exploratory_data.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "from datasets import Dataset\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "from scipy.stats import chi2_contingency, pointbiserialr\n",
    "from sklearn.feature_extraction.text import CountVectorizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# load data\n",
    "train = Dataset.load_from_disk(\"unaugmented_data/train\")\n",
    "val = Dataset.load_from_disk(\"unaugmented_data/val\")\n",
    "dev = Dataset.load_from_disk(\"unaugmented_data/dev\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = {}\n",
    "data['train'] = pd.DataFrame(train)\n",
    "data['val'] = pd.DataFrame(val)\n",
    "data['dev'] = pd.DataFrame(dev)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "10468\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "993"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# determine the size of dataset\n",
    "temp = pd.concat([df for df in data.values()])\n",
    "print(len(temp.par_id.unique())) # total size\n",
    "len(temp.par_id.unique()[temp.pcl == 1]) # total number of PCL paragraphs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "8375\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "794"
      ]
     },
     "execution_count": 67,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# determine size of dataset for training data\n",
    "temp = pd.concat([data['train'], data['val']])\n",
    "print(len(temp.par_id.unique())) # total size\n",
    "len(temp.par_id.unique()[temp.pcl == 1]) # total number of PCL paragraphs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "3832"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# check the test data size\n",
    "test_data = pd.read_csv(\"data/task4_test.tsv\", sep='\\t', header=None)\n",
    "len(test_data[0].unique())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Remove extra prefix in *art_id* column"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "for df in data.values():\n",
    "    np.all(df.art_id.apply(lambda x: True if x[:2] == \"@@\" else False)) # check that all art_id has same prefix '@@'\n",
    "    df['art_id'] = df.art_id.apply(lambda x: x[2:]).astype('int64') # perform removal of prefix\n",
    "    df['art_id'] = df['art_id'].astype('int64') # convert to integer type"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Exploratory Data Analysis (on training data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "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>par_id</th>\n",
       "      <th>art_id</th>\n",
       "      <th>keyword</th>\n",
       "      <th>country_code</th>\n",
       "      <th>text</th>\n",
       "      <th>label_score</th>\n",
       "      <th>label</th>\n",
       "      <th>label_category_vector</th>\n",
       "      <th>input_ids</th>\n",
       "      <th>attention_mask</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>99</td>\n",
       "      <td>25281193</td>\n",
       "      <td>in-need</td>\n",
       "      <td>my</td>\n",
       "      <td>The departures from London will barely put a d...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>[0, 0, 0, 0, 0, 0, 0]</td>\n",
       "      <td>[101, 1996, 6712, 2015, 2013, 2414, 2097, 4510...</td>\n",
       "      <td>[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3162</td>\n",
       "      <td>20664287</td>\n",
       "      <td>migrant</td>\n",
       "      <td>nz</td>\n",
       "      <td>\" Singh is the most common surname for our dri...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>[0, 0, 0, 0, 0, 0, 0]</td>\n",
       "      <td>[101, 1000, 5960, 2003, 1996, 2087, 2691, 1198...</td>\n",
       "      <td>[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>246</td>\n",
       "      <td>20000497</td>\n",
       "      <td>homeless</td>\n",
       "      <td>au</td>\n",
       "      <td>What causes someone to become homeless ? Brain...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>[0, 0, 0, 0, 0, 0, 0]</td>\n",
       "      <td>[101, 2054, 5320, 2619, 2000, 2468, 11573, 102...</td>\n",
       "      <td>[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>7613</td>\n",
       "      <td>15679266</td>\n",
       "      <td>vulnerable</td>\n",
       "      <td>za</td>\n",
       "      <td>- An extended-hour 's clinic pilot project was...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>[0, 0, 0, 0, 0, 0, 0]</td>\n",
       "      <td>[101, 1011, 2019, 3668, 1011, 3178, 1005, 1055...</td>\n",
       "      <td>[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>98</td>\n",
       "      <td>15438640</td>\n",
       "      <td>immigrant</td>\n",
       "      <td>pk</td>\n",
       "      <td>The Golden State is home to the country 's lar...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>[0, 0, 0, 0, 0, 0, 0]</td>\n",
       "      <td>[101, 1996, 3585, 2110, 2003, 2188, 2000, 1996...</td>\n",
       "      <td>[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1670</th>\n",
       "      <td>1029</td>\n",
       "      <td>16905169</td>\n",
       "      <td>homeless</td>\n",
       "      <td>za</td>\n",
       "      <td>Muizenberg residents have lambasted the Muizen...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>[0, 0, 0, 0, 0, 0, 0]</td>\n",
       "      <td>[101, 14163, 4697, 11144, 3901, 2031, 12559, 1...</td>\n",
       "      <td>[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1671</th>\n",
       "      <td>2767</td>\n",
       "      <td>23925</td>\n",
       "      <td>disabled</td>\n",
       "      <td>in</td>\n",
       "      <td>Christian missionaries care for the destitute ...</td>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>[1, 1, 0, 0, 0, 0, 0]</td>\n",
       "      <td>[101, 3017, 11743, 2729, 2005, 1996, 4078, 377...</td>\n",
       "      <td>[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1672</th>\n",
       "      <td>5532</td>\n",
       "      <td>10345914</td>\n",
       "      <td>women</td>\n",
       "      <td>gh</td>\n",
       "      <td>I bid you fraternal greetings Hon. Alex Bapula...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>[0, 0, 0, 0, 0, 0, 0]</td>\n",
       "      <td>[101, 1045, 7226, 2017, 25312, 16451, 2389, 14...</td>\n",
       "      <td>[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1673</th>\n",
       "      <td>3641</td>\n",
       "      <td>70027514</td>\n",
       "      <td>hopeless</td>\n",
       "      <td>jm</td>\n",
       "      <td>\" He had one good Ashes series and what he del...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>[0, 0, 0, 0, 0, 0, 0]</td>\n",
       "      <td>[101, 1000, 2002, 2018, 2028, 2204, 11289, 218...</td>\n",
       "      <td>[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1674</th>\n",
       "      <td>3763</td>\n",
       "      <td>7428468</td>\n",
       "      <td>in-need</td>\n",
       "      <td>gb</td>\n",
       "      <td>' We also thought that the baby in her uterus ...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>[0, 0, 0, 0, 0, 0, 0]</td>\n",
       "      <td>[101, 1005, 2057, 2036, 2245, 2008, 1996, 3336...</td>\n",
       "      <td>[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>8375 rows × 10 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      par_id    art_id     keyword country_code  \\\n",
       "0         99  25281193     in-need           my   \n",
       "1       3162  20664287     migrant           nz   \n",
       "2        246  20000497    homeless           au   \n",
       "3       7613  15679266  vulnerable           za   \n",
       "4         98  15438640   immigrant           pk   \n",
       "...      ...       ...         ...          ...   \n",
       "1670    1029  16905169    homeless           za   \n",
       "1671    2767     23925    disabled           in   \n",
       "1672    5532  10345914       women           gh   \n",
       "1673    3641  70027514    hopeless           jm   \n",
       "1674    3763   7428468     in-need           gb   \n",
       "\n",
       "                                                   text  label_score  label  \\\n",
       "0     The departures from London will barely put a d...            0      0   \n",
       "1     \" Singh is the most common surname for our dri...            0      0   \n",
       "2     What causes someone to become homeless ? Brain...            1      0   \n",
       "3     - An extended-hour 's clinic pilot project was...            0      0   \n",
       "4     The Golden State is home to the country 's lar...            0      0   \n",
       "...                                                 ...          ...    ...   \n",
       "1670  Muizenberg residents have lambasted the Muizen...            0      0   \n",
       "1671  Christian missionaries care for the destitute ...            4      1   \n",
       "1672  I bid you fraternal greetings Hon. Alex Bapula...            0      0   \n",
       "1673  \" He had one good Ashes series and what he del...            0      0   \n",
       "1674  ' We also thought that the baby in her uterus ...            0      0   \n",
       "\n",
       "      label_category_vector  \\\n",
       "0     [0, 0, 0, 0, 0, 0, 0]   \n",
       "1     [0, 0, 0, 0, 0, 0, 0]   \n",
       "2     [0, 0, 0, 0, 0, 0, 0]   \n",
       "3     [0, 0, 0, 0, 0, 0, 0]   \n",
       "4     [0, 0, 0, 0, 0, 0, 0]   \n",
       "...                     ...   \n",
       "1670  [0, 0, 0, 0, 0, 0, 0]   \n",
       "1671  [1, 1, 0, 0, 0, 0, 0]   \n",
       "1672  [0, 0, 0, 0, 0, 0, 0]   \n",
       "1673  [0, 0, 0, 0, 0, 0, 0]   \n",
       "1674  [0, 0, 0, 0, 0, 0, 0]   \n",
       "\n",
       "                                              input_ids  \\\n",
       "0     [101, 1996, 6712, 2015, 2013, 2414, 2097, 4510...   \n",
       "1     [101, 1000, 5960, 2003, 1996, 2087, 2691, 1198...   \n",
       "2     [101, 2054, 5320, 2619, 2000, 2468, 11573, 102...   \n",
       "3     [101, 1011, 2019, 3668, 1011, 3178, 1005, 1055...   \n",
       "4     [101, 1996, 3585, 2110, 2003, 2188, 2000, 1996...   \n",
       "...                                                 ...   \n",
       "1670  [101, 14163, 4697, 11144, 3901, 2031, 12559, 1...   \n",
       "1671  [101, 3017, 11743, 2729, 2005, 1996, 4078, 377...   \n",
       "1672  [101, 1045, 7226, 2017, 25312, 16451, 2389, 14...   \n",
       "1673  [101, 1000, 2002, 2018, 2028, 2204, 11289, 218...   \n",
       "1674  [101, 1005, 2057, 2036, 2245, 2008, 1996, 3336...   \n",
       "\n",
       "                                         attention_mask  \n",
       "0     [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...  \n",
       "1     [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...  \n",
       "2     [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...  \n",
       "3     [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...  \n",
       "4     [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...  \n",
       "...                                                 ...  \n",
       "1670  [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...  \n",
       "1671  [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...  \n",
       "1672  [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...  \n",
       "1673  [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...  \n",
       "1674  [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...  \n",
       "\n",
       "[8375 rows x 10 columns]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_val_df = pd.concat([data['train'], data['val']])\n",
    "train_val_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "12      0\n",
       "99      0\n",
       "117     1\n",
       "118     0\n",
       "129     0\n",
       "       ..\n",
       "1579    0\n",
       "1587    0\n",
       "1610    0\n",
       "1650    0\n",
       "1653    0\n",
       "Name: label, Length: 475, dtype: int64"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_val_df['pcl'][train_val_df['art_id'].map(train_val_df['art_id'].value_counts()) > 1]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Analysis of class labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create helper function to determine correlation between features and label\n",
    "def cramer_V(freq_table):\n",
    "    x2 = chi2_contingency(freq_table, correction=False)[0]\n",
    "    return np.sqrt((x2 / freq_table.values.sum()) / (min(freq_table.shape) - 1))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### By length of text "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "SignificanceResult(statistic=0.03701849535853825, pvalue=0.0007030057772265076)"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# determine the point biserial correlation coefficient between length of text and prediction\n",
    "text_len = train_val_df['text'].apply(len)\n",
    "pointbiserialr(train_val_df.pcl, text_len)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### By number of tokens"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "SignificanceResult(statistic=0.04939425266677241, pvalue=6.114179022448945e-06)"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# determine the point biserial correlation coefficient between length of text and prediction\n",
    "vectorizer = CountVectorizer()\n",
    "train_val_tokens = vectorizer.fit_transform(train_val_df.text.values)\n",
    "num_tokens = np.sum(train_val_tokens.todense(), axis=1).flat\n",
    "pointbiserialr(train_val_df.pcl, num_tokens)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### By repetition of tokens / maximum occurence of token"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "SignificanceResult(statistic=0.04923775027429294, pvalue=6.542441400616269e-06)"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# determine the point biserial correlation coefficient between maximum occurence of token and prediction\n",
    "vectorizer = CountVectorizer()\n",
    "train_val_tokens = vectorizer.fit_transform(train_val_df.text.values)\n",
    "num_tokens = np.max(train_val_tokens.todense(), axis=1).flat\n",
    "pointbiserialr(train_val_df.pcl, num_tokens)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### By Country"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "country_code\n",
      "gh    100\n",
      "ng     94\n",
      "ph     92\n",
      "za     80\n",
      "ie     74\n",
      "jm     74\n",
      "gb     74\n",
      "ca     62\n",
      "pk     62\n",
      "my     60\n",
      "lk     58\n",
      "bd     58\n",
      "nz     54\n",
      "sg     54\n",
      "us     54\n",
      "ke     52\n",
      "in     48\n",
      "au     48\n",
      "tz     44\n",
      "hk     26\n",
      "Name: count, dtype: int64\n",
      "country_code\n",
      "sg    656\n",
      "au    650\n",
      "pk    644\n",
      "us    640\n",
      "ke    630\n",
      "in    626\n",
      "my    624\n",
      "za    618\n",
      "ph    618\n",
      "lk    616\n",
      "ca    616\n",
      "hk    616\n",
      "nz    614\n",
      "gh    612\n",
      "bd    598\n",
      "ng    592\n",
      "ie    586\n",
      "gb    554\n",
      "jm    544\n",
      "tz    478\n",
      "Name: count, dtype: int64\n"
     ]
    }
   ],
   "source": [
    "# examine the number of PCL by country\n",
    "print(train_val_df[train_val_df['pcl'] == 1].country_code.value_counts())\n",
    "print(train_val_df[train_val_df['pcl'] != 1].country_code.value_counts())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# plot frequency plot\n",
    "train_val_df.groupby(['country_code', 'pcl']).size().unstack().plot(\n",
    "    kind='bar', stacked=True, color=['#a51900', '#02893b'], xlabel=\"country_code\", ylabel=\"Count\"\n",
    ")\n",
    "plt.legend(['No', 'Yes'], title=\"pcl\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.07618400627076896"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# determine the correlation between country and label using Cramer's V\n",
    "cramer_V(pd.crosstab(train_val_df['country_code'], train_val_df['pcl']))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### By Keyword"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "keyword\n",
       "disabled         0.080336\n",
       "homeless         0.172254\n",
       "hopeless         0.124365\n",
       "immigrant        0.027284\n",
       "in-need          0.167056\n",
       "migrant          0.035147\n",
       "poor-families    0.155772\n",
       "refugee          0.082955\n",
       "vulnerable       0.068886\n",
       "women            0.045400\n",
       "dtype: float64"
      ]
     },
     "execution_count": 71,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# get percentage of postive for each keyword\n",
    "keyword_tab = pd.crosstab(train_val_df['keyword'], train_val_df['pcl'])\n",
    "keyword_tab[1] / (keyword_tab[0] + keyword_tab[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "train_val_df.groupby(['keyword', 'pcl']).size().unstack().plot(\n",
    "    kind='bar', stacked=True, color=['#a51900', '#02893b'], xlabel=\"keyword\", ylabel=\"Count\"\n",
    ")\n",
    "plt.legend(['No', 'Yes'], title=\"pcl\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.1786041630371666"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# determine the correlation between keyword and label using Cramer's V\n",
    "cramer_V(pd.crosstab(train_val_df['keyword'], train_val_df['pcl']))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### By article ID"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.989978409493731"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# determine the correlation between keyword and label using Cramer's V\n",
    "cramer_V(pd.crosstab(train_val_df['art_id'], train_val_df['pcl']))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Qualitative Assessment of Dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Print several examples of text based on label score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['Two elderly mainland Chinese women have been arrested for allegedly setting off firecrackers outside the US Consulate , shortly after they entered Hong Kong . They said they had met with unfair treatment in the mainland , but their motive is unknown . <h> Support independent media <h> Partner with HKFP <h> What is Hong Kong Free Press ?',\n",
       "       \"2018 Commonwealth Games Ghana 's Hor Halutie advances to 100m final <h> Ghanaian sprinter , Hor Halutie , has qualified to Monday 's women 's 100m final at the on-going Commonwealth Games .\",\n",
       "       'Slate : Is the dominance of women right now a good thing ?',\n",
       "       'She denied that she had organised any events under the women \\'s league banner , but admitted that , in Women \\'s Month such as August is , and throughout the year she \\'s been getting a lot of invites \" as a gender activist in different capacities \" .',\n",
       "       '\" It is worth trying , \" he said , because \" there are women , children , innocent civilians who can get food and medical supplies to get some relief from the terror of constant bombings . \" Citing \" grave differences , \" Obama said that \" it \\'s premature for us to say there \\'s a clear path forward , but there \\'s the possibility , at least for us , to make some progress . \"',\n",
       "       '\" So many women and people of color and underrepresented people in general think , \\' I \\'m too young \\' or \\' people are going to think that this is silly or ill-advised , \\' \" said Baxter , who currently works for the Op-Ed Project , a non-profit dedicated to lifting up voices of underrepresented experts . \" Your knowledge and experience matters if you step up and say it does . Even just saying , \\' I do want to run for office one day , \\' will put me in a better position to connect with people who know how to make that happen . \"',\n",
       "       \"But League Central is also conscious of the backlash from the increased involvement of gambling companies in the sport , banning all wagering sponsorship and signage from its new women 's competition .\",\n",
       "       'Further , regardless of self-perceived social status , both men and women were found to use a higher pitch while answering to people they think are dominant , prestigious and have a higher social status .',\n",
       "       'Really Thomas , that is the most pathetic attempt at diversion . In fact , it borders on the misogynistic . Are you suggesting that women can not be called control freaks without it being sexist ?',\n",
       "       'Rape had been used as a systematized instrument of war by the racist forces of Sri Lankan Army against female LTTE for physical and moral destruction and now the occupying Sri Lankan Army still continues to do so and so for Sri Lankan Army is concerned rape is perpetrated as a weapon of war . In all the videos and photos that have come into the public domain indicate that Eelam women and female LTTE captives were raped and shot in the head .',\n",
       "       'In addition , \" having ICT ( infrastructure ) would be very beneficial for women , because it allows them to work from home , \" said Betti Alisjahbana , founder of ICT services and leadership development company QB International .',\n",
       "       'Mwakyembe said the challenge dubbed \" Women pre-Olympic Trials towards the 2020 Tokyo Olympics \" seeks to uncover new women talents , ripe to compete in international tournaments .',\n",
       "       'A report in Harian Metro said there were roughly 57 Malaysians with IS groups in Syria alone -- including women and children -- trapped as the IS groups continue to fall . Of the 57 Malaysians , 20 are females .',\n",
       "       \"Among China 's best known businesswomen , Zhang Xin leads real estate developer Soho China with her husband Pan Shiyi . Forbes places Zhang 's net worth at $3.2 billion , which is good enough for 62nd place on the rich list . The one-time factory girl has become a celebrity entrepreneur in recent years , but she her high profile has not saved SOHO from painful struggles in the country 's shifting property market . SOHO sold a Pudong commercial project and put three Shanghai office buildings up for sale in 2016 as the company continues to fall short of revenue targets following a shift away from its former strategy of building office properties to sell as individual units to small investors .\",\n",
       "       \"But how often should we shave if we no longer have in-between days ? Well , we are all born different and so is our hair . Everyone 's hair grows differently so there is no right or wrong answer as to how often we should shave . Some women do it every day , while others have the privilege of waiting longer in-between shaves . As time goes by , we will slowly no more about our body and eventually find the right schedule that works . <h> 4 ) ' I can shave without water when I 'm in a hurry '\",\n",
       "       'The media was also partly to blame for normalising and glorifying violence among young women , she said . Popular movies , television shows and video games celebrated violent but gorgeous lead female characters , a trend that was continuing in the latest hit novel and film Hunger Games , in which the lead character , 16-year-old Katniss Everdeen , is selected to compete in a televised battle to the death with other teenagers .',\n",
       "       \"Smashing and crashing into other girls on a roller skating rink might not be everyone 's cup of tea , but for a group of Timaru women , the fun of roller derby thrills them to the core .\",\n",
       "       'She said Tanzania has obligations to the Maputo Protocol that ensures women are protected against discrimination , having ratified it in March 2007 .',\n",
       "       'In your letter , you also said that you want to become pregnant . It should be noted that women can develop ovarian cysts at any time from infancy to menopause and the majority of ovarian cysts cause no symptoms at all . In some cases , however , they become so large that they twist the ovaries .',\n",
       "       '\" Out of the 30 samples of women , three tested positive for Stanozolol , Methenedionone and Nandrolone . Out of the 30 samples of men , 1 sample turned positive for Nandrolone and Stanozolol , \" a NADA release said .'],\n",
       "      dtype=object)"
      ]
     },
     "execution_count": 93,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_val_df[train_val_df.keyword == 'women'].text.values[:20]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[  99 3162 7613   98 1855 3688 1021 2431 7037 6783 2634 4045 7493 6948\n",
      " 3867 6679 6321 4374 3571 5265]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([\"The departures from London will barely put a dent in S&amp;P 's overall presence in Europe 's main financial centre . But Peterson warned Britain needed to provide clarity on key post-Brexit regulatory arrangements to ensure there is n't more upheaval .\",\n",
       "       '\" Singh is the most common surname for our drivers across the country . Not all of them are immigrants , but that gives an idea of how many are .',\n",
       "       \"- An extended-hour 's clinic pilot project was launched at the Princess Clinic in Roodepoort , the start of a comprehensive plan to increase access to healthcare for the most vulnerable residents of Johannesburg . The clinic operates a 15 hour day during the week , from 07h00 to 22h00 , as well as from 07h00 to 13h30 on Saturdays ;\",\n",
       "       \"The Golden State is home to the country 's largest number of illegal immigrants -- an estimated 2.4 million -- and officials say members of the community are terrified since the mogul pulled off one of the most improbable presidential wins in modern history on November 8 .\",\n",
       "       'Two elderly mainland Chinese women have been arrested for allegedly setting off firecrackers outside the US Consulate , shortly after they entered Hong Kong . They said they had met with unfair treatment in the mainland , but their motive is unknown . <h> Support independent media <h> Partner with HKFP <h> What is Hong Kong Free Press ?',\n",
       "       '- Get the homeless people shelters ; assist families with health-care plans that will cover a wider range of issues and surgeries ;',\n",
       "       'As her co-workers recently scurried out of the office to get lunch , Phsar Thmey 1 commune chief Si Ton sat by her phone , awaiting a call from a resident in need of her signature .',\n",
       "       '\" ( It ) reflects the government \\'s commitment to addressing the high incidence of crime and violence , which undermines productivity and engenders hopelessness in society , \" he said , adding that \" there can be no development without peace \" .',\n",
       "       \"2018 Commonwealth Games Ghana 's Hor Halutie advances to 100m final <h> Ghanaian sprinter , Hor Halutie , has qualified to Monday 's women 's 100m final at the on-going Commonwealth Games .\",\n",
       "       'As a result , the statute of limitations , which sets the maximum time within which legal proceedings must begin after an occurrence , was abolished for sex crimes against minors and the disabled .',\n",
       "       \"They were outclassed yesterday , hapless amid an onslaught by their hungry opponents but , even in hopeless defeat , the Singapore Slingers vowed to come back stronger and win this year 's Asean Basketball League ( ABL ) title .\",\n",
       "       'Comments have been disabled as the case has appeared before the courts .',\n",
       "       'Among the values Leitch believes Canada should require support for from future immigrants , gender equality has pride of place . Fair enough . But many religions do not treat men and women equally . The Catholic Church for one does not . It denies women access to the priesthood . It frowns on contraception , has long been at the forefront of the fight against abortion rights and opposes same-sex marriage .',\n",
       "       \"The president 's executive orders , bypassing Congress , to protect more than four million undocumented immigrants from deportation , have infuriated Republicans and been blocked by states in the courts .\",\n",
       "       \"Moreover , absent employees are vulnerable to losing their jobs when workplaces are n't aware of family violence -- whether due to physical violence , illness or stress .\",\n",
       "       'PIE replaces the common law action whereby owners can quickly and easily reclaim their property against an unlawful occupier ( what lawyers call the rei vindicatio ) . Ownership in terms of this common law remedy was a very powerful right that trumped other considerations such as the vulnerable circumstances of occupiers and their interest in protecting the space they call home .',\n",
       "       \"Study : Cheetahs sprinting towards extinction <h> The study found that 77% of the animal 's remaining habitat falls outside protected areas , leaving it especially vulnerable to human interference\",\n",
       "       \"Four years ago Sri Lanka 's navy rescued 138 refugees from Bangladesh and Myanmar whose boat had been drifting off Sri Lanka for over 10 days .\",\n",
       "       'A parallel society exists in modern Britain in which a \" sub-current of vulnerable people \" is routinely exploited as modern slaves by ruthless gangs , the National Crime Agency said .',\n",
       "       '\" It \\'s entirely unacceptable for disabled people , their family , friends and carers not to be able to access all high street shops and facilities . \"'],\n",
       "      dtype=object)"
      ]
     },
     "execution_count": 82,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# print a few examples with label score 0 (negative label)\n",
    "text_df = train_val_df[['par_id', 'text', 'labels']]\n",
    "print(text_df[text_df.labels == 0]['par_id'].values[:20])\n",
    "text_df[text_df.labels == 0]['text'].values[:20]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 246 7742 5841   45 6722 3915 2762 7318 1303 7935 1821 1691 5718  610\n",
      " 7598 5261 6026 4416 1717 6285]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array(['What causes someone to become homeless ? Brainstorm as many reasons as you can and check the story for ideas .',\n",
       "       \"I help run a nonprofit called Boulder Food Rescue that picks up food that would otherwise be thrown out from grocery stores and caterers and transports it by bicycle to agencies that serve hungry , homeless and low income folks in Boulder , Colorado , USA . We have 120 hard working volunteers who bike food around town , deliver that food to over 50 recipient organization and in the past year and a half have rescued over 250,000 lbs of fresh produce . Our bike powered model is replicable in nearly every city , and we 've created a step by step guide to how to do so ! Thus far , six different cities have adopted our model and we 're always looking for more dedicated people to do the same . Please visit boulderfoodrescue.org for more information about how to start your own bike powered food rescue and begin addressing waste and want at the same time in your hometown !\",\n",
       "       'Slate : Is the dominance of women right now a good thing ?',\n",
       "       \"Currently , what 's left to the Palestinian people is nothing more than the tiny territory of Gaza Strip , which is a de facto prison of Palestinian refugees , as well as the scattered and isolated villages and cities of the West Bank .\",\n",
       "       'Father Anton Pascual , director of Caritas Manila , said that although Manila Archdiocese has prepared relief goods for victims of calamities , the organization is open to receive more donations. ? The priest said the evacuees are in need of basic emergency items such as food , water , medicine , clothing and hygiene and first aid kits .',\n",
       "       '\" It is worth trying , \" he said , because \" there are women , children , innocent civilians who can get food and medical supplies to get some relief from the terror of constant bombings . \" Citing \" grave differences , \" Obama said that \" it \\'s premature for us to say there \\'s a clear path forward , but there \\'s the possibility , at least for us , to make some progress . \"',\n",
       "       '\" The villagers have been wanting help about what \\'s happened because , of course , this has had a huge impact on them ... Many are disabled and heavily traumatised , particularly the parents of little Fatima . \"',\n",
       "       'The emir said the call became imperative in order to encourage the younger generation especially those from poor families to achieve their educational aspirations .',\n",
       "       'In addition , \" having ICT ( infrastructure ) would be very beneficial for women , because it allows them to work from home , \" said Betti Alisjahbana , founder of ICT services and leadership development company QB International .',\n",
       "       'Rightly so the description of Food Aid as an \" NGO that rescues surplus food from the supply chain and distributes it to people in need \" .',\n",
       "       '\" People are in fear . The government of India is accusing us of being a security threat . So people are feeling helpless and hopeless . We are like birds in a cage - stuck , \" said Mr Ali .',\n",
       "       'EcoPost also provides an income for poor families who collect scrap plastic and other waste . Rutto says about 550 people have gained employment through her recycling efforts : \" Ours is a social enterprise . We \\'re looking at eradicating poverty , not just making money . We have given job opportunities to many youths \"',\n",
       "       'Mr Tuju , the Cabinet Secretary without portfolio , says the government is on top of the humanitarian crisis situation across the country and will ensure those in need receive the help they need .',\n",
       "       '\" This is the first step in building more new homes to help those most in need in our regions . In total , it will build more than 270 bedrooms for individuals or families in the areas where they are needed most .',\n",
       "       \"Through the project the World Bank will provide technical and financial support . The move will enable the world 's poorest and most vulnerable people to gain access to critical services , from education to healthcare and financing , while also advancing their legal and political rights .\",\n",
       "       '\" I will not encourage anybody else to do same . It is not a good song and it can not be classified as a gospel song because Christ came to give even the hopeless hope , \" Noble Nketsiah , who is out with \\' Meto \\' after a long break , added .',\n",
       "       \"One of the many LifeLink supported agencies , The Shopfront , is the focus of the Archbishop 's Christmas Appeal this year . The Shopfront is a drop-in and referral centre offering support and friendship to people experiencing difficulties dealing with homelessness , financial stress , hunger , depression , loneliness , addiction and violence .\",\n",
       "       'We have a judicial system that is intended to uphold the principle of equality of all before law . Yet our common people find it hard to access the system when in need',\n",
       "       'I am sure that you are all aware of the flooding and mudslide disaster taking place in Sri Lanka . As of the time of this writing , 64 people have been reported dead , over 135 people are still missing , and thousands are homeless . We have a long tradition at Dharma Vijaya for coming to the aid of disaster victims : 2004 tsunami , Myanmar typhoon , Nepal earthquake , and others .',\n",
       "       'Hazera Begum still remembers how she had run away from her family in the mid-1970s as an eight-year old . In an attempt to escape from the torment of her stepmother , she initially began her homeless life as a street urchin , @ @'],\n",
       "      dtype=object)"
      ]
     },
     "execution_count": 96,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# print a few examples with label score 1 (negative label)\n",
    "print(text_df[text_df.labels == 1]['par_id'].values[:20])\n",
    "text_df[text_df.labels == 1]['text'].values[:20]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 635 4279 7375 9196 3055 2697 2520 8279 4258 2125 5105 2655 8268  179\n",
      "  545 7770  913 3820 4358 3662]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([\"He would n't tolerate nonsense and he never allowed anybody to intimidate him . He was wealthy and very comfortable , and with his wealth was very generous and kind . In every three months , we had banquet for all the chiefs in my community . After that of chiefs , he would organise ours for us and the children from poor families . He used to tell us that being comfortable did not mean we were special , as God had created everyone equally .\",\n",
       "       'Over the last year , they have helped provide various types of support to numerous struggling individuals and families by matching those wanting to give help to those who need it , from painting and decorating family homes in poor condition , organising days out for underprivileged children and helping the homeless .',\n",
       "       \"As the firms ' names , locations and charges suggest , they are n't best known for their work dealing with marital scraps within poor families , or even prosperous upper-middle-class ones . One of the most high-profile lawyers in the field is Fiona Shackleton , who acted for Prince Charles and Sir Paul McCartney during their divorces . Nicknamed the Steel Magnolia , she is easily identifiable by a head of blonde bouffant hair that was once famously drenched and straightened by a glass of water thrown by Sir Paul 's ex-wife Heather Mills after a bitterly contested court hearing .\",\n",
       "       'The hundreds of families that have gathered at the enormous NRG centre in the southern outskirts of Houston may have a roof over their heads , but their plight is similar to those who left the Middle East for Europe . They are refugees -- people who have lost their homes , their possessions and , in many cases , their livelihoods .',\n",
       "       'DMK working president M K Stalin on Sunday condemned the proposal of the Centre to conduct NEET online through the National Testing Agency . The move is an attempt to destroy the dreams of students from urban poor families and those from rural areas aspiring for medical education .',\n",
       "       \"By God 's grace , the church supports 15 charities ( educational , health and Christian missions ) on a monthly basis . This includes the children 's cancer unit of Korle Bu which has publicly indicated that our monthly donations for the past 5 years , which are the largest they receive , have allowed children from poor families all over the country and West Africa to receive life-saving treatment for free . This is verifiable .\",\n",
       "       'Dole , France -- French baker Michel Flamant , who owes his life to the homeless man who begged for handouts outside his bakery , knows more than anyone that man does not live by bread alone .',\n",
       "       \"Far more important than the implications for the Economy Gods ( is the dollar up or down ? ) last night 's outcome will also mean many , many vulnerable New Zealanders will have a better life over the next three years at least .\",\n",
       "       '\" He must have thought : \\' I do n\\'t want to do this anymore \\' . \" Isaac was a child beaten down and feeling hopeless .',\n",
       "       '\" The government is implementing several schemes that would change the economic position of poor families , \" she added .',\n",
       "       'This movie gets a lot of slack , from a clunky script to horrific accents , but it \\'s an Irish immigration story that is second only to \" Gang of New York . \" It shows the cruelty and injustices that faced Irish immigrants in the cold new world and their struggle for a better life .',\n",
       "       'Relief programmes are being scaled up to meet rising needs , but to save lives and prevent decades of development progress from being irreversibly eroded , more must be done . We need to provide more people with desperately needed cash and food assistance . We need to reach more children with specialised nutritional support to stop greater numbers from becoming malnourished . And we need to expand social safety nets that buffer the impact of this crisis , such as school feeding programmes so children from poor families get at least one good meal a day . The urgency of the situation is irrefutable . But humanitarian aid is not keeping pace with the tremendous needs . Not even close . In fact , there is a staggering funding gap -- and it is a tragedy in the making .',\n",
       "       'Ramirez was only 13 when he was first introduced to the sport and honed his skills at the Tuloy Foundation in Muntinlupa , a haven built through the efforts of Fr . Rocky Evangelista that provides hundreds of underprivileged , abandoned and homeless children the opportunity to learn academics and sports in a safe and caring environment .',\n",
       "       'Jenny Neal , regional director of the Grandmothers Advocacy Network for Saskatchewan and Manitoba , displays her orange scarf at the sculpture Prairie Wind at River Landing , Thursday , November 24 , 2016 , to promote 16 days of orange to raise awareness about violence against women.Greg Pender / The StarPhoenix',\n",
       "       'This year TEAM HOPE shoeboxes from Ireland will go to street children , orphans , children in hospital , poor families , victims of war and other needy situations in Eastern Europe and former Soviet Union e.g. Romania , Kosovo , Albania and Africa e.g. Cameroon , Congo , Rwanda .',\n",
       "       'BEIRUT , Lebanon - Film star and UN special envoy Angelina Jolie said on Wednesday she was moved at how Lebanese families were opening their homes to Syria refugees , after Beirut ruled out setting up camps for them .',\n",
       "       '\" There are people who are struggling to make ends meet and it just seems ironic that in a country like Australia where we are so blessed with so much land and so much fertile country that we \\'re not able to get the food to people in need . \" <h> Feeding Sydney \\'s west',\n",
       "       'AVC : Could you relate to that kind of hopelessness and despair ?',\n",
       "       \"The figures mentioned above are inclusive of men but have decided to single out women- our mothers- as an appreciation of their love to children . They paint a very gloomy picture and calls for every effort to protect women and children all over the world . As you are reading through the lines , right now you have a very touching story to tell about your own mother or your neighbour 's mother .\",\n",
       "       \"I 've been bothered by that moment since then . Why did n't I help him ? I had a few hundred dollars with me and could have spared it . I felt ashamed when I got back home because I knew that my own selfish pride got in my way of helping someone desperately in need .\"],\n",
       "      dtype=object)"
      ]
     },
     "execution_count": 98,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# print a few examples with label score 2 (positive label)\n",
    "print(text_df[text_df.labels == 2]['par_id'].values[:20])\n",
    "text_df[text_df.labels == 2]['text'].values[:20]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['Together with wife Vicky , Cornell established the Chris &amp; Vicky Cornell Foundation , a charity that works with vulnerable children . The charity currently focuses on protecting these children from the ravages of homelessness , poverty , abuse and neglect . <h> 7 . Cuba concert',\n",
       "       '\" There is no one who can tell immigrants they ca n\\'t achieve a success that is rightfully theirs . Even if the goal is n\\'t to get a degree , any other success for immigrants in this country is deserved and waiting for them to be grasped , \" says Jacqueline . \" All it takes is patience and the will to fight and move forward . No doubt in my mind any immigrant has that . \"',\n",
       "       'Discrimination of the disabled by society is one of the major problems undermining the progress of democratic practice in the country . It is always the dream of people with disabilities that so long as the disability bill is passed , their position in society will be influenced positively .',\n",
       "       'An organisation representing Kashmiri Pandits on Saturday demanded that the Centre announce a smart city in Kashmir for the displaced community members who have been living a life of \" refugees in their own country \" .',\n",
       "       'Today , homeless women are still searching for the same thing . A place to sleep and be safe .',\n",
       "       \"Family law aid cut will hit those most in need , says Bar Council <h> ' We have a duty to ensure that access to justice is delivered to all members of society '\",\n",
       "       'The Portuguese superstar has raised money for various causes , including donating more than $83,000 to a 10-year-old fan in need of brain surgery . He is also a global spokesperson for organisations that tackle issues around childhood hunger , obesity and biodiversity , among others .',\n",
       "       'Remarkable too are the ways in which a childhood spent in Pakistan forced one to dispense with fatalistic views of the world and hopeless imaginations of the future . Lifetimes are spent under skies dotted with flying bullets , with threats to kidnap loved ones , and in the thick of crowded markets that unexpectedly explode . Even amid the looming dangers , we knew there were other lives to be lived . And so we went on , fiercely living -- despite the noise in the background .',\n",
       "       'This aspect of the Constitution has required the Court on several occasions to assess policy adopted by the government . In the seminal early case , Government of the RSA and Others v Grootboom and Others , 49 the Court held that the government \\'s housing policy was in breach of the obligations imposed upon government by section 26 of the Constitution in that it failed to \" provide for any form of relief to those desperately in need of access to housing \" 50 and ordered the government to to amend its program \" to provide relief for people who have no access to land , no roof over their heads , and who are living in intolerable conditions or crisis situations . \" 51',\n",
       "       '\" Malaria strikes an astounding number of refugees , but it does n\\'t need to \" said Chris Helfrich , the director of Nothing But Nets . \" A simple bed net can help protect mothers and children from a deadly mosquito bite-and this is the help that they desperately need . We \\'re so grateful to Curry and all our Nothing But Nets Champions for ensuring that we can cover this camp with life-saving bed nets . \"',\n",
       "       '\" She was in a vulnerable , helpless , hopeless situation when the two men entered her home .',\n",
       "       'Our hearts go out to the hundreds of thousands of refugees to be found in Tanzania , other parts of Africa and elsewhere , and we feel compelled to hail the Japanese Government and people for their gesture of goodwill and love that will definitely make a huge difference in the lives of the beneficiaries of the assistance .',\n",
       "       'What we have instead observed has been an ever-escalating education arms race , in which the poorest children have ended up hopelessly ill-equipped to fight . The signs are all around us : the booming billion-pound industry of private tutors paid to boost pupils ? ? ? grades ; the sharp-elbowed tiger parents stopping at nothing to get their children into the best schools , and the stressed-out students trying to navigate a complex and often baffling university admissions system .',\n",
       "       '\" If I die ... I know there is a day that I will die , when that day reaches then I would like to be remembered for helping many children from the poor families to have access to modern and quality education ... that is why I have been moving up and down to ensure that we have good schools that will provide an opportunity for our youth to get a good education ... in these schools we do have Germany teachers and others ( volunteers ) from other developed world so as to ensure quality education is provided to our children , \" he said .',\n",
       "       'When politicians discuss immigration , it is usually in high-flying terms . Jeb Bush says that \" immigrants create an engine of economic prosperity . \" Politicians always talk of importing the best and the brightest from abroad . But New York City \\'s salons capture the tawdry reality of illegal immigration , which creates islands of lawlessness where people can be mistreated with little consequence .',\n",
       "       '\" Because our democracies are inclusive , we are able to welcome people , and refugees in need to our country \\'s and nowhere have we seen that compassion more evident than here in Greece , \" he said after visiting the Acropolis hill in Athens .',\n",
       "       'Machakos Governor Alfred Mutua has said it is a shame for Ukambani residents to continue languishing in the ashes of poverty , neglect and hopelessness despite having certain local leaders in senior government positions in the past .',\n",
       "       'IS first captured the city in 2014 and tens of thousands who fled are badly in need of aid , with charities saying \" camps are bursting at the seams \" .',\n",
       "       '\" Woke up today to see I \\'ve reached 1M followers ! So here \\'s to you guys : donating 1 house to a family in need of a home in our GK Village Ave Maria , Tanay Rizal . My way to say thank you for following me and our way of giving back , \" she wrote .',\n",
       "       'With the assistances of the Shoal Livelihoods Programme through different NGOs , over 1,00,000 poor families living in the areas of these districts have achieved the success to overcome poverty since 2004 .'],\n",
       "      dtype=object)"
      ]
     },
     "execution_count": 65,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# print a few examples with label score 3 (positive label)\n",
    "text_df[text_df.labels == 3]['text'].values[:20]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 3233   925   931   438  3138 10454  1976  6327  9584  2960  7797  1933\n",
      "  8237  3406  4863  1951  2901  4125  2449  3115]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array(['A few kilometres past Eldoret , we drove past hundreds , perhaps thousands , of unwashed , hungry and disheveled refugees milling aimlessly around a small shopping centre , hopelessness and fear etched on their faces . They were running away from death ; running off to nowhere .',\n",
       "       'The pope is urging leaders of the Group of 20 nations to make the poor and refugees a priority of their summit .',\n",
       "       \"KOLKATA : He has never been an officer who sat back and allowed things to remain as they were . When in charge of correctional homes in West Bengal , he gave top priority to human rights and reform and rehabilitation of inmates . Later , when in charge of the Border Security Force ( BSF ) in India 's eastern theatre , BD Sharma not only helped improve ties with Bangladesh 's border guarding force Border Guard Bangladesh ( BGB ) but won the hearts of people in the neighbouring country by ensuring education for an illegal immigrant 's child and finally sending her back home .\",\n",
       "       'Instead of passively paying a sickness benefit for 40 years , for example , we want to take steps to intervene now to help vulnerable New Zealanders get a job , lead a better life , and save the Government money in the long run .',\n",
       "       'The spirit the Rwandese have shown to accommodate the people , almost a tribe in their tiny country with approximately 19 million people with meager resources just like any other third world country , without planning to ditch them in refugee camps , but offer them to live just like any other citizen should actually be emulated in the human race .',\n",
       "       '\" We are challenged , I suggest , to turn this time of celebration into a sustained commitment of awareness and care for those for whom each day is an act of survival , \" said Mr Higgins , thanking those who work to support the vulnerable .',\n",
       "       '\" The American people , as we always have been , are there to partner with African countries to ensure their most vulnerable populations receive life-saving assistance .',\n",
       "       \"And so what is this hustling scam-artist talking about ? The rump-CPP presidential candidate has also been screaming a lot in the name of Ghana 's communities of the disabled and abjectly deprived . Now , let Mr. Greenstreet , who has promised to supply free wheelchairs to the crippled and poor if voted into power , point us to a single substantive or significant project that he has undertaken for the especial benefit of the country 's poor and destitute , besides endlessly running his loud mouth .\",\n",
       "       \"According to Betty-Ann Blaine , executive director of Hear the Children 's Cry , deterioration of family life ; instability/shifting households ; crippling levels of poverty ; lack of adequate social support systems and heartbreaking levels of hopelessness are key factors leading to children running away from homes .\",\n",
       "       'He went on to urge women to learn from that and refrain from being \" all over \" the place .',\n",
       "       \"I 'm sorry to say that the average Nigerian woman is so much afraid of losing the grounds she has gained . This is unlike the men , who , when they climb up , leave the ladder for other men to climb same ladder to come and meet them there . The average Nigerian woman climbs the ladder and uses her left foot to push the ladder down . It 's unfortunate . You know why ? She lacks confidence in herself . Only a few women leaders ever think that once they are up there , no matter how they open doors for others , no other woman will come bring them down . That is the reason why a Nigerian woman today will have so much and will still have friends dying of lack .\",\n",
       "       'At that time , the handful of bhikkhunis was not only illiterate but also homeless and hungry , says Bhikkhuni Lekshe . \" They needed some help . \"',\n",
       "       'Even people who are disabled can still practice karate and have a sense of accomplishment . It matters not what your state of being is , people are encouraged to excel .',\n",
       "       'As Maas put it , \" the loss of this organisation could unleash an uncontrollable chain reaction . \" Kids would be pushed from UNRWA classrooms onto the streets , where they would be more vulnerable to dangerous scenarios such as recruitment efforts by terrorists , who will surely jump at the chance to argue that if we ca n\\'t keep our aid promises , peaceful coexistence with the West is impossible . Child marriage , child labour , and child trafficking would rise . A generation of children and young people would be lost , in a region more unstable than ever .',\n",
       "       'The march was led by eThekwini mayor , Cllr Zandile Gumede . KwaZulu-Natal premier , Willies Mchunu said , \" Real men must stand up and defend the most vulnerable members of society- women and children . We must not stand by and watch this terrible scourge tear apart the fabric of our communities and society . \"',\n",
       "       'Many of them sought refuge in Malaysia only to be treated as outcasts to be whipped , incarcerated , deported , trafficked or abused . Why was no consideration given to allowing the thousands of these refugees to work as there is now for the Rohingya ? Did the fact that they were non-Muslim disqualify them from compassionate treatment ?',\n",
       "       'Our judicial system works , we have a free press , a wonderful , best in the world Constitution , and our Press is free . We must be very grateful for these huge benefits . The blacks want all our farmland without compensation . Give it to them . Let the farmers flock into the cities and make a new life for themselves . With their resilience I am sure it will not be so difficult for them to establish a new , happy and productive life . They will have no money but the clothes on their back to start off with , but that is what so many immigrant Americans had to face . Through guts , determination and sheer will power , they rose above it all , and look what America is today .',\n",
       "       'Yong is not the first Sabahan to find himself on the wrong side of the law , and neither will he be the last . Another boy from Sabah was Lee Siaw Foo , who was convicted of trafficking heroin into Singapore . Lee , too , came from a family facing hard times ; his mother a bankrupt , his father in need of an operation after a heart attack . As the only child and sole breadwinner , he agreed to deliver packages in Malaysia and Singapore . The packages turned out to be heroin . Yong told his lawyer that in 2009 Lee was dragged kicking and screaming from his cell to the gallows .',\n",
       "       'Regardless of your situation , being generous is a good idea . Why ? \" The generous will prosper ; those who refresh others will themselves be refreshed . \" Proverbs 11:25 , NLT . \" And I have been a constant example of how you can help those in need by working hard . You should remember the words of the Lord Jesus : \\' It is more blessed to give than to receive . \\' \" Acts 20:35 , NLT',\n",
       "       'While the future of the peace process remains uncertain , the hard fact that the government must tackle now is to not just take back the \" failed \" refugees but also ensure that they are not devoured by the same demon they had fled from in the first instance .'],\n",
       "      dtype=object)"
      ]
     },
     "execution_count": 80,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# print a few examples with label score 4 (positive label)\n",
    "print(text_df[text_df.labels == 4]['par_id'].values[:20])\n",
    "text_df[text_df.labels == 4]['text'].values[:20]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "label_score\n",
       "0    6825\n",
       "1     756\n",
       "3     369\n",
       "4     299\n",
       "2     126\n",
       "Name: count, dtype: int64"
      ]
     },
     "execution_count": 102,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_val_df.labels.value_counts()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "venv",
   "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": "3.10.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}