(Model 5) CNN_BILSTM_CRF .ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import csv\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import future"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "DS = pd.read_csv(\"ObjectSubset150SentenceLength.csv\")"
   ]
  },
  {
   "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>text_ID</th>\n",
       "      <th>word</th>\n",
       "      <th>tag</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>46809</td>\n",
       "      <td>RR</td>\n",
       "      <td>object</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>46809</td>\n",
       "      <td>130/80</td>\n",
       "      <td>object</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>46809</td>\n",
       "      <td>mmHg.</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>47482</td>\n",
       "      <td>RR142/89mmHg,</td>\n",
       "      <td>object</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>47482</td>\n",
       "      <td>HR</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>157275</th>\n",
       "      <td>1624218</td>\n",
       "      <td>saamiseks.</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>157276</th>\n",
       "      <td>1624218</td>\n",
       "      <td>RR</td>\n",
       "      <td>object</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>157277</th>\n",
       "      <td>1624218</td>\n",
       "      <td>140/90,</td>\n",
       "      <td>object</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>157278</th>\n",
       "      <td>1624218</td>\n",
       "      <td>VS</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>157279</th>\n",
       "      <td>1624218</td>\n",
       "      <td>4,1.</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>157280 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "        text_ID           word     tag\n",
       "0         46809             RR  object\n",
       "1         46809         130/80  object\n",
       "2         46809          mmHg.     NaN\n",
       "3         47482  RR142/89mmHg,  object\n",
       "4         47482             HR     NaN\n",
       "...         ...            ...     ...\n",
       "157275  1624218     saamiseks.     NaN\n",
       "157276  1624218             RR  object\n",
       "157277  1624218        140/90,  object\n",
       "157278  1624218             VS     NaN\n",
       "157279  1624218           4,1.     NaN\n",
       "\n",
       "[157280 rows x 3 columns]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "DS"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "DS['tag'].fillna('text', inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "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>text_ID</th>\n",
       "      <th>word</th>\n",
       "      <th>tag</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>46809</td>\n",
       "      <td>RR</td>\n",
       "      <td>object</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>46809</td>\n",
       "      <td>130/80</td>\n",
       "      <td>object</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>46809</td>\n",
       "      <td>mmHg.</td>\n",
       "      <td>text</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>47482</td>\n",
       "      <td>RR142/89mmHg,</td>\n",
       "      <td>object</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>47482</td>\n",
       "      <td>HR</td>\n",
       "      <td>text</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>47482</td>\n",
       "      <td>76</td>\n",
       "      <td>text</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>47482</td>\n",
       "      <td>x</td>\n",
       "      <td>text</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>47482</td>\n",
       "      <td>min,</td>\n",
       "      <td>text</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>47482</td>\n",
       "      <td>p167cm,</td>\n",
       "      <td>text</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>47482</td>\n",
       "      <td>k89kg.</td>\n",
       "      <td>text</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>47482</td>\n",
       "      <td>EKG-s</td>\n",
       "      <td>text</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>47482</td>\n",
       "      <td>siinusrütmi</td>\n",
       "      <td>text</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>47482</td>\n",
       "      <td>foonil</td>\n",
       "      <td>text</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>47482</td>\n",
       "      <td>ventrikualarsed</td>\n",
       "      <td>text</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>47482</td>\n",
       "      <td>ekstrasüstolid.</td>\n",
       "      <td>text</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>53393</td>\n",
       "      <td>RR143/87mmHg,</td>\n",
       "      <td>object</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>53393</td>\n",
       "      <td>HR60</td>\n",
       "      <td>text</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>53393</td>\n",
       "      <td>x</td>\n",
       "      <td>text</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>53393</td>\n",
       "      <td>min,</td>\n",
       "      <td>text</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>53393</td>\n",
       "      <td>p180cm,</td>\n",
       "      <td>text</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    text_ID             word     tag\n",
       "0     46809               RR  object\n",
       "1     46809           130/80  object\n",
       "2     46809            mmHg.    text\n",
       "3     47482    RR142/89mmHg,  object\n",
       "4     47482               HR    text\n",
       "5     47482               76    text\n",
       "6     47482                x    text\n",
       "7     47482             min,    text\n",
       "8     47482          p167cm,    text\n",
       "9     47482           k89kg.    text\n",
       "10    47482            EKG-s    text\n",
       "11    47482      siinusrütmi    text\n",
       "12    47482           foonil    text\n",
       "13    47482  ventrikualarsed    text\n",
       "14    47482  ekstrasüstolid.    text\n",
       "15    53393    RR143/87mmHg,  object\n",
       "16    53393             HR60    text\n",
       "17    53393                x    text\n",
       "18    53393             min,    text\n",
       "19    53393          p180cm,    text"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "DS.head(n=20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "class SentenceGetter(object):\n",
    "    def __init__(self, dataset):\n",
    "        self.n_sent = 1\n",
    "        self.dataset = dataset\n",
    "        self.empty = False\n",
    "        agg_func = lambda s: [(w, t) for w,t in zip(s[\"word\"].values.tolist(),\n",
    "                                                        s[\"tag\"].values.tolist())]\n",
    "        self.grouped = self.dataset.groupby(\"text_ID\").apply(agg_func)\n",
    "        self.sentences = [s for s in self.grouped]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "getter = SentenceGetter(DS)\n",
    "sentences = getter.sentences"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('Pt.', 'text'),\n",
       " ('tehtud', 'text'),\n",
       " ('3', 'text'),\n",
       " ('plasmaferees,', 'text'),\n",
       " ('talus', 'text'),\n",
       " ('hästi.', 'text'),\n",
       " ('Eemaldatud', 'text'),\n",
       " ('700', 'text'),\n",
       " ('ml', 'text'),\n",
       " ('plasmat.', 'text'),\n",
       " ('RR', 'object'),\n",
       " ('112/60', 'object'),\n",
       " ('mmHg.', 'text'),\n",
       " ('Tgasi', 'text'),\n",
       " ('22.10.09.', 'text')]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sentences[1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "13643"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(sentences)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# import pickle\n",
    "# with open('sentences5000.txt', 'wb') as f:\n",
    "#   pickle.dump(sentences, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# with open('sentences2000.txt', 'rb') as f:\n",
    "#     sentences = pickle.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "LengthsList = []\n",
    "for i in range(len(sentences)):\n",
    "    LengthsList.append(len(sentences[i]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "10.0"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.median(LengthsList)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "from math import nan\n",
    "\n",
    "words = list(set(DS[\"word\"].values))\n",
    "n_words = len(words)\n",
    "\n",
    "tags = list(set(DS[\"tag\"].values))\n",
    "n_tags = len(tags); \n",
    "\n",
    "\n",
    "from future.utils import iteritems\n",
    "\n",
    "word2idx = {w: i for i, w in enumerate(words)}\n",
    "# tag2idx[\"PAD\"] = 0\n",
    "tags2 = [\"PAD\",\"text\",\"object\"]\n",
    "tag2idx = {t: i for i, t in enumerate(tags2)}\n",
    "idx2tag = {v: k for k, v in iteritems(tag2idx)}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{0: 'PAD', 1: 'text', 2: 'object'}"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "idx2tag"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "27741"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(words)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "150"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "max([len(s) for s in sentences])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from keras.preprocessing.sequence import pad_sequences\n",
    "from keras.utils import to_categorical\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "# maxlen = max([len(s) for s in sentences])\n",
    "maxlen = 150\n",
    "X = [[word2idx[w[0]] for w in s] for s in sentences]\n",
    "X = pad_sequences(maxlen=maxlen, sequences=X, padding=\"post\",value=tag2idx[\"PAD\"], truncating='post')\n",
    "\n",
    "y = [[tag2idx[w[1]] for w in s] for s in sentences]\n",
    "y = pad_sequences(maxlen=maxlen, sequences=y, padding=\"post\", value=tag2idx[\"PAD\"], truncating='post')\n",
    "y = [to_categorical(i, num_classes=n_tags +1) for i in y]\n",
    "\n",
    "# Split train and test data\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2,random_state=2018)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'PAD': 0, 'text': 1, 'object': 2}"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tag2idx"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       [0., 0., 1.],\n",
       "       [0., 0., 1.],\n",
       "       [0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.]], dtype=float32)"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train[3]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       [0., 0., 1.],\n",
       "       [0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       [0., 1., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.],\n",
       "       [1., 0., 0.]], dtype=float32)"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train[2]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 3812, 11371, 18925,  4399,  7346,     0,     0,     0,     0,\n",
       "           0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
       "           0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
       "           0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
       "           0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
       "           0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
       "           0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
       "           0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
       "           0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
       "           0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
       "           0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
       "           0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
       "           0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
       "           0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
       "           0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
       "           0,     0,     0,     0,     0,     0,     0,     0,     0,\n",
       "           0,     0,     0,     0,     0,     0])"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train[1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "# import pickle\n",
    "# def StoreData(myObject,ObjectName):\n",
    "#         with open(ObjectName+'.txt', 'wb') as f:\n",
    "#           pickle.dump(myObject, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "# StoreData(X_train,\"X_train\")\n",
    "# StoreData(X_test,\"X_test\")\n",
    "# StoreData(y_train,\"y_train\")\n",
    "# StoreData(y_test,\"y_test\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "# with open('X_test.txt', 'rb') as f:\n",
    "#     # Step 3\n",
    "#     X_test = pickle.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(10914, 150)"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "# output_dim = 1 + (input_dim - kernel_size)/stride\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "# n_words + 2 - 2 / 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.models import Model, Input\n",
    "from keras.layers import TimeDistributed,SpatialDropout1D,LSTM,MaxPooling1D,Flatten,Embedding, Dense, TimeDistributed, Conv1D,Dropout, Bidirectional\n",
    "import keras as k\n",
    "from keras_contrib.layers import CRF\n",
    "\n",
    "input = Input(shape=(maxlen,))\n",
    "# Embedding Layer\n",
    "model = Embedding(input_dim=n_words + 2  , output_dim=20, input_length=maxlen)(input)\n",
    "\n",
    "\n",
    "# (number of sentences,  words, 100 output dimension or filters)   \n",
    "# batchSize = number of sentences    \n",
    "# length = number of words in each sentence   \n",
    "# channels = dimension of the embedding's output.  \n",
    "model = Conv1D(filters=20,\n",
    "              kernel_size= 5,  # 3 means 3 words\n",
    "              padding='same',  # valid means no padding\n",
    "              strides=1,  # see explnation above\n",
    "              activation='relu')(model)\n",
    "\n",
    "# model = MaxPooling1D(pool_size=2,padding=\"same\")(model) # (?, 27, 10), (?, 24, 10)\n",
    "# model = Flatten()(model) # (?, 270), (?, 240)\n",
    "\n",
    "# model = LSTM(units=20, \n",
    "#              return_sequences=True, \n",
    "#              recurrent_dropout=0.5)(model)     \n",
    "\n",
    "model = SpatialDropout1D(0.3)(model)\n",
    "\n",
    "# BI-LSTM Layer\n",
    "model = Bidirectional(LSTM(units=20, \n",
    "                           return_sequences=True, \n",
    "                           recurrent_dropout=0.5))(model)\n",
    "\n",
    "\n",
    "\n",
    "# TimeDistributed Layer\n",
    "model = TimeDistributed(Dense(n_tags+1, activation=\"softmax\"))(model)  \n",
    "\n",
    "# CRF Layer\n",
    "crf = CRF(n_tags + 1)\n",
    "\n",
    "out = crf(model)  # output\n",
    "model = Model(input, out)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Kasutaja\\Anaconda3\\envs\\myEnv\\lib\\site-packages\\keras_contrib-2.0.8-py3.7.egg\\keras_contrib\\layers\\crf.py:346: UserWarning: CRF.loss_function is deprecated and it might be removed in the future. Please use losses.crf_loss instead.\n",
      "C:\\Users\\Kasutaja\\Anaconda3\\envs\\myEnv\\lib\\site-packages\\keras_contrib-2.0.8-py3.7.egg\\keras_contrib\\layers\\crf.py:353: UserWarning: CRF.accuracy is deprecated and it might be removed in the future. Please use metrics.crf_accuracy\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_1\"\n",
      "____________________________________________________________________________________________________\n",
      "Layer (type)                                 Output Shape                            Param #        \n",
      "====================================================================================================\n",
      "input_1 (InputLayer)                         (None, 150)                             0              \n",
      "____________________________________________________________________________________________________\n",
      "embedding_1 (Embedding)                      (None, 150, 20)                         554860         \n",
      "____________________________________________________________________________________________________\n",
      "conv1d_1 (Conv1D)                            (None, 150, 20)                         2020           \n",
      "____________________________________________________________________________________________________\n",
      "spatial_dropout1d_1 (SpatialDropout1D)       (None, 150, 20)                         0              \n",
      "____________________________________________________________________________________________________\n",
      "bidirectional_1 (Bidirectional)              (None, 150, 40)                         6560           \n",
      "____________________________________________________________________________________________________\n",
      "time_distributed_1 (TimeDistributed)         (None, 150, 3)                          123            \n",
      "____________________________________________________________________________________________________\n",
      "crf_1 (CRF)                                  (None, 150, 3)                          27             \n",
      "====================================================================================================\n",
      "Total params: 563,590\n",
      "Trainable params: 563,590\n",
      "Non-trainable params: 0\n",
      "____________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "from keras.callbacks import ModelCheckpoint\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "\n",
    "# Compile model\n",
    "model.compile(optimizer=\"adam\", loss=crf.loss_function, metrics=[crf.accuracy, 'accuracy'])\n",
    "\n",
    "model.summary(line_length=100)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: GraphViz in c:\\users\\kasutaja\\anaconda3\\envs\\myenv\\lib\\site-packages (0.13.2)\n"
     ]
    }
   ],
   "source": [
    "!pip install GraphViz "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from keras.utils import plot_model\n",
    "plot_model(model, to_file='CNNmodel.png')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from keras.utils.vis_utils import plot_model\n",
    "plot_model(model, to_file='CNNmodel_plot.png', show_shapes=True, show_layer_names=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Kasutaja\\Anaconda3\\envs\\myEnv\\lib\\site-packages\\tensorflow_core\\python\\framework\\indexed_slices.py:433: UserWarning: Converting sparse IndexedSlices to a dense Tensor of unknown shape. This may consume a large amount of memory.\n",
      "  \"Converting sparse IndexedSlices to a dense Tensor of unknown shape. \"\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 9822 samples, validate on 1092 samples\n",
      "Epoch 1/10\n",
      "9822/9822 [==============================] - 36s 4ms/step - loss: 0.6624 - crf_viterbi_accuracy: 0.7980 - accuracy: 0.9226 - val_loss: 0.2856 - val_crf_viterbi_accuracy: 0.9842 - val_accuracy: 0.9842\n",
      "\n",
      "Epoch 00001: val_accuracy improved from -inf to 0.98415, saving model to ner-CNN-bi-lstm-model-OBJECT-0.98.hdf5\n",
      "Epoch 2/10\n",
      "9822/9822 [==============================] - 34s 3ms/step - loss: 0.1806 - crf_viterbi_accuracy: 0.9851 - accuracy: 0.9226 - val_loss: 0.1176 - val_crf_viterbi_accuracy: 0.9857 - val_accuracy: 0.9856\n",
      "\n",
      "Epoch 00002: val_accuracy improved from 0.98415 to 0.98564, saving model to ner-CNN-bi-lstm-model-OBJECT-0.99.hdf5\n",
      "Epoch 3/10\n",
      "9822/9822 [==============================] - 34s 3ms/step - loss: 0.0938 - crf_viterbi_accuracy: 0.9901 - accuracy: 0.9226 - val_loss: 0.0766 - val_crf_viterbi_accuracy: 0.9910 - val_accuracy: 0.9911\n",
      "\n",
      "Epoch 00003: val_accuracy improved from 0.98564 to 0.99107, saving model to ner-CNN-bi-lstm-model-OBJECT-0.99.hdf5\n",
      "Epoch 4/10\n",
      "9822/9822 [==============================] - 34s 3ms/step - loss: 0.0663 - crf_viterbi_accuracy: 0.9909 - accuracy: 0.9226 - val_loss: 0.0549 - val_crf_viterbi_accuracy: 0.9911 - val_accuracy: 0.9912\n",
      "\n",
      "Epoch 00004: val_accuracy improved from 0.99107 to 0.99117, saving model to ner-CNN-bi-lstm-model-OBJECT-0.99.hdf5\n",
      "Epoch 5/10\n",
      "9822/9822 [==============================] - 34s 3ms/step - loss: 0.0476 - crf_viterbi_accuracy: 0.9911 - accuracy: 0.9226 - val_loss: 0.0417 - val_crf_viterbi_accuracy: 0.9911 - val_accuracy: 0.9912\n",
      "\n",
      "Epoch 00005: val_accuracy improved from 0.99117 to 0.99121, saving model to ner-CNN-bi-lstm-model-OBJECT-0.99.hdf5\n",
      "Epoch 6/10\n",
      "9822/9822 [==============================] - 34s 3ms/step - loss: 0.0371 - crf_viterbi_accuracy: 0.9912 - accuracy: 0.9226 - val_loss: 0.0343 - val_crf_viterbi_accuracy: 0.9911 - val_accuracy: 0.9913\n",
      "\n",
      "Epoch 00006: val_accuracy improved from 0.99121 to 0.99125, saving model to ner-CNN-bi-lstm-model-OBJECT-0.99.hdf5\n",
      "Epoch 7/10\n",
      "9822/9822 [==============================] - 34s 3ms/step - loss: 0.0303 - crf_viterbi_accuracy: 0.9917 - accuracy: 0.9226 - val_loss: 0.0287 - val_crf_viterbi_accuracy: 0.9926 - val_accuracy: 0.9926\n",
      "\n",
      "Epoch 00007: val_accuracy improved from 0.99125 to 0.99263, saving model to ner-CNN-bi-lstm-model-OBJECT-0.99.hdf5\n",
      "Epoch 8/10\n",
      "9822/9822 [==============================] - 34s 3ms/step - loss: 0.0252 - crf_viterbi_accuracy: 0.9974 - accuracy: 0.9226 - val_loss: 0.0248 - val_crf_viterbi_accuracy: 0.9982 - val_accuracy: 0.9982\n",
      "\n",
      "Epoch 00008: val_accuracy improved from 0.99263 to 0.99824, saving model to ner-CNN-bi-lstm-model-OBJECT-1.00.hdf5\n",
      "Epoch 9/10\n",
      "9822/9822 [==============================] - 34s 3ms/step - loss: 0.0213 - crf_viterbi_accuracy: 0.9994 - accuracy: 0.9226 - val_loss: 0.0220 - val_crf_viterbi_accuracy: 0.9986 - val_accuracy: 0.9986\n",
      "\n",
      "Epoch 00009: val_accuracy improved from 0.99824 to 0.99857, saving model to ner-CNN-bi-lstm-model-OBJECT-1.00.hdf5\n",
      "Epoch 10/10\n",
      "9822/9822 [==============================] - 34s 3ms/step - loss: 0.0185 - crf_viterbi_accuracy: 0.9997 - accuracy: 0.9226 - val_loss: 0.0199 - val_crf_viterbi_accuracy: 0.9986 - val_accuracy: 0.9986\n",
      "\n",
      "Epoch 00010: val_accuracy improved from 0.99857 to 0.99860, saving model to ner-CNN-bi-lstm-model-OBJECT-1.00.hdf5\n"
     ]
    }
   ],
   "source": [
    "\n",
    "# Saving the best model only\n",
    "filepath=\"ner-CNN-bi-lstm-model-OBJECT-{val_accuracy:.2f}.hdf5\"\n",
    "checkpoint = ModelCheckpoint(filepath, monitor='val_accuracy', verbose=1, save_best_only=True, mode='max')\n",
    "callbacks_list = [checkpoint]\n",
    "\n",
    "# Fit the best model\n",
    "history = model.fit(X_train, np.array(y_train), batch_size=32, epochs=10, validation_split=0.1, verbose=1, callbacks=callbacks_list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x360 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot the graph \n",
    "plt.style.use('ggplot')\n",
    "\n",
    "def plot_history(history):\n",
    "    accuracy = history.history['accuracy']\n",
    "    val_accuracy = history.history['val_accuracy']\n",
    "    loss = history.history['loss']\n",
    "    val_loss = history.history['val_loss']\n",
    "    x = range(1, len(accuracy) + 1)\n",
    "\n",
    "    plt.figure(figsize=(12, 5))\n",
    "    plt.subplot(1, 2, 1)\n",
    "    plt.plot(x, accuracy, 'b', label='Training acc')\n",
    "    plt.plot(x, val_accuracy, 'r', label='Validation acc')\n",
    "    plt.title('Training and validation accuracy')\n",
    "    plt.legend()\n",
    "    plt.subplot(1, 2, 2)\n",
    "    plt.plot(x, loss, 'b', label='Training loss')\n",
    "    plt.plot(x, val_loss, 'r', label='Validation loss')\n",
    "    plt.title('Training and validation loss')\n",
    "    plt.legend()\n",
    "\n",
    "plot_history(history)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{0: 'PAD', 1: 'text', 2: 'object'}"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "idx2tag"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2729/2729 [==============================] - 2s 904us/step\n"
     ]
    }
   ],
   "source": [
    "test_pred = model.predict(X_test, verbose=1)   \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "def pred2label(pred):\n",
    "    out = []\n",
    "    for pred_i in pred:\n",
    "        out_i = []\n",
    "        for p in pred_i:\n",
    "            p_i = np.argmax(p)\n",
    "            out_i.append(idx2tag[p_i].replace(\"PAD\",\"text\"))\n",
    "        out.append(out_i)\n",
    "    return out\n",
    "# test_pred = model.predict(X_test, verbose=1)   \n",
    "pred_labels = pred2label(test_pred)\n",
    "test_labels = pred2label(y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "F1-score: 86.2%\n"
     ]
    }
   ],
   "source": [
    "from seqeval.metrics import precision_score, recall_score, f1_score, classification_report\n",
    "print(\"F1-score: {:.1%}\".format(f1_score(test_labels, pred_labels)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "           precision    recall  f1-score   support\n",
      "\n",
      "     text       0.86      0.86      0.86      4461\n",
      "   object       0.85      0.87      0.86      2104\n",
      "\n",
      "micro avg       0.86      0.87      0.86      6565\n",
      "macro avg       0.86      0.87      0.86      6565\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(test_labels, pred_labels))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "      object       0.91      0.93      0.92      3653\n",
      "        text       1.00      1.00      1.00    405697\n",
      "\n",
      "    accuracy                           1.00    409350\n",
      "   macro avg       0.96      0.97      0.96    409350\n",
      "weighted avg       1.00      1.00      1.00    409350\n",
      "\n"
     ]
    }
   ],
   "source": [
    "from  sklearn_crfsuite.metrics import flat_classification_report  \n",
    "report = flat_classification_report(y_pred=pred_labels, y_true=test_labels)\n",
    "print(report)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Word           ||True ||Pred\n",
      "30.04.2013     : text  text\n",
      "-              : text  text\n",
      "kaal           : text  text\n",
      "58             : text  text\n",
      "kg,            : text  text\n",
      "pikkus         : text  text\n",
      "164            : text  text\n",
      "cm,            : text  text\n",
      "KMI            : text  text\n",
      "20             : text  text\n",
      "30.04.2013     : text  text\n",
      "-              : text  text\n",
      "RR             : object object\n",
      "145/79,        : object object\n",
      "fr.            : text  text\n",
      "86             : text  text\n",
      "x'             : text  text\n",
      ".              : text  text\n"
     ]
    }
   ],
   "source": [
    "print(\"{:15}||{:5}||{}\".format(\"Word\", \"True\", \"Pred\"))\n",
    "\n",
    "i=5\n",
    "for w,true, pred in zip(X_test[i], test_labels[i] ,pred_labels[i]):\n",
    "    if words[w] != \"Urea\":\n",
    "        print(\"{:15}: {:5} {}\".format(words[w],true,pred))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_sentence = \" Tupp vaba, em.kael puhas, RR 155/85 mmhg .\"\n",
    "test_sentence = ''.join(test_sentence).split() "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_test_sent = pad_sequences(sequences=[[word2idx.get(w, 0) for w in test_sentence]],\n",
    "                            padding=\"post\", value=tag2idx[\"PAD\"], truncating='post', maxlen=maxlen)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Word           ||Prediction\n",
      "==============================\n",
      "Tupp           : text \n",
      "vaba,          : text \n",
      "em.kael        : text \n",
      "puhas,         : text \n",
      "RR             : object\n",
      "155/85         : object\n",
      "mmhg           : text \n",
      ".              : text \n"
     ]
    }
   ],
   "source": [
    "p = model.predict(x_test_sent)\n",
    "p = pred2label(p)\n",
    "# p = np.argmax(p, axis=-1)\n",
    "print(\"{:15}||{}\".format(\"Word\", \"Prediction\"))\n",
    "print(30 * \"=\")\n",
    "for w, pred in zip(test_sentence, p[0]):\n",
    "    print(\"{:15}: {:5}\".format(w, pred))\n"
   ]
  }
 ],
 "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}