8/f2Fz5544gm8+OKLrV53w4YNqK6uRkpKCgICAgAA0dHRGDt2LNavX4/PP/9ciE1NTcWVK1ewatUqvPzyywCAKVOmIC4uDuvWrcPYsWPh6upqriFbnEwGJKarMWdcfxaKRERE9NCsuiYxLCxMVCACgFwuR1RUFICGKeMHabVaVFVVNXvNe/fuISsrCxEREUKBCAAqlQrR0dE4efIkKioqhPb09HR07NgRkydPFvUhLi4O5eXlOHXq1MMOz2r45hUiIiJ6VO1y40pxcTEAoGvXrqL269evY9CgQQgLC0N4eDjeffdd3L59WxJTX1+PkJAQyXVDQ0Oh0+mgVqsBAHq9Hmq1Gv3794dCoZDEAmhyvSMRERGRvbPqdHNTSktLsW/fPvTu3RthYWFCu7+/P2JjY+Hr64u6ujqcOnUKBw4cwPnz5/H111+jS5cuAO4XmD169JBc29hWVFQEALhz5w5qa2vbFEtERETkSNpVkajRaLBw4UJUVlbik08+ET3da7yOEAAmTJiAwMBArF27Fjt37hQ2tNTU1ACA5Mlg4zZjTG1tbbOxxmlwY4ypunXr9FC/11YqVWdJm7OzM7p16wxnZ2d07y793F41lQtHxnyIMR9izIcY83EfcyHGfLSjIlGr1WLRokW4ePEi/vCHP2DYsGGt/k58fDw+/fRTnDp1SigSjcWdRqORxBvbjBtRWoqtq6sTxZqqrKwKer1ldo6oVJ1RUlIpapPJGnJYVlYJrVaL0tJKh9i40lQuHBnzIcZ8iDEfYszHfcyFmCPlQy6XNftgq12sSdTpdFi6dCmysrLw3nvvYerUqW36Pblcjp49e4o2onh5eQG4P+3cmLHN09MTAODu7g6lUtlibFNT0URERET2zupFol6vx4oVK3D06FG8/fbbmDFjRpt/V6vV4tatW+jWrZvQFhAQABcXF1y+fFkSn5OTAycnJ/Tv3x9AQ5E5YMAAqNVqydPEnJwcAEBwcPDDDIuIiIjIplm1SNTr9Vi5ciXS09OxZMkSzJ49u8m4xk8KG9u2bRtqamowYsQIoc3NzQ0jRozA2bNnkZ+fL7SXlJTg6NGjiIyMhIeHh9AeExOD6upq7N+/X9SvlJQUuLu7IzIy8tEGSURERGSDrLom8aOPPsLBgwcRHBwMLy8vHDp0SPT5kCFD0KdPHxw8eBD79+/Hc889B29vb2g0Gnz33Xc4efIkgoKCJE8flyxZgjNnzmDWrFmYOXMmlEolkpKSoNPpsHz5clHstGnTsH//fnz44YcoLCyEj48PMjMzcenSJaxduxYdOnSweB6IiIiI2hurv7sZaDiLcMWKFZLP165diz59+iA4OBhnz55FRkaGcC6ij48P3nzzTbz++uuSzSV9+/ZFcnIyEhISsGXLFuHdzRs3bhQdsA00bF7ZuXMnPv74Yxw4cABVVVXw8/PDxo0bERMTY6GRExEREbVvVi0S9+zZ06a4sLAw0ZmJbREYGIjExMQ2xXp4eGD16tVYvXq1Sd9BREREZK+svnGFiIiIiNofFolEREREJMEi0Y7JZNbuAREREdkqFol2SiYDvjqiZqFIRERED4VFop1oqhjU1Osef0eIiIjILrBItAMyGbD76DUoXOTYffSatbtDREREdoBFop3QaHWi/xIRERE9ChaJRERERCTBIpGIiIiIJFgkEhEREZEEi0QiIiIikmCRSEREREQSLBKJiIiISIJFoh3gW1WIiIjI3Fgk2jjjQdqtxRARERGZgkWiHWjpAG2+w5mIiIgeBotEB8B3OBMREZGpWCQSERERkQSLRDukcJG3uk6RiIiIqCUsEu1US+sUiYiIiFrDIpGIiIiIJFgkEhEREZEEi0QiIiIikmCRSEREREQSLBKJiIiISMLZml/+ww8/4PDhwzh79iwKCwtRX18PX19fxMbGYvr06ZA98JqQY8eOYevWrcjPz4ebmxtGjhyJZcuWoWvXrpJrX79+HQkJCbhw4QIMBgMGDx6M5cuXIzAwUBJbUVGBDRs24Pjx46isrISfnx/eeOMNREdHW2zsRERERO2ZVYvExMREnD59GmPGjMG0adOg0WiQmZmJVatW4cqVK/jggw+E2IyMDLz11lsICgrCypUrUVZWhu3btyM3NxdpaWlwdXUVYgsKChAXFwdXV1fMnz8fSqUSycnJiIuLQ2pqKvz8/IRYjUaDWbNmIT8/H/Hx8XjyySeRmZmJxYsXo66uDhMnTnycKSEiIiJqF6xaJL766qtYt24dlEql0DZjxgy89tprSE1NRXx8PPz9/VFfX481a9bA29sbSUlJ6NChAwAgODgYc+fORXJyMubMmSNcY8OGDaiurkZKSgoCAgIAANHR0Rg7dizWr1+Pzz//XIhNTU3FlStXsGrVKrz88ssAgClTpiAuLg7r1q3D2LFjRQWoLZPJAIPB2r0gIiIiW2DVNYlhYWGiAhEA5HI5oqKiADRMGQNAdnY2SkpKEBsbKxSIADB8+HD4+PggPT1daLt37x6ysrIQEREhFIgAoFKpEB0djZMnT6KiokJoT09PR8eOHTF58mRRH+Li4lBeXo5Tp06ZdczWIpMBXx1R44EZfCIiIqImtcuNK8XFxQAgrDXMy8sDAISGhkpiQ0JCcO3aNWg0GgANhWV9fT1CQkIksaGhodDpdFCr1QAAvV4PtVqN/v37Q6FQSGIBIDc310yjevwefD2fpp5vYSEiIqK2sep0c1NKS0uxb98+9O7dG2FhYQDuF409evSQxPfo0QM6nQ6lpaXo1atXq7EAUFRUBAC4c+cOamtr2xRrqm7dOj3U77WVStVZ+LOzU/N/jXoD4OzsjG7dOsPZ2Rndu3duNtZWNc4FMR8PYj7EmA8x5uM+5kKM+WhnRaJGo8HChQtRWVmJTz75RHi6V1NTAwCSp30AhOlqY0xLsQ9er7a2ttXrGmNMVVZWBb3eMgsAVarOKCmpBNAwjazVaVuM12r1KCurhFarRWlppV2tS2ycC2I+HsR8iDEfYszHfcyFmCPlQy6XNftgq90UiVqtFosWLcLFixfxhz/8AcOGDRM+MxZsxinlxurq6gBA2FzSUqyxrS2xD16XiIiIyJG0izWJOp0OS5cuRVZWFt577z1MnTpV9LmXlxeA+9POjRUXF8PJyQkqlapNsQDg6ekJAHB3d4dSqWwxtqmpaCIiIiJ7Z/UiUa/XY8WKFTh69CjefvttzJgxQxIzcOBAAEBOTo7ks9zcXPj7+wtTxgEBAXBxccHly5clsTk5OXByckL//v0BNOxiHjBgANRqteRpovG7goODH22ARERERDbIqkWiXq/HypUrkZ6ejiVLlmD27NlNxg0dOhQqlQppaWmiNYLffvstCgoKMH78eKHNzc0NI0aMwNmzZ5Gfny+0l5SU4OjRo4iMjISHh4fQHhMTg+rqauzfv1/Ur5SUFLi7uyMyMtKcQyYiIiKyCVZdk/jRRx/h4MGDCA4OhpeXFw4dOiT6fMiQIejTpw8UCgXeeecdLF26FK+88gqmTJmCsrIy7NixA35+fnj11VdFv7dkyRKcOXMGs2bNwsyZM6FUKpGUlASdTofly5eLYqdNm4b9+/fjww8/RGFhIXx8fJCZmYlLly5h7dq1onMZiYiIiByF1d/dDDRMGa9YsULy+dq1a9GnTx8AwPjx46FQKLBt2zasWbMGbm5uiIqKwrJlyySFXN++fZGcnIyEhARs2bJFeHfzxo0bRQdsAw2bV3bu3ImPP/4YBw4cQFVVFfz8/LBx40bExMRYaORERERE7ZtVi8Q9e/aYFB8VFSW8jaU1gYGBSExMbFOsh4cHVq9ejdWrV5vUHyIiIiJ7ZfWNK0RERETU/rBIJCIiIiIJFolEREREJMEikYiIiIgkWCQSERERkQSLRAckk1m7B0RERNTesUh0AAoXOXYfvQagoUD86oiahSIRERG1iEWig9Bodff/XK9rIZKIiIiIRSIRERERNYFFIhERERFJsEgkIiIiIgkWiUREREQkYVKR+P333+PWrVstxvzyyy/4/vvvH6lTRERERGRdJhWJ8fHxOHDgQIsx//M//4P4+PhH6hQRERERWZdJRaLBYGg1Rq/XQ8ZD+No9/hURERFRS8y+JvHatWvo0qWLuS9LZqRwkSMxnQdqExERUfOcWwt4cOr4m2++QXZ2tiROr9ejqKgIhYWFiIqKMl8PySJ4oDYRERG1pNUisXFBKJPJcPPmTdy8eVMSJ5fL8cQTT+CFF17Au+++a95eklk0fj0fERERUUtaLRKvXr0q/DkwMBALFizAggULLNopspzGr+cjIiIiak6rRWJju3fvRu/evS3VFyIiIiJqJ0wqEocOHWqpfpCVyGRAGzatExERkYMxqUgEAI1Gg2PHjiE3NxeVlZXQ6aTTlzKZDGvWrDFLB8lyZDIgMV2NOeP6s1AkIiIiEZOKxJ9//hmzZ89GYWFhi2cmski0HdzlTERERE0xqUj88MMP8fPPP+Oll17CpEmT4OnpCScnJ0v1jYiIiIisxKQiMTs7G88++yzWrl1rtg7cu3cPO3bsQF5eHvLy8lBSUoIXXngBmzZtksSOGjWqyeN3FAoFcnNzJe3Xr19HQkICLly4AIPBgMGDB2P58uUIDAyUxFZUVGDDhg04fvw4Kisr4efnhzfeeAPR0dHmGSgRERGRDTGpSJTL5QgICDBrB8rLy7F582aoVCoEBQXhxIkTLcZ7e3tj4cKForamnmYWFBQgLi4Orq6umD9/PpRKJZKTkxEXF4fU1FT4+fkJsRqNBrNmzUJ+fj7i4+Px5JNPIjMzE4sXL0ZdXR0mTpxolrFaAt+aQkRERJZg8u5mtVpt1g706NEDf/vb3+Dp6QkA6NevX4vxTzzxBF588cVWr7thwwZUV1cjJSVFKGyjo6MxduxYrF+/Hp9//rkQm5qaiitXrmDVqlV4+eWXAQBTpkxBXFwc1q1bh7Fjx8LV1fVhh2gxMhl4ODYRERFZhEnvbn7nnXegVquxc+dOs3VAoVAIBWJbabVaVFVVNfv5vXv3kJWVhYiICNGTT5VKhejoaJw8eRIVFRVCe3p6Ojp27IjJkycLbXK5HHFxcSgvL8epU6dM6t/jxMOxiYiIyBJMepK4detW+Pv7449//CNSUlLQr18/dOrUSRJnyd3N169fx6BBg1BfX48uXbpgzJgxWLZsGbp27SqKqa+vR0hIiOT3Q0NDkZaWBrVajWeeeQZ6vR5qtRoDBw6EQqGQxAJAbm4uRo8ebZHxEBEREbVHJhWJ33zzjfDnn376CT/99FOTcZYqEv39/REbGwtfX1/U1dXh1KlTOHDgAM6fP4+vv/4aXbp0AQAUFxcDaJjKfpCxraioCABw584d1NbWtimWiIiIyFGYVCQeP37cUv1ok8brCAFgwoQJCAwMxNq1a7Fz505hQ0tNTQ0ASJ4MNm4zxtTW1jYbq1QqRTGm6NZN+oTVnFSqzgAAZyeTz0MHAOgNenTr1hnOzs7o3r2zObv22BlzQQ2YDzHmQ4z5EGM+7mMuxJgPE4vE9vje5vj4eHz66ac4deqUUCQaizuNRiOJN7YZN6K0FFtXVyeKNUVZWRX0esu8xkSl6oySkkrIZIBWp32oa8hlwMbk89BqtSgtrbTZN64Yc0ENmA8x5kOM+RBjPu5jLsQcKR9yuazZB1smbVxpj+RyOXr27CnaiOLl5QXg/rRzY8Y242YZd3d3KJXKFmObmoq2B9z0QkRERM0x6UniwYMH2xz7uM4W1Gq1uHXrFgYMGCC0BQQEwMXFBZcvX5bE5+TkwMnJCf379wfQUGQOGDAAarUaGo1GNO2ck5MDAAgODrbwKKxPJoPNPk0kIiIi8zOpSHznnXcga+X0ZoPBAJlMZvYisaKiAu7u7pL2bdu2oaamBiNGjBDa3NzcMGLECGRlZSE/Px/+/v4AgJKSEhw9ehSRkZHw8PAQ4mNiYnDp0iXs379fOCdRr9cjJSUF7u7uiIyMNOtY2huZDEhMV2POuP4sFImIiAiAiUVic6/jq6qqwpUrV3D48GGMHDkSo0aNMqkTSUlJuHv3rvDzjRs3sGXLFgBAeHg4wsPDcfDgQezfvx/PPfccvL29odFo8N133+HkyZMICgrCjBkzRNdcsmQJzpw5g1mzZmHmzJlQKpVISkqCTqfD8uXLRbHTpk3D/v378eGHH6KwsBA+Pj7IzMzEpUuXsHbtWnTo0MGk8dgiTT2nnomIiOg+k4rEl156qcXPX375ZcTHx2P69OkmdWL79u2idzLn5+fj008/BQAsWLAA4eHhCA4OxtmzZ5GRkYHbt28DAHx8fPDmm2/i9ddfl2wu6du3L5KTk5GQkIAtW7YI727euHGj5NWCSqUSO3fuxMcff4wDBw6gqqoKfn5+2LhxI2JiYkwaCxEREZE9eLjzU5oREhKCUaNGYdOmTSZN0WZlZbUaExYWhrCwMJP6ExgYiMTExDbFenh4YPXq1Vi9erVJ30FERERkj8y+u9nT0xPXrvF9wkRERES2zKxFYn19PU6dOgU3NzdzXpaIiIiIHjOzHIGj0+lQVFSEzMxM/OMf/zB5TSIRERERtS9mOQLH8H/npshkMkRFReHtt982T++IiIiIyCrMcgSOTCZDly5dMHDgQOFNJmS7eLA2ERERmfUIHLJ9Mhnw1REerE1EROTobP7dzWR+PFibiIiIHuqcxEOHDuHgwYNQq9WoqqpCp06dMGDAALz44ot48cUXzd1HIiIiInrMTCoSdTodFi1ahOPHj8NgMKBz587o06cPysrKcPr0aZw5cwZ//vOfsXnzZjg5OVmqz0RERERkYSZNN+/atQvHjh3DsGHDcPDgQXz//ffIzMxEdnY2Dh48iMjISJw4cQI7d+60UHeJiIiI6HEwqUj85ptv4O/vjy+//BKBgYGizwIDA/HFF1/Az8+v2fMUiYiIiMg2mFQk/vTTTxg+fDjk8qZ/TS6XY/jw4fjpp5/M0jmyPIWLHLuP8jWKREREJGZSkahUKnH37t0WY+7evQulUvlInaLHS6PlbmYiIiISM6lIDA4ORkZGBv75z382+flPP/2EjIwMhISEmKVzRERERGQdJu1unj9/PmbOnInJkydj2rRpGDp0KLp164aysjJ8//33SE1NRU1NDebNm2ep/hIRERHRY2BSkRgeHo7169fj/fffx44dO0S7mA0GAzp16oSPPvoI4eHh5u4nPQZNvJabiIiIHJTJh2nHxMTg17/+NY4dO4arV68Kh2kHBgbi+eefR6dOnSzRT7IwhYscW77Jg8KFL+EhIiKiNhaJJ06cwJ07dzBu3Di4uLigU6dOmDhxoihGo9EgMzMTTzzxBEaMGGGBrpKl1Wl0MBjuP1GUycD3NxMRETmoVh8bXblyBb/73e9w5coVuLi4NBunUCigVqvx29/+FlevXjVrJ+nxk8mAr46oOQVNRETkoFotEtPS0qBUKrFgwYJWL/bb3/4Wrq6u2Lt3r1k6R9alqefROERERI6q1SLx3LlzeOaZZ9ClS5dWL9alSxcMGzYM2dnZZukcEREREVlHq0Xiv//9b/Tt27fNF/Tx8cGtW7ceqVNkXXwLCxEREbVaJOr1ehi4e8Hh8C0sREREjq3VIlGlUuHHH39s8wV//PFHdO/e/ZE6RURERETW1WqROGTIEJw5cwZFRUWtXqyoqAinT5/GU089ZZbOEREREZF1tFokTp8+HXV1dVi0aBHu3LnTbNydO3ewaNEi1NfXY9q0aW3uwL179/DZZ59h/vz5ePbZZ9GvXz8sXLiw2fhjx45h8uTJCAkJwTPPPIN3330Xt2/fbjL2+vXrmDt3LsLCwjBkyBDMmTOn2eN5Kioq8P777yMyMhIhISGYNGkSMjMz2zwOevx4PA8REZHltHqY9pAhQzB9+nTs3bsXMTExmD59OoYOHQpPT08ADU8Pz507h9TUVJSWliIuLg6DBw9ucwfKy8uxefNmqFQqBAUF4cSJE83GZmRk4K233kJQUBBWrlyJsrIybN++Hbm5uUhLS4Orq6sQW1BQgLi4OLi6umL+/PlQKpVITk5GXFwcUlNT4efnJ8RqNBrMmjUL+fn5iI+Px5NPPonMzEwsXrwYdXV1koPDyfqM5zi+/dpQa3eFiIjILrXpjSvvv/8+OnTogF27dmHLli3YsmWL6HODwQC5XI7Zs2dj6dKlJnWgR48e+Nvf/iYUnf369Wsyrr6+HmvWrIG3tzeSkpLQoUMHAEBwcDDmzp2L5ORkzJkzR4jfsGEDqqurkZKSgoCAAABAdHQ0xo4di/Xr1+Pzzz8XYlNTU3HlyhWsWrUKL7/8MgBgypQpiIuLw7p16zB27FhRAepo2uubV3iOIxERkeW06UW9crkcb7/9NjIyMjB37lyEh4fjV7/6FX71q18hPDwc8+bNQ0ZGBlasWAEnJyeTOqBQKIQCsSXZ2dkoKSlBbGysUCACwPDhw+Hj44P09HSh7d69e8jKykJERIRQIAINm3Cio6Nx8uRJVFRUCO3p6eno2LEjJk+eLBpzXFwcysvLcerUKZPGZE/45hUiIiLH1KYniUZPPvkk3nrrLUv1pUV5eXkAgNDQUMlnISEhyMjIgEajgUKhwPXr11FfX4+QkBBJbGhoKNLS0qBWq/HMM89Ar9dDrVZj4MCBUCgUklgAyM3NxejRoy0wqkfzuAo3PrEjIiJyPG16ktgeFBcXA2iYnn5Qjx49oNPpUFpa2qZYAMJu7Tt37qC2trZNse2JTAaLH3jNQ7WJiIgcl0lPEq2ppqYGACRP+wBAqVSKYlqKNbYZY2pra1u9rjHGFN26dTL5d0zRvXtn6A0yODuZ76/Q2cUAZxc5nBstGdAbAGdnZ3Tv3tls32Muzs4NY1ep2l/frIn5EGM+xJgPMebjPuZCjPmwoSLRWLBpNBrJZ3V1dQAgbC5pKdbY1pbYB69rirKyKuj1ltntoVJ1RmlpJbQ6rVmvq63XQg4nOMnE/dZq9Sgrq0RiuhpzxvVvF5tYZDJAq20Yf0lJpZV7036oVJ2Zj0aYDzHmQ4z5uI+5EHOkfMjlsmYfbNnMdLOXlxeA+1PJjRUXF8PJyQkqlapNsQCEzTLu7u5QKpUtxjY1Fe2IuDaRiIjIcdhMkThw4EAAQE5OjuSz3Nxc+Pv7C1PGAQEBcHFxweXLlyWxOTk5cHJyQv/+/QE07GIeMGAA1Gq15Gmi8buCg4PNOhYiIiKi9s5misShQ4dCpVIhLS1NtEbw22+/RUFBAcaPHy+0ubm5YcSIETh79izy8/OF9pKSEhw9ehSRkZHw8PAQ2mNiYlBdXY39+/cLbXq9HikpKXB3d0dkZKSFR0dERETUvrSLNYlJSUm4e/eu8PONGzeEA7vDw8MRHh4OhUKBd955B0uXLsUrr7yCKVOmoKysDDt27ICfnx9effVV0TWXLFmCM2fOYNasWZg5cyaUSiWSkpKg0+mwfPlyUey0adOwf/9+fPjhhygsLISPjw8yMzNx6dIlrF27VnQuIxEREZEjaBdF4vbt23Hz5k3h5/z8fHz66acAgAULFiA8PBwAMH78eCgUCmzbtg1r1qyBm5sboqKisGzZMkkh17dvXyQnJyMhIQFbtmyBwWDA4MGDsXHjRtEB20DD5pWdO3fi448/xoEDB1BVVQU/Pz9s3LgRMTExFh49ERERUfvTLorErKysNsdGRUUhKiqqTbGBgYFITExsU6yHhwdWr16N1atXt7kvRERERPbKZtYkEhEREdHjwyKRWvXgm1f4HmciIiL7xyKR2kSjbTgjUSYDvjqiZqFIRERk51gkksl4qDYREZH9Y5FIRERERBIsEomIiIhIgkUiEREREUmwSCQiIiIiCRaJRERERCTBIpGIiIiIJFgk0kORyXioNhERkT1jkUgmM76BhYdqExER2S9na3eAbEfj1/NptDpo6vVW7hERERFZCp8kkkmMr+cz4pNEIiIi+8QikR6awkWOxHROORMREdkjFon0SPgeZyIiIvvEIpGIiIiIJFgkkllwypmIiMi+sEikRyaTgcfhEBER2RkWiWQWXJtIRERkX1gkEhEREZEEi0QiIiIikmCRSEREREQSLBKJiIiISIJFIhERERFJOFu7A6YoLCzE6NGjm/zs2WefxVdffSVqO3bsGLZu3Yr8/Hy4ublh5MiRWLZsGbp27Sr5/evXryMhIQEXLlyAwWDA4MGDsXz5cgQGBlpkLERERETtmU0ViUbDhw/HuHHjRG09evQQ/ZyRkYG33noLQUFBWLlyJcrKyrB9+3bk5uYiLS0Nrq6uQmxBQQHi4uLg6uqK+fPnQ6lUIjk5GXFxcUhNTYWfn99jGRcRERFRe2GTRaKvry9efPHFZj+vr6/HmjVr4O3tjaSkJHTo0AEAEBwcjLlz5yI5ORlz5swR4jds2IDq6mqkpKQgICAAABAdHY2xY8di/fr1+Pzzzy07ICIiIqJ2xmbXJNbV1aGmpqbJz7Kzs1FSUoLY2FihQAQankD6+PggPT1daLt37x6ysrIQEREhFIgAoFKpEB0djZMnT6KiosJi47B1Chc5dh+9JvzMt64QERHZB5ssEvfu3YvQ0FAMGjQII0eOxNatW6HVaoXP8/LyAAChoaGS3w0JCcG1a9eg0WgANKxFrK+vR0hIiCQ2NDQUOp0OarXaQiOxDxptw9tW+Ho+IiIi+2FT081yuRwREREYM2YMevbsibKyMhw6dAiffPIJ1Go1Nm3aBAAoLi4GIF2naGzT6XQoLS1Fr169Wo0FgKKiIksNye7w9XxERET2waaKxF69emHXrl2ittjYWCxYsAB//vOfcfbsWURERAjT0AqFQnINpVIJAEJMS7HGtuamtVvSrVsnk3/HFN27d4azk3n/+pxdDHB2kcPZycnk39Ub9OjWrTOcnZ3RvXtns/arOc7ODeNXqR7P99kK5kOM+RBjPsSYj/uYCzHmw8aKxKbIZDLMmzcPx44dw6lTpxARESEUgsYp5cbq6uoAQNjd3FKssa3xTui2Kiurgl5vMPn32kKl6ozS0kpoddrWg02grddCDic4yUzvt1wGbEw+D61Wi9LSShgsM3SBTAZhiUFJSaVlv8yGqFSdmY9GmA8x5kOM+biPuRBzpHzI5bJmH2zZ5JrEB/Xu3RsAhA0mXl5eAO5POzdWXFwMJycnqFSqNsUCgKenp9n7bI+MaxMBbmAhIiKydXZRJP7rX/8CAHTr1g0AMHDgQABATk6OJDY3Nxf+/v7CVHJAQABcXFxw+fJlSWxOTg6cnJzQv39/S3XdLnEDCxERke2zqSKxvLxc0qbVavHZZ58BAEaOHAkAGDp0KFQqFdLS0lBbWyvEfvvttygoKMD48eOFNjc3N4wYMQJnz55Ffn6+0F5SUoKjR48iMjISHh4elhqS3eIGFiIiIttmU2sS//M//xPV1dUYNGgQvLy8UFZWhiNHjiA/Px9xcXHCkTcKhQLvvPMOli5dildeeQVTpkxBWVkZduzYAT8/P7z66qui6y5ZsgRnzpzBrFmzMHPmTCiVSiQlJUGn02H58uXWGCoRERGRVdlUkTh8+HAcOnQIe/fuxd27d6FUKtGvXz/88Y9/xMSJE0Wx48ePh0KhwLZt27BmzRq4ubkhKioKy5YtEx2wDQB9+/ZFcnIyEhISsGXLFuHdzRs3bhQdsE2te/BwbSIiIrJNNlUkxsbGIjY2ts3xUVFRiIqKalNsYGAgEhMTH7Zr1MiDG1gssdOZ6x2JiIgsy6bWJJJtsdQGFpkMfFpJRERkYSwSySKM086W2sDS+GklERERmR+LRLKYxu905vQwERGRbWGRSBZlfKLIcxOJiIhsi01tXCHbpNHqoKnXW7sbREREZAI+SSSbwyeSRERElscikR4bc6xN5M5mIiKix4NFIj0Wjdcmyh/xruPOZiIiIstjkUiPTUNxZ0Bi+qMXikRERGRZ/KearKChUHzUqWeFixwfJ1/gGkUiIiILYJFoo2y9MNLU68wyBu6aJiIisgwWiTbK1jdvKFzkwtNEWy94iYiI7BGLRBtlD5s3jE8TjQdtt1YsspgkIiJ6fFgkktW1pVhs/DkRERFZHotEshrjsTjG/zYuFhvvfjYWhpp62396SkREZCtYJJJVGafNxdPn94/Jkcv5BJGIiMga+O5majeMTxQbGIQ/G58gij8nIiIiS2KRSO1K4yeKTW3OsYcNO0RERLaA083U7vEJIhER0ePHIpFsAp8gEhERPV4sEomIiIhIgkUiEREREUmwSCQiIiIiCRaJRERERCTBIpGIiIiIJFgk/h+dTocvvvgCY8aMQVBQEJ5//nls2bIFWq3W2l0jIiIieux4mPb/+eCDD/CnP/0J0dHRmDt3LnJzc7Fp0yYUFhZizZo11u4eERER0WPFIhHAtWvXsHfvXsTExGDjxo0AgNjYWHTu3BmJiYmYPn06QkJCrNxLIiIioseH080AMjIyYDAYMGPGDFF7fHw8AODIkSPW6BYRERGR1fBJIoC8vDzI5XIEBQWJ2j09PeHp6Ync3FyTrymXy8zVvSZ1f6KD2a/ZQekCF2cZFM5OZr+2peghg5OTDAaDtXvSfjg5WfbeszXMhxjzIcZ83MdciLWHfDyOf9taqldYJAIoLi6Gh4cHFAqF5LMePXqgqKjI5Gt6eLiZo2vNmjOR09/UtK5dO1m7C+0K8yHGfIgxH/cxF2LMB6ebAQA1NTVNFogAoFQqUVtb+5h7RERERGRdLBLRUAhqNJomP6urq4NSqXzMPSIiIiKyLhaJALy8vFBeXt5koVhcXAxPT08r9IqIiIjIelgkAhg4cCD0ej3y8vJE7UVFRSgqKpJsaCEiIiKydywSAcTExEAmk2HPnj2i9t27dwMAxo8fb41uEREREVkNdzcDCAwMxNSpU7Fv3z4YDAZERkYiLy8P+/btw0svvYTQ0FBrd5GIiIjosZIZDDxhDgC0Wi2++uorpKWl4ZdffkGPHj0wadIkzJs3Dy4uLtbuHhEREdFjxSKRiIiIiCS4JpGIiIiIJFgkEhEREZEEN67YAJ1OJ6yX/Pe//w0vLy9MmjQJc+fOhbOz7f8VFhYWYvTo0U1+9uyzz+Krr74StR07dgxbt25Ffn4+3NzcMHLkSCxbtgxdu3aV/P7169eRkJCACxcuwGAwYPDgwVi+fDkCAwMtMhZT3bt3Dzt27EBeXh7y8vJQUlKCF154AZs2bWoy3lJjr6iowIYNG3D8+HFUVlbCz88Pb7zxBqKjo80+5uaYkotRo0bh5s2bknaFQtHku9ZtLRcA8MMPP+Dw4cM4e/YsCgsLUV9fD19fX8TGxmL69OmQycTvW7Xne8OUXDjCvXHjxg189tln+OGHH1BSUgKDwQBvb29ER0cjPj4enTqJXydnz/cGYFo+HOH+MCfbrzAcwAcffIA//elPiI6Oxty5c5Gbm4tNmzahsLAQa9assXb3zGb48OEYN26cqK1Hjx6inzMyMvDWW28hKCgIK1euRFlZGbZv347c3FykpaXB1dVViC0oKEBcXBxcXV0xf/58KJVKJCcnIy4uDqmpqfDz83ss42pJeXk5Nm/eDJVKhaCgIJw4caLZWEuNXaPRYNasWcjPz0d8fDyefPJJZGZmYvHixairq8PEiRMtmQKBKbkAAG9vbyxcuFDU5uTkJImzxVwAQGJiIk6fPo0xY8Zg2rRp0Gg0yMzMxKpVq3DlyhV88MEHQqy93xum5AKw/3ujqKgIZWVlGDt2LDw9PSGTyZCXl4ctW7bg2LFj2Lt3r/CqWXu/N0zNB2D/94dZGahdu3r1qqFfv36GxYsXi9o/+ugjQ0BAgCEnJ8dKPTOfn3/+2RAQEGBYt25di3EajcYQGRlpGDVqlKG6ulpo/+tf/2oICAgwJCYmiuLffPNNQ//+/Q3Xrl0T2oqLiw1DhgwxzJ0717yDeEh1dXWGX375Rfg5ICDA8Oabb0riLDn2PXv2GAICAgwpKSlCm06nM0ybNs3w9NNPG2pqah55nG3R1lwYDAbDyJEjDS+99FKbrmuLuTAYDIbz588bamtrRW06nc7w6quvGgICAgzXr183GAyOcW+0NRcGg2PcG8358ssvDQEBAYZjx44ZDAbHuDda8mA+DAbHvj8eBtcktnMZGRkwGAyYMWOGqD0+Ph4AcOTIEWt0y2Lq6upQU1PT5GfZ2dkoKSlBbGwsOnToILQPHz4cPj4+SE9PF9ru3buHrKwsREREICAgQGhXqVSIjo7GyZMnUVFRYbFxtJVCoWjTax8tOfb09HR07NgRkydPFtrkcjni4uJQXl6OU6dOPeIo26atuWhMq9Wiqqqq2c9tNRcAEBYWJnlvvFwuR1RUFICGaTDAMe6NtuaiMXu+N5rTu3dvAMDdu3cBOMa90ZIH89GYI94fD4NFYjuXl5cHuVwueTWgp6cnPD09m1xDYav27t2L0NBQDBo0CCNHjsTWrVuh1WqFz42vTWzqcPOQkBBcu3ZNeP/29evXUV9fj5CQEElsaGgodDod1Gq1hUZifpYau16vh1qtRv/+/UXTMY2/q73eY9evX8egQYMQFhaG8PBwvPvuu7h9+7Ykxt5yUVxcDADCejJHvjcezIWRo9wbtbW1uH37Nv7973/j+PHjSEhIgEKhwNChQwE43r3RWj6MHOX+MAeuSWzniouL4eHhIbnxgIb1ekVFRVbolXnJ5XJERERgzJgx6NmzJ8rKynDo0CF88sknUKvVwsYF4z8ID65TNLbpdDqUlpaiV69ercYCsKncWWrsd+7cQW1trc3lyd/fH7GxsfD19UVdXR1OnTqFAwcO4Pz58/j666/RpUsXAK3nDbCtXJSWlmLfvn3o3bs3wsLCADjuvdFULgDHujd2796Njz/+WPjZ19cX//3f/y08QXO0e6O1fACOdX+YA4vEdq6mpqbJAhEAlEolamtrH3OPzK9Xr17YtWuXqC02NhYLFizAn//8Z5w9exYRERHCNHRT+TBORRljWoo1tjU3rd0eWWrsxvunpeu2x3vs888/F/08YcIEBAYGYu3atdi5c6ewKN2ecqHRaLBw4UJUVlbik08+kfTfke6N5nIBONa9MW7cOAQFBeHu3bu4ePEisrOzUV1dLXzuaPdGa/kAHOv+MAdON7dzSqVSmA54UF1dnWSdjr2QyWSYN28eAAhrOYxjbSofdXV1ACDs1Gsp1tjWeFdfe2epsZty3fYuPj4eHTt2FK39sZdcaLVaLFq0CBcvXsTvf/97DBs2TPjM0e6NlnLRHHu9N3r37o1hw4Zh7NixePfdd/H6669j0aJFOH36NADHuzday0dz7PX+MAcWie2cl5cXysvLm7z5iouLTV7ob0uMUwTGxcFeXl4A7k8DNFZcXAwnJyeoVKo2xQKwqdxZauzu7u5QKpUtxjY1hdIeyeVy9OzZU7SY3B5yodPpsHTpUmRlZeG9997D1KlTRZ870r3RWi6aY6/3xoNeeOEFuLi4YP/+/QAc695oyoP5aI6j3B8Pg0ViOzdw4EDo9XphAbJRUVERioqKJBta7Mm//vUvAEC3bt0ANOQCAHJyciSxubm58Pf3Fx73BwQEwMXFBZcvX5bE5uTkwMnJCf3797dU183OUmOXy+UYMGAA1Gq15H9EjN8VHBxs1rFYilarxa1bt4T7BbD9XOj1eqxYsQJHjx7F22+/LTnlAHCce6MtuWiOPd4bTdFqtdDpdMJuXke5N5rzYD5ainOE++NhsEhs52JiYiCTybBnzx5R++7duwEA48ePt0a3zKq8vFzSptVq8dlnnwEARo4cCQAYOnQoVCoV0tLSROs7vv32WxQUFIhy4ebmhhEjRuDs2bPIz88X2ktKSnD06FFERkbCw8PDUkMyO0uOPSYmBtXV1aL/29br9UhJSYG7uzsiIyMtPDrTNHd00bZt21BTU4MRI0YIbbacC71ej5UrVyI9PR1LlizB7Nmzm4xzhHujrblwlHujtLS0yfa9e/dCr9cLO2od4d4A2p4PR7k/zElmMBgM1u4Etez999/Hvn37EB0djcjISOTl5WHfvn2YOHEi1q1bZ+3uPbIFCxaguroagwYNgpeXF8rKynDkyBHk5+cjLi4O//Vf/yXEpqenY+nSpQgKCsKUKVNQVlaGHTt2wMvLC19//bXoLLAbN24I54PNnDkTSqUSSUlJKC4uxr59+0RnX1lTUlKS8H+6n376Kfz9/RETEwMACA8PR3h4OADLjb2urg5Tp07Fjz/+iNdeew0+Pj7IzMzE6dOnsXbtWkyaNKld5WLnzp3Yv38/nnvuOXh7e0Oj0eC7777DyZMnERQUhKSkJLvIxbp167Bjxw4EBwc3+dRsyJAh6NOnDwD7vzfamgtHuTd+97vfoaysDBEREejVqxeqqqqQnZ2NEydOwNfXF/v27UPnzp0B2P+9YUo+HOX+MCcWiTZAq9UK727+5Zdf0KNHD0yaNAnz5s2Di4uLtbv3yNLS0nDo0CHcuHEDd+/ehVKpRL9+/TB16tQmX2X0l7/8Bdu2bZO8h7TxVIHR1atXkZCQgIsXLwrv3Vy2bBkGDBjwGEbWNs29SxRoKKDffPNN4WdLjb28vBwff/wxjh8/jqqqKuGdo8YC7XFpSy4uXLiAL7/8ElevXhXONvPx8cELL7yA119/vckF4raYixkzZiA7O7vZzx/8h8ee74225sJR7o2MjAx88803uHr1KsrLy+Hs7AwfHx88//zzmDVrluTdzfZ8bwBtz4ej3B/mxCKRiIiIiCS4JpGIiIiIJFgkEhEREZEEi0QiIiIikmCRSEREREQSLBKJiIiISIJFIhERERFJsEgkIiIiIgkWiURENmbGjBno16+ftbtBRHbO2dodICKyBYWFhRg9erSozcXFBd27d0d4eDjmz58PX19fK/WOiMj8+MYVIqI2MBaJv/rVrzBu3DgAQFVVFS5duoScnBx07NgRaWlp8PPzs3hfbt26hZqaGhalRGRRfJJIRGSCvn37it6nDQC///3vkZKSgi+++AIfffSRxfvQq1cvi38HERHXJBIRPaKXXnoJAPDDDz+I2n/66SesXLkSv/71rxEUFIRf//rXWLVqFW7fvi3EFBYWIjAwEPPmzWvy2kVFRejfvz/mzJkjtDW3JlGj0SAxMRG/+c1vEBoairCwMMycORPZ2dmiuNWrV6Nfv364ceOGqH3SpEno168ftm7dKmpPTExEv379cObMmTZkg4jsBYtEIiIzcXa+Pzlz6dIlvPTSSzh8+DAGDRqE+Ph4DBgwAHv37sXUqVNRUVEBAPD29saQIUPw3XffiYpHo/T0dOj1ekyYMKHF766rq8OsWbOwfv16KBQKTJ06FdHR0bh27RpmzpyJv/zlL0Ls008/DQA4d+6c0FZZWYmrV69K2gEgOzsbLi4uGDx4sGkJISKbxiKRiOgRff311wAgFFEajQZLliyBk5MTDh06hE2bNmHFihXYtm0bNm7ciJ9//hmbNm0Sfn/ChAnQarXIzMyUXPvw4cNwdXXFmDFjWuzDZ599hvPnz2Pp0qX4+uuv8d5772H16tU4cuQIvLy88P7776O2thYAEB4eDplMJnrCeP78eeh0Ojz99NO4dOkSNBoNAECn0+HChQsICQmBq6vroyWKiGwKi0QiIhPcuHEDmzdvxubNm7Fu3TrExsZi37596NOnD/7jP/4DAPDXv/4Vt27dwrx58ySbS6KjozFw4EAcOXJE1Obi4oL09HRR7D/+8Q+o1WqMHj0abm5uzfZJr9dj79698Pf3x9y5c0Wfde3aFbNnz0Z5ebkwXezu7g5/f39RkZidnY2OHTtizpw5qK2txeXLlwE0TKFXVVVh6NChD5EtIrJl3LhCRGSCf/7zn/jss89Ebd7e3vjTn/4ElUoFAMjJyQEA5OfnY/PmzZJr1NXVoaKiArdv30bXrl3h7u6O5557DidOnEBhYSG8vb0BNDxFBIDf/OY3rfbp7t276NWrV5PfV1BQAKChwB05ciSAhinnPXv24Mcff4Svry/OnTuHIUOG4Omnn4aLiwvOnTuHp556Sph6Nk5RE5HjYJFIRGSC0aNHY8uWLQCA0tJSJCcnY8uWLVi8eDF27doFZ2dn3LlzBwDwzTfftHitmpoa4c8TJkxAVlYWjhw5gnnz5sFgMODw4cPw8PDAs88+2+J1jOsbr169KqwrbO37hg4dij179uDcuXPw9PTE1atXsXjxYri6uiI4OBjZ2dn43e9+x/WIRA6MRSIR0UPq3r07Fi1ahNLSUqSmpmL37t2YPXs2OnXqBADYsWMHhg0b1qZrjRo1Cm5ubjh8+DDmzZuHixcv4ubNm3jllVdEG2KaYvy+CRMmICEhoU3f99RTT0Emk+HcuXPo2bOnsB4RaCggd+zYgerqaq5HJHJgXJNIRPSI3nrrLXTs2BHbtm1DVVUVQkJCAAB///vf23wNV1dXREVFIT8/H1evXm3zVDMA+Pr6ws3NDbm5udDpdG36vq5du8Lf3x/ff/89srOz4ebmhoEDBwJomFquq6tDSkoK7t27x/WIRA6KRSIR0SPq2rUrXnnlFdy5cwe7du3C888/j549e+LLL78UNoA0VltbK6xbbMxYEH7zzTfIzMzE//t//w+DBg1q9fudnZ0xffp0FBQU4JNPPmmyUMzJyRFNNwMNTwzLyspw8OBBPPXUU8ITy8GDB8PFxQVfffUVAK5HJHJUnG4mIjKD2bNnIzk5Gbt27cJrr72GTz/9FG+88QamTp2KyMhI+Pn5QafTobCwEN9//z0GDRokFGFGERERUKlUSE5ORn19PV555ZU2f/+iRYuQl5eHL774Av/7v/+Lp556Ck888QR++eUX5OXloaCgAN999x06dOgg/E54eDiSkpJw+/ZtUSHYoUMHBAcH4+LFi1yPSOTA+CSRiMgMunbtiunTp+POnTvYvXs3QkNDcfDgQcTFxeFf//oXkpOTcejQIdy8eRMTJ07EokWLJNeQy+UYN24c6uvrAbRtqtlIqVRi+/bteP/999GlSxdkZGRgz549+Pvf/w5fX1/88Y9/hIeHh+h3hg4dCplMJvz5wc8AcD0ikQOTGQwGg7U7QURERETtC58kEhEREZEEi0QiIiIikmCRSEREREQSLBKJiIiISIJFIhERERFJsEgkIiIiIgkWiUREREQkwSKRiIiIiCRYJBIRERGRBItEIiIiIpL4/9VFUqngJ3VdAAAAAElFTkSuQmCC",
      "text/plain": [
       "<Figure size 720x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Lenght of reviews\n",
    "plt.figure(figsize=(10,4))\n",
    "sns.histplot(data[\"Review\"].apply(lambda x : len(x)))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c82725dd",
   "metadata": {
    "papermill": {
     "duration": 0.016087,
     "end_time": "2023-02-20T21:21:21.159612",
     "exception": false,
     "start_time": "2023-02-20T21:21:21.143525",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "We'll also do some basic **data cleaning**."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "f0da2c09",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:21:21.195307Z",
     "iopub.status.busy": "2023-02-20T21:21:21.194221Z",
     "iopub.status.idle": "2023-02-20T21:21:24.536291Z",
     "shell.execute_reply": "2023-02-20T21:21:24.534930Z"
    },
    "papermill": {
     "duration": 3.362991,
     "end_time": "2023-02-20T21:21:24.539498",
     "exception": false,
     "start_time": "2023-02-20T21:21:21.176507",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Clean text function\n",
    "def clean_text(text):\n",
    "    text = emoji.demojize(text)                # Remove emojis\n",
    "    text = text.replace('_', ' ')              # Replace underscores with space\n",
    "    text = text.strip()                        # Remove extra spaces\n",
    "    return text\n",
    "\n",
    "# Applying cleaning function\n",
    "data[\"Review\"] = data[\"Review\"].apply(lambda x: clean_text(x))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9a3aa394",
   "metadata": {
    "papermill": {
     "duration": 0.015553,
     "end_time": "2023-02-20T21:21:24.571433",
     "exception": false,
     "start_time": "2023-02-20T21:21:24.555880",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "Let's take a look at the **target distribution**. We can see there is an **imbalance** with more ratings leaning towards the **extremes**. We going to treat this problem as a **multiclass classification** one. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "10f0c653",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:21:24.605759Z",
     "iopub.status.busy": "2023-02-20T21:21:24.605315Z",
     "iopub.status.idle": "2023-02-20T21:21:24.844548Z",
     "shell.execute_reply": "2023-02-20T21:21:24.843154Z"
    },
    "papermill": {
     "duration": 0.259909,
     "end_time": "2023-02-20T21:21:24.847314",
     "exception": false,
     "start_time": "2023-02-20T21:21:24.587405",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAwQAAAFQCAYAAAABe7kqAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8qNh9FAAAACXBIWXMAAAsTAAALEwEAmpwYAAAxkklEQVR4nO3de3iM577/8U8SySCoQ3Oow09bcQ6hDo1Gt1NpErod4xAEVaFraamiaKu6HVtUi5Vio5U6h2JtEu2FVtXhslouEoLsZdOVHpKIBBE5zvz+cGWWaUIzmmQSz/v1V3M/33nmO9O52uczc9/342SxWCwCAAAAYEjOjm4AAAAAgOMQCAAAAAADIxAAAAAABkYgAAAAAAyMQAAAAAAYGIEAAAAAMDACAQAAAGBglRzdAKS0tNsym7kdBAAAAEqHs7OTatVyL/IYgaAcMJstBAIAAAA4BFOGAAAAAAMjEAAAAAAGRiAAAAAADIxAAAAAABgYgQAAAAAwMAIBAAAAYGAEAgAAAMDACAQAAACAgREIAAAAAAPjTsUAAADQYzWqyM3EpWFFkJOdpxs375TY+fi3DgAAALmZKmnB2zsc3QaKYdb8QSV6PqYMAQAAAAZGIAAAAAAMjEAAAAAAGBiBAAAAADAwAgEAAABgYAQCAAAAwMAIBAAAAICBEQgAAAAAAyMQAAAAAAZGIAAAAAAMjEAAAAAAGBiBAAAAADAwAgEAAABgYAQCAAAAwMAIBAAAAICBEQgAAAAAAyMQAAAAAAZGIAAAAAAMzKGB4Ny5c1q0aJH69eun9u3by8/PTwMGDNCWLVtksVgK1R84cEADBw5U69at1alTJ82aNUvXr18v8tyXLl1SeHi42rVrp2eeeUZjx47VhQsXiqxNT0/X7NmzFRAQoNatW2vAgAGKiYkpsjY/P19r1qxRz5495evrqxdeeEERERHKy8t7+DcCAAAAcJBKjnzytWvX6tixY+rZs6eGDBminJwcxcTEaM6cOTp//rzmzp1rrY2OjtYbb7whX19fzZw5U6mpqVq/fr1iY2MVFRWlypUrW2uvXLmi0NBQVa5cWRMmTJDJZNKmTZsUGhqq7du3y8fHx1qbk5OjMWPGKCEhQWFhYXryyScVExOjyZMnKzs7W/369bPpee7cudqyZYuCgoIUHh6u2NhYLV++XImJiVqwYEGpv2cAAABASXKyFPVVfBn58ccf5evrK5PJZB0zm80aNWqUTp48qb1796px48bKzc1Vt27dZDKZtHfvXlWpUkWSdPjwYYWHh2v69OkaO3as9Ryvv/66Dhw4oN27d6tJkyaSpJSUFAUGBqp9+/ZavXq1tXbjxo2aO3eu5syZo2HDhll7CA0N1ZUrV/Ttt99aw8bFixfVt29fBQUFadmyZdZzLF68WGvXrlVUVJRat25t9/uQmpohs9lh/xoAAADk4VFdC97e4eg2UAyz5g9SSsotux7j7OykOnWqFX2sJJp6WO3atbMJA5Lk7OysXr16Sbo77UeSTp48qZSUFIWEhFjDgCR16dJFDRs21N69e61jt2/f1qFDh+Tv728NA5Lk4eGhoKAgHTlyROnp6dbxvXv3qmrVqho4cKBND6GhoUpLS9PRo0et49HR0bJYLBo5cqRNz2FhYZKkffv2PexbAQAAADhEuVxUnJycLEmqXbu2JCkuLk6S5OfnV6i2devWunjxonJyciTdDRG5ublFflPv5+en/Px8xcfHS7r7S0B8fLyaN28uNze3QrWSFBsbax2Li4uTs7OzfH19bWq9vLzk5eVlUwsAAABUBA5dQ1CUa9euadu2bapXr57atWsn6d8BwdPTs1C9p6en8vPzde3aNdWtW/cPayUpKSlJknTjxg1lZWUVq7agj1q1ahUKDwX199ba434/3wAAAABF8fCoXmLnKleBICcnR6+//rpu3bqljz/+2HrhfefOHUkq8kK8YMpRQc2Dan9/vqysrD88b0FNweOKqi2ov7fWHqwhAAAAjlaSF5gofY/MGoJ75eXladKkSTp16pTef/99Pffcc9ZjBRfnBdOC7pWdnS1J1oW/D6otGCtO7e/PW1BfVG1B/e/XQwAAAADlXbkIBPn5+XrzzTd16NAhvf322xo8eLDNcW9vb0n/njp0r+TkZLm4uMjDw6NYtdLdOf+SVLNmTZlMpgfW3judyNvbW2lpaUWGguTkZOt5AQAAgIrC4YHAbDZr+vTp2r9/v956661CO/hIUsuWLSVJZ86cKXQsNjZWjRs3tk7ladKkiVxdXXX27NlCtWfOnJGLi4uaN28u6e5uQi1atFB8fHyhi/yC52rVqpVNH2az2brIuUBSUpKSkpIKLTYGAAAAyjuHBgKz2ayZM2dq7969mjJlil5++eUi6zp27CgPDw9FRUXZzNM/fPiwrly5oj59+ljH3N3d1bVrV504cUIJCQnW8ZSUFO3fv18BAQGqVauWdTw4OFiZmZnauXOnTV+bN29WzZo1FRAQYFPr5OSkL774wqa/yMhISbLpAwAAAKgIHLqo+MMPP9Tu3bvVqlUreXt7a8+ePTbHn3nmGTVo0EBubm6aMWOG3nzzTQ0fPlyDBg1SamqqPvvsM/n4+GjEiBE2j5syZYqOHz+uMWPGaPTo0TKZTNq4caPy8/M1bdo0m9ohQ4Zo586dmj9/vhITE9WwYUPFxMTo9OnTWrhwoc19D5o1a6bBgwdr27ZtslgsCggIUFxcnLZt26b+/fsXuS0qAAAAUJ459E7FI0eO1MmTJ+97fOHChRowYID176+//lqrVq1SQkKC3N3d1a1bN02dOlV16tQp9NgLFy5oyZIlOnXqlCwWi9q2baupU6eqRYsWhWrT0tK0dOlSHTx4UBkZGfLx8dG4ceMUHBxcqDYvL0/r1q1TVFSUfvvtN3l6emrAgAEaP368XF1dH+p9YJchAADgaNypuOIo6TsVOzQQ4C4CAQAAcDQCQcVR0oHA4YuKAQAAADgOgQAAAAAwMAIBAAAAYGAEAgAAAMDACAQAAACAgREIAAAAAAMjEAAAAAAGRiAAAAAADIxAAAAAABgYgQAAAAAwMAIBAAAAYGAEAgAAAMDACAQAAACAgREIAAAAAAMjEAAAAAAGRiAAAAAADIxAAAAAABgYgQAAAAAwsEqObgD2q16jsiqbXB3dBoohKztXt25mOboNAACA+yIQVECVTa4Knb7J0W2gGDZ/OFy3RCAAAADlF1OGAAAAAAMjEAAAAAAGRiAAAAAADIxAAAAAABgYgQAAAAAwMAIBAAAAYGAEAgAAAMDACAQAAACAgREIAAAAAAMjEAAAAAAGRiAAAAAADIxAAAAAABgYgQAAAAAwMAIBAAAAYGAEAgAAAMDACAQAAACAgREIAAAAAAMjEAAAAAAGRiAAAAAADIxAAAAAABgYgQAAAAAwMAIBAAAAYGAEAgAAAMDACAQAAACAgVVydAO3b9/WZ599pri4OMXFxSklJUUvvviili9fXqi2e/fu+vnnnwuNu7m5KTY2ttD4pUuXtGTJEv3444+yWCxq27atpk2bpmbNmhWqTU9P10cffaSDBw/q1q1b8vHx0bhx4xQUFFSoNj8/X+vWrVNUVJR+/fVXeXt7a8CAAQoPD1elSg5/SwEAAIBic/jVa1pamlasWCEPDw/5+vrqm2++eWB9/fr19frrr9uMubi4FKq7cuWKQkNDVblyZU2YMEEmk0mbNm1SaGiotm/fLh8fH2ttTk6OxowZo4SEBIWFhenJJ59UTEyMJk+erOzsbPXr18/m3HPnztWWLVsUFBSk8PBwxcbGavny5UpMTNSCBQse/s0AAAAAypjDA4Gnp6e+++47eXl5SZKaNm36wPrHHntMffv2/cPzfvTRR8rMzNTmzZvVpEkTSVJQUJACAwO1ePFirV692lq7fft2nT9/XnPmzNGwYcMkSYMGDVJoaKgWLVqkwMBAVa5cWZJ08eJFbd26VcHBwVq2bJkkKSQkRNWrV9fatWs1dOhQtW7d2v43AgAAAHAAh68hcHNzs4aB4srLy1NGRsZ9j9++fVuHDh2Sv7+/NQxIkoeHh4KCgnTkyBGlp6dbx/fu3auqVatq4MCB1jFnZ2eFhoYqLS1NR48etY5HR0fLYrFo5MiRNs8ZFhYmSdq3b59drwUAAABwJIcHAntdunRJbdq0Ubt27dShQwfNmjVL169fL1STm5tb5Df1fn5+ys/PV3x8vCTJbDYrPj5ezZs3l5ubW6FaSTbrE+Li4uTs7CxfX1+bWi8vL3l5eRW5lgEAAAAorxw+ZcgejRs3VkhIiBo1aqTs7GwdPXpUX375pX744Qft2LFDNWrUkCQlJydLujsd6fcKxpKSkiRJN27cUFZWVrFqC85dq1atQuGhoP7eWgAAAKC8q1CB4N55/5L00ksvqVmzZlq4cKE+//xz62LjO3fuSFKRF+0FYwU1WVlZ9601mUw2NQWPK6q2oP7e2uKqU6ea3Y9BxeHhUd3RLQAAgEdMSV5fVKhAUJSwsDB98sknOnr0qDUQFFzI5+TkFKovGCtYJPyg2uzsbJvagvrMzMwie8nOzraezx6pqRkymy3FrucCs2JJSbnl6BYAAPhDXF9ULPZeXzg7O933S+gKt4bg95ydnfXEE0/YLBL29vaW9O+pQ/cqGCtYyFyzZk2ZTKYH1t47ncjb21tpaWlFBojk5GS7F0gDAAAAjlThA0FeXp5++eUX1alTxzrWpEkTubq66uzZs4Xqz5w5IxcXFzVv3lzS3UDRokULxcfHF7rIP3PmjCSpVatW1rGWLVvKbDYrLi7OpjYpKUlJSUmFFhsDAAAA5VmFCQT3/gJwr1WrVunOnTvq2rWrdczd3V1du3bViRMnlJCQYB1PSUnR/v37FRAQoFq1alnHg4ODlZmZqZ07d1rHzGazNm/erJo1ayogIMCm1snJSV988YVNH5GRkZKkPn36/JmXCQAAAJSpcrGGYOPGjbp586b178uXLysiIkKS1KFDB3Xo0EG7d+/Wzp079fzzz6t+/frKycnR999/ryNHjsjX17fQfQGmTJmi48ePa8yYMRo9erRMJpM2btyo/Px8TZs2zaZ2yJAh2rlzp+bPn6/ExEQ1bNhQMTExOn36tBYuXKgqVapYa5s1a6bBgwdr27ZtslgsCggIUFxcnLZt26b+/ftbtyoFAAAAKoJyEQjWr1+vn3/+2fp3QkKCPvnkE0nSxIkT1aFDB7Vq1UonTpxQdHS09b4DDRs21GuvvaZXXnnFZuGvJD399NPatGmTlixZooiICFksFrVt21bLli2zuVmZdHeh8Oeff66lS5fqyy+/VEZGhnx8fLRs2TIFBwcX6nf27NmqV6+eoqKidODAAXl6emrixIkaP358Sb81AAAAQKlyslgsxd/eBqXiYXYZCp2+qRQ7QknZ/OFwdhkCAFQIHh7VteDtHY5uA8Uwa/4gdhkCAAAAUDIIBAAAAICBEQgAAAAAAyMQAAAAAAZGIAAAAAAMjEAAAAAAGBiBAAAAADAwuwLB7t27deHChQfWXLp0Sbt37/4zPQEAAAAoI3YFghkzZujAgQMPrPn22281c+bMP9UUAAAAgLJR4lOG8vLy5OzMTCQAAACgIrD7yt3Jyem+x3Jzc3Xq1CnVqlXrTzUFAAAAoGxU+qOCHj162Py9YcMGffnll4XqzGaz0tLSlJ2drYEDB5ZchwAAAABKzR8GAovFYv1nJycnWSwWmzHriSpVko+Pj5599lm9+uqrJdslAAAAgFLxh4Hg0KFD1n9u1qyZRo0apYkTJ5ZqUwAAAADKxh8GgnsdPHhQNWrUKK1eAAAAAJQxuwJBvXr1SqsPAAAAAA5gVyCQpNTUVO3cuVNxcXG6deuW8vPzC9U4OTlpw4YNJdIgAAAAgNJjVyA4f/68xowZo5s3bxa5sLjAg7YmBQAAAFB+2BUIFi5cqBs3bmjixIkaMGCAvLy85OLiUlq9AQAAAChldgWC2NhY9erVi12GAAAAgEeEXXcqrly5MguLAQAAgEeIXYHgP/7jP3Tq1KnS6gUAAABAGbMrELz11ltKS0vTwoULlZWVVVo9AQAAACgjdq0heOONN+Tu7q7IyEhFRUXpySefVLVq1QrVse0oAAAAUDHYFQhOnjxp/efMzEydP3++yDq2HQUAAAAqBrsCwYULF0qrDwAAAAAOYNcaAgAAAACPFgIBAAAAYGB2TRlauXJlseqcnJz017/+9aEaAgAAAFB2SjQQODk5yWKxEAgAAACACsKuQBAZGVnkeEZGhs6fP68vvvhCzz77rEaMGFEizQEAAAAoXXYFgo4dO973WPfu3dW3b1/169dPvXr1+tONAQAAACh9JbqouEGDBurZs6fWrFlTkqcFAAAAUEpKfJehxx57TFeuXCnp0wIAAAAoBSUaCDIyMnTw4EHVrl27JE8LAAAAoJSUyC5DZrNZSUlJ+uabb5SWlqbx48eXSHMAAAAASleJbjtapUoVjRkzRq+//vqfagoAAABA2SiRbUednZ1VvXp1PfXUU3JzcyuRxgAAAACUvhLbdhQAAABAxfOnFhVnZmYqOTlZmZmZJdUPAAAAgDJk1y8EkpSTk6O1a9dq9+7d+te//mUdb9Cggfr166dXXnmFaUMAAABABWFXILhz545Gjx6ts2fPytXVVU2bNtXjjz+u1NRUXb58WStWrNA333yjyMhIValSpbR6BgAAAFBC7AoEq1ev1pkzZzRw4EBNmTJFderUsR67fv26li1bpqioKK1evVqTJ08u6V4BAAAAlDC71hDExMSoTZs2mj9/vk0YkKTatWtr7ty5atOmjWJiYkq0SQAAAAClw65A8Msvv+jZZ599YE3Hjh31yy+//KmmAAAAAJQNuwJBtWrVlJSU9MCapKQkubu7F/uct2/f1sqVKzVhwgR17txZTZs2feCNzQ4cOKCBAweqdevW6tSpk2bNmqXr168XWXvp0iWFh4erXbt2euaZZzR27FhduHChyNr09HTNnj1bAQEBat26tQYMGHDfXzry8/O1Zs0a9ezZU76+vnrhhRcUERGhvLy8Yr9uAAAAoDywKxC0b99e0dHR+uGHH4o8fvr0acXExNh1v4K0tDStWLFCcXFx8vX1fWBtdHS0/vrXv0qSZs6cqeHDh2v//v0aNWqUsrKybGqvXLmi0NBQnT9/XhMmTNDkyZOVmJio0NBQ/e///q9NbU5OjsaMGaMvv/xSffv21TvvvKPHHntMkydP1u7duwv1MXfuXC1dulQtW7bUe++9p+eee07Lly/X7Nmzi/26AQAAgPLArkXFEydO1JEjRxQWFqYXXnhBHTt2VJ06dZSamqp//OMfOnDggNzc3PSXv/yl2Of09PTUd999Jy8vL0lS06ZNi6zLzc3VggULVL9+fW3cuNG6i1GrVq0UHh6uTZs2aezYsdb6jz76SJmZmdq8ebOaNGkiSQoKClJgYKAWL16s1atXW2u3b9+u8+fPa86cORo2bJgkadCgQQoNDdWiRYsUGBioypUrS5IuXryorVu3Kjg4WMuWLZMkhYSEqHr16lq7dq2GDh2q1q1bF/v1AwAAAI5k1y8ETZs21dq1a1WvXj19/fXXmj9/vqZMmaJ58+bpq6++Ut26dbVmzRo1a9as2Od0c3OzhoEHOXnypFJSUhQSEmKzpWmXLl3UsGFD7d271zp2+/ZtHTp0SP7+/tYwIEkeHh4KCgrSkSNHlJ6ebh3fu3evqlatqoEDB1rHnJ2dFRoaqrS0NB09etQ6Hh0dLYvFopEjR9r0FxYWJknat29fsV87AAAA4Gh235isffv2+vrrr/Xjjz8qPj5eGRkZqlatmpo3b6527drJycmpNPpUXFycJMnPz6/QsdatWys6Olo5OTlyc3PTpUuXlJubW+Q39X5+foqKilJ8fLw6deoks9ms+Ph4tWzZstAN1QqeKzY2Vj169LD24ezsXGh6k5eXl7y8vBQbG1sirxcAAAAoC3YHAklycnJS+/bt1b59+5Lu576Sk5Ml3Z1i9Huenp7Kz8/XtWvXVLdu3T+slWRdHH3jxg1lZWUVq7agj1q1ahV5N2ZPT88/XHRdlDp1qtn9GFQcHh7VHd0CAAB4xJTk9YVdgeC3337TuXPn1KFDB9WoUaPQ8Rs3buiHH36Qr69vsaYB2ePOnTuSVOSFuMlksql5UG3BWEFNwWLkB5333gXLd+7cKbK2oP73i5uLIzU1Q2azpdj1XGBWLCkptxzdAgAAf4jri4rF3usLZ2en+34JbdcagoiICM2aNcu6wPb3qlSpolmzZmnVqlV2NVgcBRfnOTk5hY5lZ2dLkrWvB9UWjBWn9vfnLagvqragvuB8AAAAQEVgVyA4ceKEAgIC7vsNuZubmzp37qxjx46VSHP38vb2lvTvqUP3Sk5OlouLizw8PIpVK8n6C0bNmjVlMpkeWHvvdCJvb2+lpaUVGQqSk5NL/JcRAAAAoDTZFQiSkpJUv379B9bUq1evyIvrP6tly5aSpDNnzhQ6Fhsbq8aNG1uDSpMmTeTq6qqzZ88Wqj1z5oxcXFzUvHlzSXd3E2rRooXi4+MLXeQXPFerVq1s+jCbzdZFzgWSkpKUlJT0h/dSAAAAAMoTuwKByWSy2a6zKOnp6XJxcfkzPRWpY8eO8vDwUFRUlM08/cOHD+vKlSvq06ePdczd3V1du3bViRMnlJCQYB1PSUnR/v37FRAQoFq1alnHg4ODlZmZqZ07d1rHzGazNm/erJo1ayogIMCm1snJSV988YVNf5GRkZJk0wcAAABQ3tm1qLh58+Y6dOiQZsyYoapVqxY6npGRoUOHDlm/fS+ujRs36ubNm9a/L1++rIiICElShw4d1KFDB7m5uWnGjBl68803NXz4cA0aNEipqan67LPP5OPjoxEjRticc8qUKTp+/LjGjBmj0aNHy2QyaePGjcrPz9e0adNsaocMGaKdO3dq/vz5SkxMVMOGDRUTE6PTp09r4cKFNvc9aNasmQYPHqxt27bJYrEoICBAcXFx2rZtm/r371/ktqgAAABAeeVksViKvb3NV199pUmTJsnPz09vv/22zT7/Z8+e1bx58xQbG6ulS5cqODi42E10795dP//8c5HHJk6cqNdee83699dff61Vq1YpISFB7u7u6tatm6ZOnao6deoUeuyFCxe0ZMkSnTp1ShaLRW3bttXUqVPVokWLQrVpaWlaunSpDh48qIyMDPn4+GjcuHFFvo68vDytW7dOUVFR+u233+Tp6akBAwZo/PjxcnV1LfbrLvAwuwyFTt9k9/Og7G3+cDi7DAEAKgQPj+pa8PYOR7eBYpg1f1CJ7jJkVyCQpHnz5mnjxo1ycnKSu7u7PD09lZycrNu3b8tisWjEiBF655137GrQ6AgEjy4CAQCgoiAQVBwlHQjsvjHZO++8o06dOmnLli2Ki4vTTz/9pOrVq+v555/XsGHD1K1bN3tPCQAAAMBBHupOxT169FCPHj1KuhcAAAAAZcyuXYYAAAAAPFoIBAAAAICBEQgAAAAAAyMQAAAAAAZGIAAAAAAMjEAAAAAAGBiBAAAAADAwAgEAAABgYAQCAAAAwMAIBAAAAICBEQgAAAAAAyMQAAAAAAZGIAAAAAAMjEAAAAAAGFglRzcAAADKVs3qbnKtbHJ0GyiG3Kxspd/KcXQbeMQRCAAAMBjXyiZFh41xdBsohuDIzyQCAUoZU4YAAAAAAyMQAAAAAAZGIAAAAAAMjEAAAAAAGBiBAAAAADAwAgEAAABgYAQCAAAAwMC4DwHwiKj1mJsquXGjoYogLydbaTfYVxwAUD4QCIBHRCU3k3788BVHt4FiaDd9rSQCAQCgfGDKEAAAAGBg/EIAAI+wGo+ZZHJzc3QbKIbsnBzdvJHt6DYAGBCBAAAeYSY3N43+bJKj20AxfD7mE0kEAgBljylDAAAAgIERCAAAAAADIxAAAAAABkYgAAAAAAyMQAAAAAAYGIEAAAAAMDACAQAAAGBgBAIAAADAwAgEAAAAgIERCAAAAAADIxAAAAAABkYgAAAAAAyMQAAAAAAYGIEAAAAAMLBKjm7AHomJierRo0eRxzp37qx169bZjB04cECffvqpEhIS5O7urm7dumnq1KmqXbt2ocdfunRJS5Ys0Y8//iiLxaK2bdtq2rRpatasWaHa9PR0ffTRRzp48KBu3bolHx8fjRs3TkFBQSXzQgEAAIAyUqECQYEuXbqod+/eNmOenp42f0dHR+uNN96Qr6+vZs6cqdTUVK1fv16xsbGKiopS5cqVrbVXrlxRaGioKleurAkTJshkMmnTpk0KDQ3V9u3b5ePjY63NycnRmDFjlJCQoLCwMD355JOKiYnR5MmTlZ2drX79+pXqawcAAABKUoUMBI0aNVLfvn3vezw3N1cLFixQ/fr1tXHjRlWpUkWS1KpVK4WHh2vTpk0aO3astf6jjz5SZmamNm/erCZNmkiSgoKCFBgYqMWLF2v16tXW2u3bt+v8+fOaM2eOhg0bJkkaNGiQQkNDtWjRIgUGBtqEDQAAAKA8q7BrCLKzs3Xnzp0ij508eVIpKSkKCQmxhgHp7i8LDRs21N69e61jt2/f1qFDh+Tv728NA5Lk4eGhoKAgHTlyROnp6dbxvXv3qmrVqho4cKB1zNnZWaGhoUpLS9PRo0dL8FUCAAAApatCBoKtW7fKz89Pbdq0Ubdu3fTpp58qLy/PejwuLk6S5OfnV+ixrVu31sWLF5WTkyPp7tqB3NxctW7dulCtn5+f8vPzFR8fL0kym82Kj49X8+bN5ebmVqhWkmJjY0vmRQIAAABloEJNGXJ2dpa/v7969uypJ554QqmpqdqzZ48+/vhjxcfHa/ny5ZKk5ORkSYXXFRSM5efn69q1a6pbt+4f1kpSUlKSJOnGjRvKysoqVi0AAABQEVSoQFC3bl1t2LDBZiwkJEQTJ07UV199pRMnTsjf3986lej33+JLkslkkiRrzYNqC8YKarKysv7wvAU19qhTp5rdj0HF4eFR3dEtoBzic4Gi8LlAUfhcoCgl+bmoUIGgKE5OTho/frwOHDigo0ePyt/f33pxXjAt6F7Z2dmSZF34+6DagrHi1P7+vPZITc2Q2Wwpdj3/YahYUlJulcnz8LmoWPhcoCh8LlAUPhcoir2fC2dnp/t+CV0h1xD8Xr169STJuvjX29tb0r+nDt0rOTlZLi4u8vDwKFatJHl5eUmSatasKZPJ9MDaoqYTAQAAAOXVIxEIrl69KkmqU6eOJKlly5aSpDNnzhSqjY2NVePGja3Tfpo0aSJXV1edPXu2UO2ZM2fk4uKi5s2bS7q7hqFFixaKj48v9CtBwXO1atWqhF4VAAAAUPoqVCBIS0srNJaXl6eVK1dKkrp16yZJ6tixozw8PBQVFWUzp//w4cO6cuWK+vTpYx1zd3dX165ddeLECSUkJFjHU1JStH//fgUEBKhWrVrW8eDgYGVmZmrnzp3WMbPZrM2bN6tmzZoKCAgouRcMAAAAlLIKtYbg3XffVWZmptq0aSNvb2+lpqZq3759SkhIUGhoqHXrTzc3N82YMUNvvvmmhg8frkGDBik1NVWfffaZfHx8NGLECJvzTpkyRcePH9eYMWM0evRomUwmbdy4Ufn5+Zo2bZpN7ZAhQ7Rz507Nnz9fiYmJatiwoWJiYnT69GktXLjQ5r4HAAAAQHlXoQJBly5dtGfPHm3dulU3b96UyWRS06ZN9cEHH6hfv342tX369JGbm5tWrVqlBQsWyN3dXb169dLUqVMLXbQ//fTT2rRpk5YsWaKIiAhZLBa1bdtWy5Yts7lZmXR3YfHnn3+upUuX6ssvv1RGRoZ8fHy0bNkyBQcHl/ZbAAAAAJSoChUIQkJCFBISUuz6Xr16qVevXsWqbdasmdauXVus2lq1amnevHmaN29esXsBAAAAyqMKtYYAAAAAQMkiEAAAAAAGRiAAAAAADIxAAAAAABgYgQAAAAAwMAIBAAAAYGAEAgAAAMDACAQAAACAgREIAAAAAAMjEAAAAAAGRiAAAAAADIxAAAAAABgYgQAAAAAwMAIBAAAAYGAEAgAAAMDACAQAAACAgREIAAAAAAMjEAAAAAAGRiAAAAAADIxAAAAAABgYgQAAAAAwMAIBAAAAYGAEAgAAAMDACAQAAACAgREIAAAAAAMjEAAAAAAGRiAAAAAADIxAAAAAABgYgQAAAAAwMAIBAAAAYGAEAgAAAMDACAQAAACAgREIAAAAAAMjEAAAAAAGRiAAAAAADIxAAAAAABgYgQAAAAAwMAIBAAAAYGAEAgAAAMDACAQAAACAgREIAAAAAAMjEAAAAAAGRiAAAAAADIxAAAAAABgYgQAAAAAwMALBQ8jPz9eaNWvUs2dP+fr66oUXXlBERITy8vIc3RoAAABgl0qObqAimjt3rrZs2aKgoCCFh4crNjZWy5cvV2JiohYsWODo9gAAAIBiIxDY6eLFi9q6dauCg4O1bNkySVJISIiqV6+utWvXaujQoWrdurWDuwQAAACKhylDdoqOjpbFYtHIkSNtxsPCwiRJ+/btc0RbAAAAwEPhFwI7xcXFydnZWb6+vjbjXl5e8vLyUmxsrN3ndHZ2svsxj9dyt/sxcIyH+ff7sNxq1Cmz58KfU5afi8er1S6z58KfU5afiyqP89+LiqIsPxeP1axaZs+FP8fez8WD6p0sFovlzzZkJC+99JJSU1N17NixQscGDRqktLQ0HTx40AGdAQAAAPZjypCd7ty5Izc3tyKPmUwmZWVllXFHAAAAwMMjENjJZDIpJyenyGPZ2dkymUxl3BEAAADw8AgEdvL29lZaWlqRoSA5OVleXl4O6AoAAAB4OAQCO7Vs2VJms1lxcXE240lJSUpKSiq02BgAAAAozwgEdgoODpaTk5O++OILm/HIyEhJUp8+fRzRFgAAAPBQ2HbUTs2aNdPgwYO1bds2WSwWBQQEKC4uTtu2bVP//v3l5+fn6BYBAACAYmPb0YeQl5endevWKSoqSr/99ps8PT01YMAAjR8/Xq6uro5uDwAAACg2AgEAAABgYKwhAAAAAAyMQAAAAAAYGIuK4TC3b9/WZ599pri4OMXFxSklJUUvvviili9f7ujW4CDnzp3T//zP/+jEiRNKTExUbm6uGjVqpJCQEA0dOlROTk6ObhEOcPnyZa1cuVLnzp1TSkqKLBaL6tevr6CgIIWFhalatWqObhHlxD//+U/17dtXubm5WrVqlbp16+bollDGEhMT1aNHjyKPde7cWevWrSvjjioGAgEcJi0tTStWrJCHh4d8fX31zTffOLolONjatWt17Ngx9ezZU0OGDFFOTo5iYmI0Z84cnT9/XnPnznV0i3CApKQkpaamKjAwUF5eXnJyclJcXJwiIiJ04MABbd26VW5ubo5uEw5msVg0e/Zsubq6Kjc319HtwMG6dOmi3r1724x5eno6qJvyj0AAh/H09NR3331nvbtz06ZNHdwRHG3EiBFatGiRTCaTdWzkyJEaNWqUtm/frrCwMDVu3NiBHcIROnXqpE6dOhUaf+qpp7R48WIdOXLkvt8Iwjh27NihuLg4jRs3TitWrHB0O3CwRo0aqW/fvo5uo8JgDQEcxs3NzRoGAElq166dTRiQJGdnZ/Xq1UuSdOnSJUe0hXKqXr16kqSbN286uBM42vXr17VkyRK9+uqrqlu3rqPbQTmRnZ2tO3fuOLqNCoFAAKDcS05OliTVrl3bwZ3AkbKysnT9+nX9+uuvOnjwoJYsWSI3Nzd17NjR0a3BwT744APVrFlTL7/8sqNbQTmxdetW+fn5qU2bNurWrZs+/fRT5eXlObqtcospQwDKtWvXrmnbtm2qV6+e2rVr5+h24ECRkZFaunSp9e9GjRrpb3/7m/WXAhjTiRMntHv3bq1fv561JJCzs7P8/f3Vs2dPPfHEE0pNTdWePXv08ccfKz4+no1L7oNAAKDcysnJ0euvv65bt27p448/5n/2Bte7d2/5+vrq5s2bOnXqlE6ePKnMzExHtwUHysnJ0Zw5cxQUFKSAgABHt4NyoG7dutqwYYPNWEhIiCZOnKivvvpKJ06ckL+/v4O6K7+YMgSgXMrLy9OkSZN06tQpvf/++3ruuecc3RIcrF69enruuecUGBioWbNm6ZVXXtGkSZN07NgxR7cGB1mzZo2SkpI0c+ZMR7eCcszJyUnjx4+XJB09etTB3ZRPBAIA5U5+fr7efPNNHTp0SG+//bYGDx7s6JZQDr344otydXXVzp07Hd0KHCA5OVmrV6/WoEGDlJWVpatXr+rq1atKTU2VJKWkpOjq1avMG4ekf29CkJ6e7thGyimmDAEoV8xms6ZPn679+/frrbfe0siRIx3dEsqpvLw85efns8uQQaWmpionJ0eRkZGKjIwsdPzdd9+VJB08eFD169cv6/ZQzly9elWSVKdOHQd3Uj4RCACUG2azWTNnztTevXs1ZcoUdgyBpLsLyx9//PFC41u3bpXZbJafn58DuoKj1a9fX5988kmh8ZMnT2rTpk0KDw9Xy5YtuQA0mLS0NNWqVctmLC8vTytXrpQk7l59HwQCONTGjRttvt27fPmyIiIiJEkdOnRQhw4dHNUaHODDDz/U7t271apVK3l7e2vPnj02x5955hk1aNDAQd3BUd577z2lpqbK399fdevWVUZGhk6ePKlvvvlGjRo10qhRoxzdIhygevXqCgwMLDResND8mWee4eLPgN59911lZmaqTZs28vb2Vmpqqvbt26eEhASFhobyBcJ9EAjgUOvXr9fPP/9s/TshIcH6jc/EiRMJBAZz7tw5SVJsbKymT59e6PjChQsJBAbUu3dv7dq1Szt37lRaWpoqVaqkhg0bauLEiRozZoyqVavm6BYBlBNdunTRnj17tHXrVt28eVMmk0lNmzbVBx98oH79+jm6vXLLyWKxWBzdBAAAAADHYJchAAAAwMAIBAAAAICBEQgAAAAAAyMQAAAAAAZGIAAAAAAMjEAAAAAAGBiBAAAAADAwAgEAAABgYAQCAECFlZiYqKZNm2rGjBmObgUAKqxKjm4AAFCxJSYmqkePHjZjrq6uevzxx9WhQwdNmDBBjRo1eujzz5gxQ7t27dLBgwdVv379P9suAOB3CAQAgBLx1FNPqXfv3pKkjIwMnT59Wn//+9914MABRUVFycfHp8Sf08vLS9HR0apevXqJnxsAjIJAAAAoEU8//bRee+01m7H3339fmzdv1po1a/Thhx+W+HO6urr+qV8fAACsIQAAlKL+/ftLks6dO2cdu3XrltasWaPQ0FAFBATI19dXXbt21ezZs5WSkmLz+O7du2vXrl2SpB49eqhp06Y2awbut4age/fu6t69uzIyMjRv3jx17txZrVq10sCBA3Xs2LEiez1+/LiGDh0qPz8/derUSTNnzlRaWpr1XADwqOIXAgBAqatU6d//u/nnP/+p5cuXy9/fXy+++KJMJpMuXryobdu26fvvv9euXbv02GOPSZLCwsK0a9cuXbhwQWFhYapRo4YkqXnz5n/4nLm5uRo7dqwyMjIUGBiomzdvat++fQoPD9eOHTvUrFkza+3hw4f16quvymQyqU+fPqpZs6aOHDmil19+Wbm5uXJ1dS3hdwQAyg8CAQCg1OzYsUOS1LZtW+tYo0aN9P3336tmzZo2tXv27NH06dO1adMm/eUvf5EkjR49WhcuXNCFCxc0atQouxYVJycnq02bNtq4caP1gr5Tp06aMWOGNm/erP/6r/+SJOXn5+u9996Tk5OTNm/ebA0bU6ZMUXh4uM6fP6969eo99HsAAOUdU4YAACXi8uXLWrFihVasWKFFixYpJCRE27ZtU4MGDfTqq69a66pXr14oDEjSf/7nf6p69eo6fvx4ifX01ltv2Xy7/9JLL6lSpUqKi4uzjv3www/69ddfFRgYaPPLg4uLiyZNmlRivQBAecUvBACAEvF///d/Wrlypc1Y/fr1tWXLFnl4eNiMHz9+XBs2bNDZs2eVnp6u/Px867Hk5OQS6adGjRqFflGoVKmS6tSpo5s3b1rHLly4IMn2V4wCvr6+TBcC8MgjEAAASkSPHj0UEREhSbp27Zo2bdqkiIgITZ48WRs2bLCuI4iOjtaUKVPk7u6u559/XvXq1VPlypUlSRs2bFBubm6J9HO/rUgrVaoks9ls/fv27duSpNq1axeqdXZ2LvLXDAB4lBAIAAAl7vHHH9ekSZN07do1bd++XZGRkXr55ZclSX/7299UuXJl7dq1S//v//0/62MsFovWrl1b5r26u7tLkq5fv17omNlsVnp6ujw9Pcu6LQAoM6whAACUmjfeeENVq1bVqlWrlJGRIUn66aef9PTTT9uEAUk6f/68srKyCp3D2fnu/6ru/Va/JBXsNnT69OlCx+Li4krsFwsAKK8IBACAUlO7dm0NHz5cN27c0IYNGyRJdevW1dWrV5Wammqty8jI0Pz584s8R8EWpElJSaXSY/v27eXt7a39+/db1xNId3cf+uSTT0rlOQGgPCEQAABK1csvv6yqVatqw4YNysjI0PDhw5WRkaH+/ftr3rx5mj17tl566SXl5+cXOTXH399fkvTee+/pk08+0Zo1a/Ttt9+WWH8uLi6aM2eOLBaLhg0bpnfeeUdLlixR//79df36dXl6esrJyanEng8AyhsCAQCgVNWuXVtDhw7VjRs3FBkZqZEjR2r27NmqVq2atm/frsOHD6tHjx5at25dkTv6dOnSRVOmTFFOTo7WrFmjpUuXav/+/SXaY7du3fTf//3fatKkif7+979rx44datmypdavX6/bt29b1xkAwKPIyWKxWBzdBAAA5dFPP/2knj17KjAwkOlDAB5Z/EIAADC827dvW7cfLZCTk6NFixZJurulKgA8qth2FABgeJcvX9aoUaPUuXNn1a9fXzdv3tSJEyf0r3/9Sx07dlTv3r0d3SIAlBqmDAEADO/atWtavHix/vGPfyg1NVVms1kNGjRQUFCQxo0bZ71xGgA8iggEAAAAgIGxhgAAAAAwMAIBAAAAYGAEAgAAAMDACAQAAACAgREIAAAAAAP7/y6XEaTdJiSkAAAAAElFTkSuQmCC",
      "text/plain": [
       "<Figure size 864x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Lenght of reviews\n",
    "plt.figure(figsize=(12,5))\n",
    "sns.countplot(data=data, x=\"Rating\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c8dd5659",
   "metadata": {
    "papermill": {
     "duration": 0.016929,
     "end_time": "2023-02-20T21:21:24.880718",
     "exception": false,
     "start_time": "2023-02-20T21:21:24.863789",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "**Features and labels**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "9ecf58b1",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:21:24.917787Z",
     "iopub.status.busy": "2023-02-20T21:21:24.917292Z",
     "iopub.status.idle": "2023-02-20T21:21:24.924001Z",
     "shell.execute_reply": "2023-02-20T21:21:24.922716Z"
    },
    "papermill": {
     "duration": 0.028288,
     "end_time": "2023-02-20T21:21:24.926574",
     "exception": false,
     "start_time": "2023-02-20T21:21:24.898286",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Define features and labels\n",
    "X = data[\"Review\"]\n",
    "y = data[\"Rating\"] - 1         # map to 0,1,2,3,4"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e4815769",
   "metadata": {
    "papermill": {
     "duration": 0.016091,
     "end_time": "2023-02-20T21:21:24.959146",
     "exception": false,
     "start_time": "2023-02-20T21:21:24.943055",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "**Train-valid-test split**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "81bfe91b",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:21:24.994537Z",
     "iopub.status.busy": "2023-02-20T21:21:24.994047Z",
     "iopub.status.idle": "2023-02-20T21:21:25.066155Z",
     "shell.execute_reply": "2023-02-20T21:21:25.063856Z"
    },
    "papermill": {
     "duration": 0.093515,
     "end_time": "2023-02-20T21:21:25.069095",
     "exception": false,
     "start_time": "2023-02-20T21:21:24.975580",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train size: (36956,)\n",
      "Valid size: (12319,)\n",
      "Test size: (12319,)\n"
     ]
    }
   ],
   "source": [
    "# Create train, validation and test sets\n",
    "X_tr, X_test, y_tr, y_test = train_test_split(X, y, test_size=0.2, shuffle=True, stratify=y, random_state=0)\n",
    "X_train, X_valid, y_train, y_valid = train_test_split(X_tr, y_tr, test_size=0.25, shuffle=True, stratify=y_tr, random_state=0) # 0.25 * 0.8 = 0.2\n",
    "\n",
    "# Print shapes\n",
    "print(\"Train size:\", X_train.shape)\n",
    "print(\"Valid size:\", X_valid.shape)\n",
    "print(\"Test size:\", X_test.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "375cc970",
   "metadata": {
    "papermill": {
     "duration": 0.015428,
     "end_time": "2023-02-20T21:21:25.100799",
     "exception": false,
     "start_time": "2023-02-20T21:21:25.085371",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "**Tokenizer**\n",
    "\n",
    "<br>\n",
    "\n",
    "Instead of using spacy, this time we will create a **tokenizer using keras** because it will more easily integrate with our keras model. We set it to **remove special characters**, **lower case** all words and only use the **most frequent 20000 words**. This allows the model to focus only on the most frequent words that are **relevant to the sentiment** and speeds up training time. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "f44c5a6c",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:21:25.133977Z",
     "iopub.status.busy": "2023-02-20T21:21:25.133311Z",
     "iopub.status.idle": "2023-02-20T21:21:26.468196Z",
     "shell.execute_reply": "2023-02-20T21:21:26.467055Z"
    },
    "papermill": {
     "duration": 1.354861,
     "end_time": "2023-02-20T21:21:26.471257",
     "exception": false,
     "start_time": "2023-02-20T21:21:25.116396",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Keras tokenizer\n",
    "tokenizer = keras.preprocessing.text.Tokenizer(num_words=20000, filters='0123456789!\"#$%&()*+,-./:;<=>?@[\\]^_`{|}~', lower=True)\n",
    "\n",
    "# Build vocabulary on train set\n",
    "tokenizer.fit_on_texts(X_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "784eeec7",
   "metadata": {
    "papermill": {
     "duration": 0.016877,
     "end_time": "2023-02-20T21:21:26.504714",
     "exception": false,
     "start_time": "2023-02-20T21:21:26.487837",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "**Vectorize**\n",
    "\n",
    "We are going to vectorize by mapping every token to an **integer ID** that we can use to look up word vectors."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "6330775c",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:21:26.540411Z",
     "iopub.status.busy": "2023-02-20T21:21:26.539991Z",
     "iopub.status.idle": "2023-02-20T21:21:28.538066Z",
     "shell.execute_reply": "2023-02-20T21:21:28.536815Z"
    },
    "papermill": {
     "duration": 2.019157,
     "end_time": "2023-02-20T21:21:28.540962",
     "exception": false,
     "start_time": "2023-02-20T21:21:26.521805",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1, 116, 8, 29, 28, 944, 25, 10, 211, 22]\n"
     ]
    }
   ],
   "source": [
    "# Vectorize\n",
    "X_train_ids = tokenizer.texts_to_sequences(X_train)\n",
    "X_valid_ids = tokenizer.texts_to_sequences(X_valid)\n",
    "X_test_ids = tokenizer.texts_to_sequences(X_test)\n",
    "\n",
    "# Example\n",
    "print(X_train_ids[0][:10])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cc2f0df3",
   "metadata": {
    "papermill": {
     "duration": 0.015793,
     "end_time": "2023-02-20T21:21:28.573004",
     "exception": false,
     "start_time": "2023-02-20T21:21:28.557211",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "Padding is a common technique used to convert all the sequences to the **same length**. Longer sequences are truncated and shorter sequences are padded with 0's at the beginning. This is required for some models (not here) but it can also make it **easier to optimize** the loss function. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "7f6862a4",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:21:28.608001Z",
     "iopub.status.busy": "2023-02-20T21:21:28.607490Z",
     "iopub.status.idle": "2023-02-20T21:21:28.892419Z",
     "shell.execute_reply": "2023-02-20T21:21:28.891232Z"
    },
    "papermill": {
     "duration": 0.305981,
     "end_time": "2023-02-20T21:21:28.895321",
     "exception": false,
     "start_time": "2023-02-20T21:21:28.589340",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[  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   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   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   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   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   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   1 116   8  29  28 944  25\n",
      "  10 211  22   5 265  74 187  22   8  27 108  89  28 944 226   5   5  98\n",
      " 273 187  22  45 944 294  30  80  39 205  40  75   5  89  75   5]\n"
     ]
    }
   ],
   "source": [
    "# Max length of sequences\n",
    "max_length = 250\n",
    "\n",
    "# Pad sequences\n",
    "X_train_pad = keras.preprocessing.sequence.pad_sequences(X_train_ids, maxlen=max_length)\n",
    "X_valid_pad = keras.preprocessing.sequence.pad_sequences(X_valid_ids, maxlen=max_length)\n",
    "X_test_pad = keras.preprocessing.sequence.pad_sequences(X_test_ids, maxlen=max_length)\n",
    "\n",
    "# Example\n",
    "print(X_train_pad[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "351dadcf",
   "metadata": {
    "papermill": {
     "duration": 0.016831,
     "end_time": "2023-02-20T21:21:28.928617",
     "exception": false,
     "start_time": "2023-02-20T21:21:28.911786",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "**Lookup dictionary**\n",
    "\n",
    "<br>\n",
    "\n",
    "We now to **map** the word ids to the word vectors in gensim so we can use them in our model. We do this by creating an **embedding matrix** that stores the word vectors in the right order if they exist (otherwise fill that row with 0's)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "1f9708c1",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:21:28.967392Z",
     "iopub.status.busy": "2023-02-20T21:21:28.966969Z",
     "iopub.status.idle": "2023-02-20T21:21:51.509252Z",
     "shell.execute_reply": "2023-02-20T21:21:51.507615Z"
    },
    "papermill": {
     "duration": 22.56566,
     "end_time": "2023-02-20T21:21:51.512466",
     "exception": false,
     "start_time": "2023-02-20T21:21:28.946806",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Load 1,000,000 most common words\n",
    "word2vec_model = gensim.models.KeyedVectors.load_word2vec_format(word2vec_path, binary=True, limit=1_000_000)\n",
    "\n",
    "# Add one for padding token\n",
    "num_tokens = len(tokenizer.word_index) + 1\n",
    "\n",
    "# Word vector dimension\n",
    "embedding_dim = 300\n",
    "\n",
    "# Initialize matrix of zeros\n",
    "embedding_matrix = np.zeros((num_tokens, embedding_dim))\n",
    "\n",
    "# Fill embedding matrix\n",
    "for word, i in tokenizer.word_index.items():\n",
    "    if word2vec_model.has_index_for(word):\n",
    "        embedding_matrix[i] = word2vec_model[word].copy()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "af97ad1f",
   "metadata": {
    "papermill": {
     "duration": 0.015453,
     "end_time": "2023-02-20T21:21:51.544367",
     "exception": false,
     "start_time": "2023-02-20T21:21:51.528914",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "Note that there are **many words** which we **could not find word vectors for**. This is likely because they were trained on a news corpus whereas our dataset is a corpus of app reviews and they each use quite a different vocabulary. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "5a3e29e5",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:21:51.578186Z",
     "iopub.status.busy": "2023-02-20T21:21:51.577716Z",
     "iopub.status.idle": "2023-02-20T21:21:51.598660Z",
     "shell.execute_reply": "2023-02-20T21:21:51.596731Z"
    },
    "papermill": {
     "duration": 0.042215,
     "end_time": "2023-02-20T21:21:51.602274",
     "exception": false,
     "start_time": "2023-02-20T21:21:51.560059",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Embedding matrix shape: (18453, 300)\n",
      "Number of words without word vectors 5563\n"
     ]
    }
   ],
   "source": [
    "print(\"Embedding matrix shape:\", embedding_matrix.shape)\n",
    "print(\"Number of words without word vectors\", int((embedding_matrix==0).sum()/300))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8f32262c",
   "metadata": {
    "papermill": {
     "duration": 0.015441,
     "end_time": "2023-02-20T21:21:51.634056",
     "exception": false,
     "start_time": "2023-02-20T21:21:51.618615",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "**Build model**\n",
    "\n",
    "We're now ready to build the model. The first layer of the neural network will be an **embedding layer** with our embedding matrix as input. We'll set `trainable = True` so the model can **fine tune** these word vectors. This is a good idea since we found out many of the vectors are 0's. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "4278e942",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:21:51.668651Z",
     "iopub.status.busy": "2023-02-20T21:21:51.667406Z",
     "iopub.status.idle": "2023-02-20T21:21:51.723562Z",
     "shell.execute_reply": "2023-02-20T21:21:51.722107Z"
    },
    "papermill": {
     "duration": 0.076781,
     "end_time": "2023-02-20T21:21:51.726530",
     "exception": false,
     "start_time": "2023-02-20T21:21:51.649749",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Embedding layer with pre-trained word vectors\n",
    "embedding_layer = layers.Embedding(\n",
    "    input_dim = num_tokens,\n",
    "    output_dim = embedding_dim,\n",
    "    embeddings_initializer = keras.initializers.Constant(embedding_matrix),\n",
    "    input_length = max_length,\n",
    "    trainable = True\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "de5d91cd",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:21:51.761332Z",
     "iopub.status.busy": "2023-02-20T21:21:51.760762Z",
     "iopub.status.idle": "2023-02-20T21:21:51.769574Z",
     "shell.execute_reply": "2023-02-20T21:21:51.768191Z"
    },
    "papermill": {
     "duration": 0.028571,
     "end_time": "2023-02-20T21:21:51.771829",
     "exception": false,
     "start_time": "2023-02-20T21:21:51.743258",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Build model\n",
    "def build_model(embedding_layer):\n",
    "    model = keras.Sequential([\n",
    "\n",
    "        # Look up word vectors from integer ids\n",
    "        embedding_layer,\n",
    "        \n",
    "        # Average all word vectors in sequence\n",
    "        layers.GlobalAveragePooling1D(),\n",
    "        \n",
    "        # Fully connected layers\n",
    "        layers.Dense(units=256, activation='relu', kernel_initializer=tf.keras.initializers.random_normal()),\n",
    "        layers.Dense(units=128, activation='relu', kernel_initializer=tf.keras.initializers.random_normal()),\n",
    "        \n",
    "        # Output layer (softmax returns a probability distribution)\n",
    "        layers.Dense(units=5, activation='softmax', kernel_initializer=tf.keras.initializers.random_normal())\n",
    "    ])\n",
    "    \n",
    "    # Define optimizer, loss function and accuracy metric\n",
    "    model.compile(optimizer='adam',\n",
    "                  loss='sparse_categorical_crossentropy',\n",
    "                  metrics=['sparse_categorical_accuracy'])\n",
    "    \n",
    "    return model"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "94bdb20a",
   "metadata": {
    "papermill": {
     "duration": 0.015489,
     "end_time": "2023-02-20T21:21:51.803329",
     "exception": false,
     "start_time": "2023-02-20T21:21:51.787840",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "Notice how most of the **model parameters** come from the embedding layer."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "22bcd764",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:21:51.837581Z",
     "iopub.status.busy": "2023-02-20T21:21:51.837051Z",
     "iopub.status.idle": "2023-02-20T21:21:52.070422Z",
     "shell.execute_reply": "2023-02-20T21:21:52.067575Z"
    },
    "papermill": {
     "duration": 0.255044,
     "end_time": "2023-02-20T21:21:52.074169",
     "exception": false,
     "start_time": "2023-02-20T21:21:51.819125",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding (Embedding)        (None, 250, 300)          5535900   \n",
      "_________________________________________________________________\n",
      "global_average_pooling1d (Gl (None, 300)               0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 256)               77056     \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 128)               32896     \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 5)                 645       \n",
      "=================================================================\n",
      "Total params: 5,646,497\n",
      "Trainable params: 5,646,497\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-02-20 21:21:51.886620: I tensorflow/core/common_runtime/process_util.cc:146] Creating new thread pool with default inter op setting: 2. Tune using inter_op_parallelism_threads for best performance.\n"
     ]
    }
   ],
   "source": [
    "# Build model\n",
    "model = build_model(embedding_layer)\n",
    "\n",
    "# Model summary\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0ec81503",
   "metadata": {
    "papermill": {
     "duration": 0.016271,
     "end_time": "2023-02-20T21:21:52.107279",
     "exception": false,
     "start_time": "2023-02-20T21:21:52.091008",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "**Train model**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "13dbee66",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:21:52.141752Z",
     "iopub.status.busy": "2023-02-20T21:21:52.141282Z",
     "iopub.status.idle": "2023-02-20T21:25:12.396059Z",
     "shell.execute_reply": "2023-02-20T21:25:12.394296Z"
    },
    "papermill": {
     "duration": 200.275513,
     "end_time": "2023-02-20T21:25:12.399342",
     "exception": false,
     "start_time": "2023-02-20T21:21:52.123829",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-02-20 21:21:52.290756: I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:185] None of the MLIR Optimization Passes are enabled (registered 2)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "289/289 [==============================] - 21s 70ms/step - loss: 1.2343 - sparse_categorical_accuracy: 0.5198 - val_loss: 1.0642 - val_sparse_categorical_accuracy: 0.5894\n",
      "Epoch 2/10\n",
      "289/289 [==============================] - 20s 70ms/step - loss: 0.9959 - sparse_categorical_accuracy: 0.6153 - val_loss: 0.9926 - val_sparse_categorical_accuracy: 0.6160\n",
      "Epoch 3/10\n",
      "289/289 [==============================] - 20s 68ms/step - loss: 0.9226 - sparse_categorical_accuracy: 0.6420 - val_loss: 0.9635 - val_sparse_categorical_accuracy: 0.6246\n",
      "Epoch 4/10\n",
      "289/289 [==============================] - 20s 69ms/step - loss: 0.8893 - sparse_categorical_accuracy: 0.6544 - val_loss: 0.9723 - val_sparse_categorical_accuracy: 0.6232\n",
      "Epoch 5/10\n",
      "289/289 [==============================] - 20s 68ms/step - loss: 0.8596 - sparse_categorical_accuracy: 0.6646 - val_loss: 0.9782 - val_sparse_categorical_accuracy: 0.6225\n",
      "Epoch 6/10\n",
      "289/289 [==============================] - 20s 69ms/step - loss: 0.8368 - sparse_categorical_accuracy: 0.6751 - val_loss: 0.9954 - val_sparse_categorical_accuracy: 0.6157\n",
      "Epoch 7/10\n",
      "289/289 [==============================] - 20s 69ms/step - loss: 0.8120 - sparse_categorical_accuracy: 0.6858 - val_loss: 1.0336 - val_sparse_categorical_accuracy: 0.5928\n",
      "Epoch 8/10\n",
      "289/289 [==============================] - 20s 68ms/step - loss: 0.7872 - sparse_categorical_accuracy: 0.6956 - val_loss: 1.0480 - val_sparse_categorical_accuracy: 0.6164\n",
      "Epoch 9/10\n",
      "289/289 [==============================] - 20s 69ms/step - loss: 0.7704 - sparse_categorical_accuracy: 0.7052 - val_loss: 1.0368 - val_sparse_categorical_accuracy: 0.5982\n",
      "Epoch 10/10\n",
      "289/289 [==============================] - 20s 69ms/step - loss: 0.7423 - sparse_categorical_accuracy: 0.7156 - val_loss: 1.0875 - val_sparse_categorical_accuracy: 0.6121\n"
     ]
    }
   ],
   "source": [
    "# Train model\n",
    "history = model.fit(\n",
    "    X_train_pad, y_train,\n",
    "    validation_data=(X_valid_pad, y_valid),\n",
    "    epochs=10,\n",
    "    batch_size=128,\n",
    "    verbose=True\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f37ae962",
   "metadata": {
    "papermill": {
     "duration": 0.20448,
     "end_time": "2023-02-20T21:25:12.809518",
     "exception": false,
     "start_time": "2023-02-20T21:25:12.605038",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "**Learning curves**\n",
    "\n",
    "<br>\n",
    "\n",
    "It is clear that the model begins to **overfit very quickly** because of the large number of **trainable parameters** in our model. If we set `trainable = False` then we would see less overfitting, although the accuracy wouldn't necessarily improve by much. In future notebooks, we will explore how to reach better performance with sequential neural networks."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "5ad7bae4",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:25:13.304174Z",
     "iopub.status.busy": "2023-02-20T21:25:13.303758Z",
     "iopub.status.idle": "2023-02-20T21:25:13.847660Z",
     "shell.execute_reply": "2023-02-20T21:25:13.846354Z"
    },
    "papermill": {
     "duration": 0.746432,
     "end_time": "2023-02-20T21:25:13.850119",
     "exception": false,
     "start_time": "2023-02-20T21:25:13.103687",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Final accuracy on validation set: 0.6120626926422119\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot learning curves\n",
    "history_df = pd.DataFrame(history.history)\n",
    "history_df.loc[:, ['loss', 'val_loss']].plot(title=\"Cross-entropy\")\n",
    "history_df.loc[:, ['sparse_categorical_accuracy', 'val_sparse_categorical_accuracy']].plot(title=\"Accuracy\")\n",
    "\n",
    "print('Final accuracy on validation set:', history_df.loc[len(history_df)-1,'val_sparse_categorical_accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "77c4104f",
   "metadata": {
    "papermill": {
     "duration": 0.204541,
     "end_time": "2023-02-20T21:25:14.256475",
     "exception": false,
     "start_time": "2023-02-20T21:25:14.051934",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "**Evaluate on test set**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "de3ae44f",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:25:14.676855Z",
     "iopub.status.busy": "2023-02-20T21:25:14.676352Z",
     "iopub.status.idle": "2023-02-20T21:25:16.427765Z",
     "shell.execute_reply": "2023-02-20T21:25:16.426421Z"
    },
    "papermill": {
     "duration": 1.96473,
     "end_time": "2023-02-20T21:25:16.430465",
     "exception": false,
     "start_time": "2023-02-20T21:25:14.465735",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test set accuracy: 0.6151\n"
     ]
    }
   ],
   "source": [
    "# Make predictions on test set\n",
    "preds = model.predict(X_test_pad)\n",
    "\n",
    "# Retrieve most likely class\n",
    "class_preds = np.argmax(preds,axis=1)\n",
    "\n",
    "# Print accuracy\n",
    "print(f\"Test set accuracy: {accuracy_score(y_test, class_preds):.4f}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "aa8faa20",
   "metadata": {
    "papermill": {
     "duration": 0.255709,
     "end_time": "2023-02-20T21:25:16.888964",
     "exception": false,
     "start_time": "2023-02-20T21:25:16.633255",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "Even though the accuracy is low, our **model did fairly well** considering there are **5 classes to predict**. In general it managed to work out if a review was positive or negative quite well."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "5ac06114",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:25:17.291800Z",
     "iopub.status.busy": "2023-02-20T21:25:17.291070Z",
     "iopub.status.idle": "2023-02-20T21:25:17.645902Z",
     "shell.execute_reply": "2023-02-20T21:25:17.644656Z"
    },
    "papermill": {
     "duration": 0.558106,
     "end_time": "2023-02-20T21:25:17.648859",
     "exception": false,
     "start_time": "2023-02-20T21:25:17.090753",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 576x432 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Confusion matrix\n",
    "cm = confusion_matrix(y_test, class_preds)\n",
    "plt.figure(figsize=(8,6))\n",
    "sns.heatmap(cm, annot=True, fmt='d', cmap='Blues')\n",
    "plt.xlabel('Predicted label')\n",
    "plt.ylabel('True label')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e8b73195",
   "metadata": {
    "papermill": {
     "duration": 0.203559,
     "end_time": "2023-02-20T21:25:18.057559",
     "exception": false,
     "start_time": "2023-02-20T21:25:17.854000",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "## <p style=\"font-family:JetBrains Mono; font-weight:normal; letter-spacing: 1px; color:#207d06; font-size:100%; text-align:left;padding: 0px; border-bottom: 3px solid #207d06;\">4.3 Modelling with word embeddings from scratch</p>\n",
    "\n",
    "Sometimes we might not want to use pretrained word embeddings and instead let our model figure out the **best word representations for our task** at hand. By default, the keras embedding layer does this **automatically**. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "40194922",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:25:18.464421Z",
     "iopub.status.busy": "2023-02-20T21:25:18.462972Z",
     "iopub.status.idle": "2023-02-20T21:25:18.471361Z",
     "shell.execute_reply": "2023-02-20T21:25:18.470480Z"
    },
    "papermill": {
     "duration": 0.216921,
     "end_time": "2023-02-20T21:25:18.473924",
     "exception": false,
     "start_time": "2023-02-20T21:25:18.257003",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Embedding layer to train from scratch\n",
    "embedding_layer = layers.Embedding(\n",
    "    input_dim = num_tokens,\n",
    "    output_dim = embedding_dim,\n",
    "    input_length = max_length\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "11c1b078",
   "metadata": {
    "papermill": {
     "duration": 0.198459,
     "end_time": "2023-02-20T21:25:18.871997",
     "exception": false,
     "start_time": "2023-02-20T21:25:18.673538",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "**Train model**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "9c3721c7",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:25:19.343735Z",
     "iopub.status.busy": "2023-02-20T21:25:19.343327Z",
     "iopub.status.idle": "2023-02-20T21:28:37.435910Z",
     "shell.execute_reply": "2023-02-20T21:28:37.434743Z"
    },
    "papermill": {
     "duration": 198.352786,
     "end_time": "2023-02-20T21:28:37.438552",
     "exception": false,
     "start_time": "2023-02-20T21:25:19.085766",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "289/289 [==============================] - 20s 69ms/step - loss: 1.3034 - sparse_categorical_accuracy: 0.4702 - val_loss: 1.1006 - val_sparse_categorical_accuracy: 0.5856\n",
      "Epoch 2/10\n",
      "289/289 [==============================] - 20s 69ms/step - loss: 1.0093 - sparse_categorical_accuracy: 0.6129 - val_loss: 0.9731 - val_sparse_categorical_accuracy: 0.6189\n",
      "Epoch 3/10\n",
      "289/289 [==============================] - 20s 69ms/step - loss: 0.9294 - sparse_categorical_accuracy: 0.6398 - val_loss: 0.9606 - val_sparse_categorical_accuracy: 0.6263\n",
      "Epoch 4/10\n",
      "289/289 [==============================] - 19s 67ms/step - loss: 0.8911 - sparse_categorical_accuracy: 0.6525 - val_loss: 0.9706 - val_sparse_categorical_accuracy: 0.6227\n",
      "Epoch 5/10\n",
      "289/289 [==============================] - 20s 69ms/step - loss: 0.8696 - sparse_categorical_accuracy: 0.6630 - val_loss: 0.9761 - val_sparse_categorical_accuracy: 0.6232\n",
      "Epoch 6/10\n",
      "289/289 [==============================] - 20s 68ms/step - loss: 0.8396 - sparse_categorical_accuracy: 0.6731 - val_loss: 1.0117 - val_sparse_categorical_accuracy: 0.6074\n",
      "Epoch 7/10\n",
      "289/289 [==============================] - 19s 67ms/step - loss: 0.8193 - sparse_categorical_accuracy: 0.6803 - val_loss: 1.0331 - val_sparse_categorical_accuracy: 0.6074\n",
      "Epoch 8/10\n",
      "289/289 [==============================] - 20s 68ms/step - loss: 0.7968 - sparse_categorical_accuracy: 0.6907 - val_loss: 1.0197 - val_sparse_categorical_accuracy: 0.6118\n",
      "Epoch 9/10\n",
      "289/289 [==============================] - 20s 69ms/step - loss: 0.7677 - sparse_categorical_accuracy: 0.7015 - val_loss: 1.0475 - val_sparse_categorical_accuracy: 0.6118\n",
      "Epoch 10/10\n",
      "289/289 [==============================] - 19s 67ms/step - loss: 0.7425 - sparse_categorical_accuracy: 0.7147 - val_loss: 1.1352 - val_sparse_categorical_accuracy: 0.5893\n"
     ]
    }
   ],
   "source": [
    "# Build and train model\n",
    "model_sc = build_model(embedding_layer)\n",
    "history_sc = model_sc.fit(\n",
    "    X_train_pad, y_train,\n",
    "    validation_data=(X_valid_pad, y_valid),\n",
    "    epochs=10,\n",
    "    batch_size=128,\n",
    "    verbose=True\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "01f6242b",
   "metadata": {
    "papermill": {
     "duration": 0.391891,
     "end_time": "2023-02-20T21:28:38.223830",
     "exception": false,
     "start_time": "2023-02-20T21:28:37.831939",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "**Learning curves**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "d4cc9a1a",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:28:39.086300Z",
     "iopub.status.busy": "2023-02-20T21:28:39.085454Z",
     "iopub.status.idle": "2023-02-20T21:28:39.608228Z",
     "shell.execute_reply": "2023-02-20T21:28:39.606681Z"
    },
    "papermill": {
     "duration": 0.93923,
     "end_time": "2023-02-20T21:28:39.610823",
     "exception": false,
     "start_time": "2023-02-20T21:28:38.671593",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Final accuracy on validation set: 0.5893335342407227\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot learning curves\n",
    "history_sc_df = pd.DataFrame(history_sc.history)\n",
    "history_sc_df.loc[:, ['loss', 'val_loss']].plot(title=\"Cross-entropy\")\n",
    "history_sc_df.loc[:, ['sparse_categorical_accuracy', 'val_sparse_categorical_accuracy']].plot(title=\"Accuracy\")\n",
    "\n",
    "print('Final accuracy on validation set:', history_sc_df.loc[len(history_sc_df)-1,'val_sparse_categorical_accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "18e6cb40",
   "metadata": {
    "papermill": {
     "duration": 0.383644,
     "end_time": "2023-02-20T21:28:40.377577",
     "exception": false,
     "start_time": "2023-02-20T21:28:39.993933",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "**Evaluate on test set**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "f581ca31",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:28:41.201335Z",
     "iopub.status.busy": "2023-02-20T21:28:41.200889Z",
     "iopub.status.idle": "2023-02-20T21:28:42.926820Z",
     "shell.execute_reply": "2023-02-20T21:28:42.925216Z"
    },
    "papermill": {
     "duration": 2.175019,
     "end_time": "2023-02-20T21:28:42.929843",
     "exception": false,
     "start_time": "2023-02-20T21:28:40.754824",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test set accuracy: 0.5939\n"
     ]
    }
   ],
   "source": [
    "# Make predictions on test set\n",
    "preds_sc = model_sc.predict(X_test_pad)\n",
    "\n",
    "# Retrieve most likely class\n",
    "class_preds_sc = np.argmax(preds_sc,axis=1)\n",
    "\n",
    "# Print accuracy\n",
    "print(f\"Test set accuracy: {accuracy_score(y_test, class_preds_sc):.4f}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f30dfb2e",
   "metadata": {
    "papermill": {
     "duration": 0.395884,
     "end_time": "2023-02-20T21:28:43.731416",
     "exception": false,
     "start_time": "2023-02-20T21:28:43.335532",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "For this task, training our own word embeddings from scratch **performed roughly the same** as using word2vec embeddings. This means there is **sufficient data** to learn good representations for our task. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "aa1361b3",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-02-20T21:28:44.566349Z",
     "iopub.status.busy": "2023-02-20T21:28:44.565871Z",
     "iopub.status.idle": "2023-02-20T21:28:44.929589Z",
     "shell.execute_reply": "2023-02-20T21:28:44.928177Z"
    },
    "papermill": {
     "duration": 0.810151,
     "end_time": "2023-02-20T21:28:44.932314",
     "exception": false,
     "start_time": "2023-02-20T21:28:44.122163",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 576x432 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Confusion matrix\n",
    "cm_sc = confusion_matrix(y_test, class_preds_sc)\n",
    "plt.figure(figsize=(8,6))\n",
    "sns.heatmap(cm_sc, annot=True, fmt='d', cmap='Blues')\n",
    "plt.xlabel('Predicted label')\n",
    "plt.ylabel('True label')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fae49a0d",
   "metadata": {
    "papermill": {
     "duration": 0.393558,
     "end_time": "2023-02-20T21:28:45.731680",
     "exception": false,
     "start_time": "2023-02-20T21:28:45.338122",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "# 5. Conclusion\n",
    "\n",
    "In this notebook, we saw how to vectorize individual words through **embeddings** to capture meaning between words. We showed how Word2vec learnt these embeddings and its remarkable ability to solve **analogies**. Finally, we trained **neural networks** to perform a classification task using pre-trained and co-trained embeddings. \n",
    "\n",
    "Using pre-trained word embeddings works great in situations where there is **little training data**. But if you have sufficient data to train your own embeddings, this will usually perform just as well if not better. \n",
    "\n",
    "There are many alternatives to Word2vec. These include\n",
    "\n",
    "* **GloVe** (Global vectors for word representations) - uses global co-occurance counts. \n",
    "* **Doc2vec** - represents whole documents as dense vectors.\n",
    "* **fastText** - vectorizes subwords, i.e. a character n-gram, which helps deal with out-of-vocabulary scenarios. \n",
    "\n",
    "Even though word embeddings were **revolutionary** when they came out less than a decade ago, they have already become **outdated**. The world has moved on from static word embeddings because they have significant **shortcomings**: \n",
    "\n",
    "* They are **difficult to scale** - models would have to be retrained as languages evolve and new words get invented. \n",
    "* They are **static** - this doesn't capture that words have different meanings depending on their context, e.g. bat can be an object or an animal. \n",
    "\n",
    "Nowaways state-of-the-art NLP models almost all rely on **transformers**, which we will discuss in a later notebook."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0ce68a21",
   "metadata": {
    "papermill": {
     "duration": 0.440349,
     "end_time": "2023-02-20T21:28:46.565138",
     "exception": false,
     "start_time": "2023-02-20T21:28:46.124789",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "**References:**\n",
    "* [NLP demystified](https://www.nlpdemystified.org/)\n",
    "\n",
    "### Coming UP\n",
    "#### [8. Recurrent Neural Networks (RNNs) and Language Modelling](./08_RNNs_LMs.ipynb)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "23843799",
   "metadata": {
    "papermill": {
     "duration": 0.384,
     "end_time": "2023-02-20T21:28:47.329940",
     "exception": false,
     "start_time": "2023-02-20T21:28:46.945940",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "Thanks for reading!"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.12"
  },
  "papermill": {
   "default_parameters": {},
   "duration": 482.192888,
   "end_time": "2023-02-20T21:28:50.813294",
   "environment_variables": {},
   "exception": null,
   "input_path": "__notebook__.ipynb",
   "output_path": "__notebook__.ipynb",
   "parameters": {},
   "start_time": "2023-02-20T21:20:48.620406",
   "version": "2.3.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}