Embedding.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import itertools as it\n",
    "import matplotlib.pyplot as plt\n",
    "from tensorflow.keras.layers import Embedding, Dense, Flatten\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.optimizers import Adam\n",
    "from sklearn.preprocessing import OneHotEncoder\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "#!pip install --upgrade numpy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'2.2.0'"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tf.__version__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "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>poet</th>\n",
       "      <th>century</th>\n",
       "      <th>book_title</th>\n",
       "      <th>poem_title</th>\n",
       "      <th>poem</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>رودکی</td>\n",
       "      <td>3</td>\n",
       "      <td>قصاید و قطعات</td>\n",
       "      <td>شمارهٔ ۱</td>\n",
       "      <td>گر من این دوستی تو ببرم تا لب گور    بزنم نعره...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>رودکی</td>\n",
       "      <td>3</td>\n",
       "      <td>قصاید و قطعات</td>\n",
       "      <td>شمارهٔ ۲</td>\n",
       "      <td>به حق نالم ز هجر دوست زارا    سحرگاهان چو بر گ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>رودکی</td>\n",
       "      <td>3</td>\n",
       "      <td>قصاید و قطعات</td>\n",
       "      <td>شمارهٔ ۳</td>\n",
       "      <td>به نام نیک تو، خواجه، فریفته نشوم    که نام نی...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>رودکی</td>\n",
       "      <td>3</td>\n",
       "      <td>قصاید و قطعات</td>\n",
       "      <td>شمارهٔ ۴</td>\n",
       "      <td>دلا، تا کی همی جویی منی را؟    چه داری دوست هر...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>رودکی</td>\n",
       "      <td>3</td>\n",
       "      <td>قصاید و قطعات</td>\n",
       "      <td>شمارهٔ ۵</td>\n",
       "      <td>گرفت خواهم زلفین عنبرین ترا    به بوسه نقش‌کنم...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    poet  century     book_title poem_title  \\\n",
       "0  رودکی        3  قصاید و قطعات   شمارهٔ ۱   \n",
       "1  رودکی        3  قصاید و قطعات   شمارهٔ ۲   \n",
       "2  رودکی        3  قصاید و قطعات   شمارهٔ ۳   \n",
       "3  رودکی        3  قصاید و قطعات   شمارهٔ ۴   \n",
       "4  رودکی        3  قصاید و قطعات   شمارهٔ ۵   \n",
       "\n",
       "                                                poem  \n",
       "0  گر من این دوستی تو ببرم تا لب گور    بزنم نعره...  \n",
       "1  به حق نالم ز هجر دوست زارا    سحرگاهان چو بر گ...  \n",
       "2  به نام نیک تو، خواجه، فریفته نشوم    که نام نی...  \n",
       "3  دلا، تا کی همی جویی منی را؟    چه داری دوست هر...  \n",
       "4  گرفت خواهم زلفین عنبرین ترا    به بوسه نقش‌کنم...  "
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "poems = pd.read_csv('../data/poems/poems_with_more_info.tsv',sep = '\\t')\n",
    "poems.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "def sliding_window(txt):\n",
    "   for i in range(len(txt) - 2):\n",
    "    yield txt[i], txt[i + 1], txt[i + 2]\n",
    "\n",
    "window = list(it.chain(*[sliding_window(_) for _ in poems.poem[:1000]]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "mapping = {c: i for i, c in enumerate(pd.DataFrame(window)[0].unique())}\n",
    "integers_in = np.array([[mapping[w[0]], mapping[w[1]]] for w in window])\n",
    "integers_out = np.array([mapping[w[2]] for w in window]).reshape(-1, 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((2095133, 2), (2095133, 1))"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "integers_in.shape, integers_out.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "گ\n",
      "ر\n",
      " \n",
      "م\n",
      "ن\n",
      "ا\n",
      "ی\n",
      "د\n",
      "و\n",
      "س\n",
      "ت\n",
      "ب\n",
      "ل\n",
      "ز\n",
      "ع\n",
      "ه\n",
      "ک\n",
      "ث\n",
      "خ\n",
      "ج\n",
      "ف\n",
      "،\n",
      "ش\n",
      "ح\n",
      "ق\n",
      "چ\n",
      "ض\n",
      "‌\n",
      "پ\n",
      "آ\n",
      "غ\n",
      "؟\n",
      "ٔ\n",
      ":\n",
      "ژ\n",
      "!\n",
      "ذ\n",
      "ص\n",
      "ط\n",
      "ئ\n",
      "ّ\n",
      "ظ\n",
      ".\n",
      "ِ\n",
      "ؤ\n",
      "-\n",
      "«\n",
      "»\n",
      "ة\n",
      "ء\n",
      "(\n",
      ")\n",
      "أ\n"
     ]
    }
   ],
   "source": [
    "for i in mapping.keys():\n",
    "    print(i)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2095133, 2, 2)"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "num_letters = len(mapping)\n",
    "\n",
    "model = Sequential()\n",
    "embedding = Embedding(num_letters,2,input_length = 2)\n",
    "model.add(embedding)\n",
    "output = model.predict(integers_in)\n",
    "output.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "idx_to_calc = list(mapping.values())\n",
    "idx_to_calc = np.array([idx_to_calc, idx_to_calc]).T\n",
    "translator = {v:k for k,v in mapping.items()}\n",
    "preds = model.predict(idx_to_calc)\n",
    "\n",
    "plt.scatter(preds[:, 0, 0], preds[:, 0, 1], alpha=0)\n",
    "for i, idx in enumerate(idx_to_calc):\n",
    "  plt.text(preds[i, 0, 0], preds[i, 0, 1], translator[idx[0]])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_emb = Sequential()\n",
    "model_emb.add(embedding)\n",
    "model_emb.add(Flatten())\n",
    "model_emb.add(Dense(num_letters,activation = 'softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_emb.compile(optimizer = Adam(learning_rate=0.001, beta_1=0.9, beta_2=0.999, amsgrad=False),\n",
    "                  loss = 'categorical_crossentropy',\n",
    "                  metrics=['accuracy']\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "to_predict = OneHotEncoder(sparse=False).fit_transform(integers_out)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/30\n",
      "65473/65473 [==============================] - 108s 2ms/step - loss: 2.4530 - accuracy: 0.3332\n",
      "Epoch 2/30\n",
      "65473/65473 [==============================] - 78s 1ms/step - loss: 2.4244 - accuracy: 0.3357\n",
      "Epoch 3/30\n",
      "65473/65473 [==============================] - 69s 1ms/step - loss: 2.4228 - accuracy: 0.3360\n",
      "Epoch 4/30\n",
      "65473/65473 [==============================] - 67s 1ms/step - loss: 2.4222 - accuracy: 0.3361\n",
      "Epoch 5/30\n",
      "65473/65473 [==============================] - 66s 1ms/step - loss: 2.4220 - accuracy: 0.3362\n",
      "Epoch 6/30\n",
      "65473/65473 [==============================] - 60s 911us/step - loss: 2.4217 - accuracy: 0.3362\n",
      "Epoch 7/30\n",
      "65473/65473 [==============================] - 64s 980us/step - loss: 2.4216 - accuracy: 0.3363\n",
      "Epoch 8/30\n",
      "65473/65473 [==============================] - 61s 937us/step - loss: 2.4215 - accuracy: 0.3362\n",
      "Epoch 9/30\n",
      "65473/65473 [==============================] - 65s 993us/step - loss: 2.4214 - accuracy: 0.3362\n",
      "Epoch 10/30\n",
      "65473/65473 [==============================] - 61s 933us/step - loss: 2.4213 - accuracy: 0.3362\n",
      "Epoch 11/30\n",
      "65473/65473 [==============================] - 61s 931us/step - loss: 2.4212 - accuracy: 0.3362\n",
      "Epoch 12/30\n",
      "65473/65473 [==============================] - 54s 827us/step - loss: 2.4212 - accuracy: 0.3362\n",
      "Epoch 13/30\n",
      "65473/65473 [==============================] - 54s 818us/step - loss: 2.4211 - accuracy: 0.3363\n",
      "Epoch 14/30\n",
      "65473/65473 [==============================] - 61s 928us/step - loss: 2.4211 - accuracy: 0.3363\n",
      "Epoch 15/30\n",
      "65473/65473 [==============================] - 61s 936us/step - loss: 2.4211 - accuracy: 0.3362\n",
      "Epoch 16/30\n",
      "65473/65473 [==============================] - 62s 951us/step - loss: 2.4210 - accuracy: 0.3363\n",
      "Epoch 17/30\n",
      "65473/65473 [==============================] - 69s 1ms/step - loss: 2.4210 - accuracy: 0.3363\n",
      "Epoch 18/30\n",
      "65473/65473 [==============================] - 57s 863us/step - loss: 2.4210 - accuracy: 0.3362\n",
      "Epoch 19/30\n",
      "65473/65473 [==============================] - 62s 952us/step - loss: 2.4209 - accuracy: 0.3360\n",
      "Epoch 20/30\n",
      "65473/65473 [==============================] - 55s 839us/step - loss: 2.4209 - accuracy: 0.3360s - loss: 2.4209 - accu\n",
      "Epoch 21/30\n",
      "65473/65473 [==============================] - 74s 1ms/step - loss: 2.4209 - accuracy: 0.3362\n",
      "Epoch 22/30\n",
      "65473/65473 [==============================] - 74s 1ms/step - loss: 2.4209 - accuracy: 0.3361\n",
      "Epoch 23/30\n",
      "65473/65473 [==============================] - 70s 1ms/step - loss: 2.4209 - accuracy: 0.3361\n",
      "Epoch 24/30\n",
      "65473/65473 [==============================] - 71s 1ms/step - loss: 2.4209 - accuracy: 0.3360\n",
      "Epoch 25/30\n",
      "65473/65473 [==============================] - 67s 1ms/step - loss: 2.4209 - accuracy: 0.3363\n",
      "Epoch 26/30\n",
      "65473/65473 [==============================] - 71s 1ms/step - loss: 2.4209 - accuracy: 0.3361\n",
      "Epoch 27/30\n",
      "65473/65473 [==============================] - 206s 3ms/step - loss: 2.4209 - accuracy: 0.3361\n",
      "Epoch 28/30\n",
      "65473/65473 [==============================] - 190s 3ms/step - loss: 2.4209 - accuracy: 0.3362\n",
      "Epoch 29/30\n",
      "65473/65473 [==============================] - 237s 4ms/step - loss: 2.4208 - accuracy: 0.3359\n",
      "Epoch 30/30\n",
      "65473/65473 [==============================] - 222s 3ms/step - loss: 2.4208 - accuracy: 0.3361\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<tensorflow.python.keras.callbacks.History at 0x25a6972c248>"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_emb.fit(x = integers_in,\n",
    "             y = to_predict,\n",
    "              epochs=30)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize = (10,10))\n",
    "preds = model.predict(idx_to_calc)\n",
    "plt.scatter(preds[:, 0, 0], preds[:, 0, 1], alpha=0)\n",
    "for i, idx in enumerate(idx_to_calc):\n",
    "  plt.text(preds[i, 0, 0], preds[i, 0, 1], translator[idx[0]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(53, 2, 2)"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "preds.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### منابع\n",
    "\n",
    "https://www.youtube.com/watch?v=mWvnlVw_LiY&list=PL75e0qA87dlG-za8eLI6t0_Pbxafk-cxb&index=5"
   ]
  }
 ],
 "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": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}