notebooks/03-classify-documents.ipynb

{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Classify Documents"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "\n",
    "from IPython.core.interactiveshell import InteractiveShell\n",
    "InteractiveShell.ast_node_interactivity = \"all\""
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "\n",
    "df_orig = pd.read_csv(\"../data/bbc-news-data-embedding.csv\", delimiter='\\t', index_col=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "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>category</th>\n",
       "      <th>filename</th>\n",
       "      <th>title</th>\n",
       "      <th>content</th>\n",
       "      <th>embedding</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>business</td>\n",
       "      <td>001.txt</td>\n",
       "      <td>Ad sales boost Time Warner profit</td>\n",
       "      <td>Quarterly profits at US media giant TimeWarne...</td>\n",
       "      <td>[-0.0012276918860152364, 0.00733763724565506, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>business</td>\n",
       "      <td>002.txt</td>\n",
       "      <td>Dollar gains on Greenspan speech</td>\n",
       "      <td>The dollar has hit its highest level against ...</td>\n",
       "      <td>[0.0009311728645116091, 0.014099937863647938, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>business</td>\n",
       "      <td>003.txt</td>\n",
       "      <td>Yukos unit buyer faces loan claim</td>\n",
       "      <td>The owners of embattled Russian oil giant Yuk...</td>\n",
       "      <td>[-0.010487922467291355, 0.009665092453360558, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>business</td>\n",
       "      <td>004.txt</td>\n",
       "      <td>High fuel prices hit BA's profits</td>\n",
       "      <td>British Airways has blamed high fuel prices f...</td>\n",
       "      <td>[0.0111119095236063, 0.004624682944267988, -0....</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>business</td>\n",
       "      <td>005.txt</td>\n",
       "      <td>Pernod takeover talk lifts Domecq</td>\n",
       "      <td>Shares in UK drinks and food firm Allied Dome...</td>\n",
       "      <td>[-0.0021637482568621635, 0.005410161800682545,...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2220</th>\n",
       "      <td>tech</td>\n",
       "      <td>397.txt</td>\n",
       "      <td>BT program to beat dialler scams</td>\n",
       "      <td>BT is introducing two initiatives to help bea...</td>\n",
       "      <td>[0.007671569474041462, 0.00624304823577404, -0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2221</th>\n",
       "      <td>tech</td>\n",
       "      <td>398.txt</td>\n",
       "      <td>Spam e-mails tempt net shoppers</td>\n",
       "      <td>Computer users across the world continue to i...</td>\n",
       "      <td>[0.0026338498573750257, 0.015989987179636955, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2222</th>\n",
       "      <td>tech</td>\n",
       "      <td>399.txt</td>\n",
       "      <td>Be careful how you code</td>\n",
       "      <td>A new European directive could put software w...</td>\n",
       "      <td>[0.007126151118427515, 0.008495588786900043, -...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2223</th>\n",
       "      <td>tech</td>\n",
       "      <td>400.txt</td>\n",
       "      <td>US cyber security chief resigns</td>\n",
       "      <td>The man making sure US computer networks are ...</td>\n",
       "      <td>[0.002447678940370679, 0.006076449993997812, -...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2224</th>\n",
       "      <td>tech</td>\n",
       "      <td>401.txt</td>\n",
       "      <td>Losing yourself in online gaming</td>\n",
       "      <td>Online role playing games are time-consuming,...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2225 rows × 5 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      category filename                              title  \\\n",
       "0     business  001.txt  Ad sales boost Time Warner profit   \n",
       "1     business  002.txt   Dollar gains on Greenspan speech   \n",
       "2     business  003.txt  Yukos unit buyer faces loan claim   \n",
       "3     business  004.txt  High fuel prices hit BA's profits   \n",
       "4     business  005.txt  Pernod takeover talk lifts Domecq   \n",
       "...        ...      ...                                ...   \n",
       "2220      tech  397.txt   BT program to beat dialler scams   \n",
       "2221      tech  398.txt    Spam e-mails tempt net shoppers   \n",
       "2222      tech  399.txt            Be careful how you code   \n",
       "2223      tech  400.txt    US cyber security chief resigns   \n",
       "2224      tech  401.txt   Losing yourself in online gaming   \n",
       "\n",
       "                                                content  \\\n",
       "0      Quarterly profits at US media giant TimeWarne...   \n",
       "1      The dollar has hit its highest level against ...   \n",
       "2      The owners of embattled Russian oil giant Yuk...   \n",
       "3      British Airways has blamed high fuel prices f...   \n",
       "4      Shares in UK drinks and food firm Allied Dome...   \n",
       "...                                                 ...   \n",
       "2220   BT is introducing two initiatives to help bea...   \n",
       "2221   Computer users across the world continue to i...   \n",
       "2222   A new European directive could put software w...   \n",
       "2223   The man making sure US computer networks are ...   \n",
       "2224   Online role playing games are time-consuming,...   \n",
       "\n",
       "                                              embedding  \n",
       "0     [-0.0012276918860152364, 0.00733763724565506, ...  \n",
       "1     [0.0009311728645116091, 0.014099937863647938, ...  \n",
       "2     [-0.010487922467291355, 0.009665092453360558, ...  \n",
       "3     [0.0111119095236063, 0.004624682944267988, -0....  \n",
       "4     [-0.0021637482568621635, 0.005410161800682545,...  \n",
       "...                                                 ...  \n",
       "2220  [0.007671569474041462, 0.00624304823577404, -0...  \n",
       "2221  [0.0026338498573750257, 0.015989987179636955, ...  \n",
       "2222  [0.007126151118427515, 0.008495588786900043, -...  \n",
       "2223  [0.002447678940370679, 0.006076449993997812, -...  \n",
       "2224                                                NaN  \n",
       "\n",
       "[2225 rows x 5 columns]"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = df_orig.copy()\n",
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2217"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# drop rows with NaN\n",
    "df.dropna(inplace=True)\n",
    "len(df)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Classify documents with their embeddings\n",
    "ref: https://github.com/openai/openai-cookbook/blob/main/examples/Classification_using_embeddings.ipynb"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 2217 entries, 0 to 2223\n",
      "Data columns (total 5 columns):\n",
      " #   Column     Non-Null Count  Dtype \n",
      "---  ------     --------------  ----- \n",
      " 0   category   2217 non-null   object\n",
      " 1   filename   2217 non-null   object\n",
      " 2   title      2217 non-null   object\n",
      " 3   content    2217 non-null   object\n",
      " 4   embedding  2217 non-null   object\n",
      "dtypes: object(5)\n",
      "memory usage: 103.9+ KB\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import classification_report\n",
    "\n",
    "# convert string to array\n",
    "df[\"embedding\"] = df['embedding'].apply(eval).apply(np.array)  \n",
    "\n",
    "df.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = df['embedding']\n",
    "y = df['category']\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.33, random_state=42, shuffle=True, stratify=y) \n",
    "\n",
    "# reshape X into 2D array\n",
    "X_train = np.vstack(X_train)\n",
    "X_test = np.vstack(X_test)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Fit a model with `XGBoost`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/anaconda/envs/azureml_py38/lib/python3.8/site-packages/xgboost/sklearn.py:888: UserWarning: The use of label encoder in XGBClassifier is deprecated and will be removed in a future release. To remove this warning, do the following: 1) Pass option use_label_encoder=False when constructing XGBClassifier object; and 2) Encode your labels (y) as integers starting with 0, i.e. 0, 1, 2, ..., [num_class - 1].\n",
      "  warnings.warn(label_encoder_deprecation_msg, UserWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[13:27:07] WARNING: /mnt/xgboost/src/learner.cc:1061: Starting in XGBoost 1.3.0, the default evaluation metric used with the objective 'multi:softprob' was changed from 'merror' to 'mlogloss'. Explicitly set eval_metric if you'd like to restore the old behavior.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "XGBClassifier(base_score=0.5, booster='gbtree', colsample_bylevel=1,\n",
       "              colsample_bynode=1, colsample_bytree=1, gamma=0, gpu_id=-1,\n",
       "              importance_type='gain', interaction_constraints='',\n",
       "              learning_rate=1, max_delta_step=0, max_depth=5,\n",
       "              min_child_weight=1, missing=nan, monotone_constraints='()',\n",
       "              n_estimators=100, n_jobs=2, num_parallel_tree=1,\n",
       "              objective='multi:softprob', random_state=0, reg_alpha=0,\n",
       "              reg_lambda=1, scale_pos_weight=None, subsample=1,\n",
       "              tree_method='exact', use_label_encoder=True,\n",
       "              validate_parameters=1, verbosity=None)"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "               precision    recall  f1-score   support\n",
      "\n",
      "     business       0.92      0.92      0.92       168\n",
      "entertainment       0.94      0.93      0.93       127\n",
      "     politics       0.90      0.90      0.90       137\n",
      "        sport       0.98      0.98      0.98       168\n",
      "         tech       0.92      0.93      0.93       132\n",
      "\n",
      "     accuracy                           0.93       732\n",
      "    macro avg       0.93      0.93      0.93       732\n",
      " weighted avg       0.93      0.93      0.93       732\n",
      "\n"
     ]
    }
   ],
   "source": [
    "from xgboost import XGBClassifier\n",
    "import pickle\n",
    "\n",
    "TRAIN = False\n",
    "LOAD = True\n",
    "\n",
    "# filename for trained model\n",
    "fname = '../output/models/xgb.pkl'\n",
    "\n",
    "if TRAIN: \n",
    "    # create model instance\n",
    "    xgb = XGBClassifier(n_estimators=100, max_depth=5, learning_rate=1, objective='multi:softprob')\n",
    "    # fit model\n",
    "    xgb.fit(X_train, y_train)\n",
    "\n",
    "    # save the model to disk\n",
    "    with open(fname, \"wb\") as f:\n",
    "        pickle.dump(xgb, f)\n",
    " \n",
    "# load the model from disk\n",
    "if LOAD:\n",
    "    with open(fname, \"rb\") as f:\n",
    "        xgb = pickle.load(f)\n",
    "\n",
    "# predict\n",
    "preds = xgb.predict(X_test)\n",
    "probas = xgb.predict_proba(X_test)\n",
    "\n",
    "# report\n",
    "report = classification_report(y_test, preds)\n",
    "print(report)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.metrics import confusion_matrix\n",
    "from sklearn.metrics import ConfusionMatrixDisplay\n",
    "\n",
    "# confusion matrix\n",
    "cm = confusion_matrix(y_test, preds)\n",
    "cm_display = ConfusionMatrixDisplay(confusion_matrix=cm, display_labels=xgb.classes_).plot()\n"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Fit a model with `RandomForest`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "               precision    recall  f1-score   support\n",
      "\n",
      "     business       0.93      0.97      0.95       168\n",
      "entertainment       0.98      0.94      0.96       127\n",
      "     politics       0.97      0.94      0.96       137\n",
      "        sport       1.00      1.00      1.00       168\n",
      "         tech       0.96      0.97      0.96       132\n",
      "\n",
      "     accuracy                           0.97       732\n",
      "    macro avg       0.97      0.96      0.96       732\n",
      " weighted avg       0.97      0.97      0.97       732\n",
      "\n"
     ]
    }
   ],
   "source": [
    "TRAIN = False\n",
    "LOAD = True\n",
    "# filename for trained model\n",
    "fname = '../output/models/rf.pkl'\n",
    "\n",
    "# fit model\n",
    "if TRAIN: \n",
    "    # train random forest classifier\n",
    "    rf = RandomForestClassifier(n_estimators=100)\n",
    "    rf.fit(X_train, y_train)\n",
    "\n",
    "    # save the model to disk\n",
    "    with open(fname, \"wb\") as f:\n",
    "        pickle.dump(rf, f)\n",
    " \n",
    "# load the model from disk\n",
    "if LOAD:\n",
    "    with open(fname, \"rb\") as f:\n",
    "        rf = pickle.load(f)\n",
    "\n",
    "# predict\n",
    "rf_preds = rf.predict(X_test)\n",
    "rf_probas = rf.predict_proba(X_test)\n",
    "\n",
    "# report\n",
    "rf_report = classification_report(y_test, rf_preds)\n",
    "print(rf_report)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Confusion matrix\n",
    "rf_cm = confusion_matrix(y_test, rf_preds)\n",
    "rf_cm_display = ConfusionMatrixDisplay(confusion_matrix=rf_cm, display_labels=rf.classes_).plot()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Save prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "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>category</th>\n",
       "      <th>prediction</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>252</th>\n",
       "      <td>business</td>\n",
       "      <td>business</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1414</th>\n",
       "      <td>sport</td>\n",
       "      <td>sport</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>567</th>\n",
       "      <td>entertainment</td>\n",
       "      <td>entertainment</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1460</th>\n",
       "      <td>sport</td>\n",
       "      <td>sport</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>775</th>\n",
       "      <td>entertainment</td>\n",
       "      <td>tech</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>421</th>\n",
       "      <td>business</td>\n",
       "      <td>business</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1987</th>\n",
       "      <td>tech</td>\n",
       "      <td>tech</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>273</th>\n",
       "      <td>business</td>\n",
       "      <td>business</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1047</th>\n",
       "      <td>politics</td>\n",
       "      <td>politics</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2135</th>\n",
       "      <td>tech</td>\n",
       "      <td>tech</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>732 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "           category     prediction\n",
       "252        business       business\n",
       "1414          sport          sport\n",
       "567   entertainment  entertainment\n",
       "1460          sport          sport\n",
       "775   entertainment           tech\n",
       "...             ...            ...\n",
       "421        business       business\n",
       "1987           tech           tech\n",
       "273        business       business\n",
       "1047       politics       politics\n",
       "2135           tech           tech\n",
       "\n",
       "[732 rows x 2 columns]"
      ]
     },
     "execution_count": 102,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_test_pred = pd.DataFrame()\n",
    "y_test_pred['category'] = y_test\n",
    "y_test_pred['prediction'] = preds\n",
    "y_test_pred"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(732, 6)"
      ]
     },
     "execution_count": 101,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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>category</th>\n",
       "      <th>filename</th>\n",
       "      <th>title</th>\n",
       "      <th>content</th>\n",
       "      <th>embedding</th>\n",
       "      <th>prediction</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>business</td>\n",
       "      <td>002.txt</td>\n",
       "      <td>Dollar gains on Greenspan speech</td>\n",
       "      <td>The dollar has hit its highest level against ...</td>\n",
       "      <td>[0.0009311728645116091, 0.014099937863647938, ...</td>\n",
       "      <td>business</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>business</td>\n",
       "      <td>005.txt</td>\n",
       "      <td>Pernod takeover talk lifts Domecq</td>\n",
       "      <td>Shares in UK drinks and food firm Allied Dome...</td>\n",
       "      <td>[-0.0021637482568621635, 0.005410161800682545,...</td>\n",
       "      <td>business</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>business</td>\n",
       "      <td>009.txt</td>\n",
       "      <td>Ethiopia's crop production up 24%</td>\n",
       "      <td>Ethiopia produced 14.27 million tonnes of cro...</td>\n",
       "      <td>[0.008323440328240395, 0.014594526030123234, -...</td>\n",
       "      <td>business</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>business</td>\n",
       "      <td>014.txt</td>\n",
       "      <td>Telegraph newspapers axe 90 jobs</td>\n",
       "      <td>The Daily and Sunday Telegraph newspapers are...</td>\n",
       "      <td>[-0.001993480371311307, 0.0038063114043325186,...</td>\n",
       "      <td>business</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>business</td>\n",
       "      <td>015.txt</td>\n",
       "      <td>Air passengers win new EU rights</td>\n",
       "      <td>Air passengers who are unable to board their ...</td>\n",
       "      <td>[0.01202255580574274, 0.0024986255448311567, -...</td>\n",
       "      <td>business</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",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2209</th>\n",
       "      <td>tech</td>\n",
       "      <td>386.txt</td>\n",
       "      <td>Video phones act as dating tools</td>\n",
       "      <td>Technologies, from e-mail, to net chatrooms, ...</td>\n",
       "      <td>[-0.0021125166676938534, -0.003049303079023957...</td>\n",
       "      <td>tech</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2212</th>\n",
       "      <td>tech</td>\n",
       "      <td>389.txt</td>\n",
       "      <td>Mobile multimedia slow to catch on</td>\n",
       "      <td>There is no doubt that mobile phones sporting...</td>\n",
       "      <td>[0.007488378323614597, 0.007830777205526829, -...</td>\n",
       "      <td>tech</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2213</th>\n",
       "      <td>tech</td>\n",
       "      <td>390.txt</td>\n",
       "      <td>Anti-spam laws bite spammer hard</td>\n",
       "      <td>The net's self-declared spam king is seeking ...</td>\n",
       "      <td>[-0.002103055827319622, 0.00393986189737916, -...</td>\n",
       "      <td>tech</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2216</th>\n",
       "      <td>tech</td>\n",
       "      <td>393.txt</td>\n",
       "      <td>Savvy searchers fail to spot ads</td>\n",
       "      <td>Internet search engine users are an odd mix o...</td>\n",
       "      <td>[0.00633224518969655, 0.007613884285092354, -0...</td>\n",
       "      <td>tech</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2222</th>\n",
       "      <td>tech</td>\n",
       "      <td>399.txt</td>\n",
       "      <td>Be careful how you code</td>\n",
       "      <td>A new European directive could put software w...</td>\n",
       "      <td>[0.007126151118427515, 0.008495588786900043, -...</td>\n",
       "      <td>tech</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>732 rows × 6 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      category filename                               title  \\\n",
       "1     business  002.txt    Dollar gains on Greenspan speech   \n",
       "4     business  005.txt   Pernod takeover talk lifts Domecq   \n",
       "8     business  009.txt   Ethiopia's crop production up 24%   \n",
       "13    business  014.txt    Telegraph newspapers axe 90 jobs   \n",
       "14    business  015.txt    Air passengers win new EU rights   \n",
       "...        ...      ...                                 ...   \n",
       "2209      tech  386.txt    Video phones act as dating tools   \n",
       "2212      tech  389.txt  Mobile multimedia slow to catch on   \n",
       "2213      tech  390.txt    Anti-spam laws bite spammer hard   \n",
       "2216      tech  393.txt    Savvy searchers fail to spot ads   \n",
       "2222      tech  399.txt             Be careful how you code   \n",
       "\n",
       "                                                content  \\\n",
       "1      The dollar has hit its highest level against ...   \n",
       "4      Shares in UK drinks and food firm Allied Dome...   \n",
       "8      Ethiopia produced 14.27 million tonnes of cro...   \n",
       "13     The Daily and Sunday Telegraph newspapers are...   \n",
       "14     Air passengers who are unable to board their ...   \n",
       "...                                                 ...   \n",
       "2209   Technologies, from e-mail, to net chatrooms, ...   \n",
       "2212   There is no doubt that mobile phones sporting...   \n",
       "2213   The net's self-declared spam king is seeking ...   \n",
       "2216   Internet search engine users are an odd mix o...   \n",
       "2222   A new European directive could put software w...   \n",
       "\n",
       "                                              embedding prediction  \n",
       "1     [0.0009311728645116091, 0.014099937863647938, ...   business  \n",
       "4     [-0.0021637482568621635, 0.005410161800682545,...   business  \n",
       "8     [0.008323440328240395, 0.014594526030123234, -...   business  \n",
       "13    [-0.001993480371311307, 0.0038063114043325186,...   business  \n",
       "14    [0.01202255580574274, 0.0024986255448311567, -...   business  \n",
       "...                                                 ...        ...  \n",
       "2209  [-0.0021125166676938534, -0.003049303079023957...       tech  \n",
       "2212  [0.007488378323614597, 0.007830777205526829, -...       tech  \n",
       "2213  [-0.002103055827319622, 0.00393986189737916, -...       tech  \n",
       "2216  [0.00633224518969655, 0.007613884285092354, -0...       tech  \n",
       "2222  [0.007126151118427515, 0.008495588786900043, -...       tech  \n",
       "\n",
       "[732 rows x 6 columns]"
      ]
     },
     "execution_count": 101,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_test_result = pd.concat([df, y_test_pred.drop(columns='category')], axis=1, join=\"inner\")\n",
    "df_test_result.shape\n",
    "df_test_result\n",
    "\n",
    "# Save results\n",
    "SAVE = False\n",
    "fname = '../output/predictions.csv'\n",
    "if SAVE:\n",
    "    df_test_result.to_csv(fname, sep='\\t')"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Wrong Predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "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>category</th>\n",
       "      <th>filename</th>\n",
       "      <th>title</th>\n",
       "      <th>content</th>\n",
       "      <th>embedding</th>\n",
       "      <th>prediction</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>105</th>\n",
       "      <td>business</td>\n",
       "      <td>106.txt</td>\n",
       "      <td>Golden rule 'intact' says ex-aide</td>\n",
       "      <td>Chancellor Gordon Brown will meet his golden ...</td>\n",
       "      <td>[0.006964212749153376, 0.0015263919485732913, ...</td>\n",
       "      <td>politics</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>212</th>\n",
       "      <td>business</td>\n",
       "      <td>213.txt</td>\n",
       "      <td>US bank boss hails 'genius' Smith</td>\n",
       "      <td>US Federal Reserve chairman Alan Greenspan ha...</td>\n",
       "      <td>[0.0021459858398884535, 0.0036910325288772583,...</td>\n",
       "      <td>entertainment</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>226</th>\n",
       "      <td>business</td>\n",
       "      <td>227.txt</td>\n",
       "      <td>Call to save manufacturing jobs</td>\n",
       "      <td>The Trades Union Congress (TUC) is calling on...</td>\n",
       "      <td>[0.0012122669722884893, 0.0076951221562922, -0...</td>\n",
       "      <td>politics</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>235</th>\n",
       "      <td>business</td>\n",
       "      <td>236.txt</td>\n",
       "      <td>Card fraudsters 'targeting web'</td>\n",
       "      <td>New safeguards on credit and debit card payme...</td>\n",
       "      <td>[7.788832590449601e-05, 0.016654228791594505, ...</td>\n",
       "      <td>tech</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>240</th>\n",
       "      <td>business</td>\n",
       "      <td>241.txt</td>\n",
       "      <td>G7 backs Africa debt relief plan</td>\n",
       "      <td>G7 finance ministers have backed plans to wri...</td>\n",
       "      <td>[-0.0027972061652690172, 0.013444329611957073,...</td>\n",
       "      <td>politics</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>241</th>\n",
       "      <td>business</td>\n",
       "      <td>242.txt</td>\n",
       "      <td>Q&amp;A: Malcolm Glazer and Man Utd</td>\n",
       "      <td>The battle for control of Manchester United h...</td>\n",
       "      <td>[0.002724364399909973, 0.004842962138354778, -...</td>\n",
       "      <td>sport</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>254</th>\n",
       "      <td>business</td>\n",
       "      <td>255.txt</td>\n",
       "      <td>Call to overhaul UK state pension</td>\n",
       "      <td>The UK pension system has been branded inadeq...</td>\n",
       "      <td>[0.004907403141260147, 0.0119895339012146, -0....</td>\n",
       "      <td>politics</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>292</th>\n",
       "      <td>business</td>\n",
       "      <td>293.txt</td>\n",
       "      <td>Glazer makes new Man Utd approach</td>\n",
       "      <td>Malcolm Glazer has made a fresh approach to b...</td>\n",
       "      <td>[-0.0023372862488031387, 0.006699692457914352,...</td>\n",
       "      <td>sport</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>387</th>\n",
       "      <td>business</td>\n",
       "      <td>388.txt</td>\n",
       "      <td>Venezuela identifies 'idle' farms</td>\n",
       "      <td>Venezuelan authorities have identified more t...</td>\n",
       "      <td>[0.006353371310979128, 0.014112972654402256, -...</td>\n",
       "      <td>politics</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>412</th>\n",
       "      <td>business</td>\n",
       "      <td>413.txt</td>\n",
       "      <td>EC calls truce in deficit battle</td>\n",
       "      <td>The European Commission (EC) has called a tru...</td>\n",
       "      <td>[0.003549534361809492, 0.005724851507693529, -...</td>\n",
       "      <td>politics</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>451</th>\n",
       "      <td>business</td>\n",
       "      <td>452.txt</td>\n",
       "      <td>McDonald's boss Bell dies aged 44</td>\n",
       "      <td>Charlie Bell, the straight-talking former hea...</td>\n",
       "      <td>[0.004609732422977686, 0.006802281830459833, -...</td>\n",
       "      <td>entertainment</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>456</th>\n",
       "      <td>business</td>\n",
       "      <td>457.txt</td>\n",
       "      <td>Germany calls for EU reform</td>\n",
       "      <td>German Chancellor Gerhard Schroeder has calle...</td>\n",
       "      <td>[-0.0018439812120050192, 0.009731155820190907,...</td>\n",
       "      <td>politics</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>459</th>\n",
       "      <td>business</td>\n",
       "      <td>460.txt</td>\n",
       "      <td>Jobs go at Oracle after takeover</td>\n",
       "      <td>Oracle has announced it is cutting about 5,00...</td>\n",
       "      <td>[0.0036041021812707186, 0.009482146240770817, ...</td>\n",
       "      <td>tech</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>518</th>\n",
       "      <td>entertainment</td>\n",
       "      <td>009.txt</td>\n",
       "      <td>Da Vinci Code is 'lousy history'</td>\n",
       "      <td>The plot of an international bestseller that ...</td>\n",
       "      <td>[-0.002073341980576515, 0.007545760832726955, ...</td>\n",
       "      <td>politics</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>657</th>\n",
       "      <td>entertainment</td>\n",
       "      <td>148.txt</td>\n",
       "      <td>Pupils to get anti-piracy lessons</td>\n",
       "      <td>Lessons on music piracy and copyright issues ...</td>\n",
       "      <td>[-0.004451509099453688, 0.014561662450432777, ...</td>\n",
       "      <td>tech</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>682</th>\n",
       "      <td>entertainment</td>\n",
       "      <td>173.txt</td>\n",
       "      <td>UK TV channel rapped for CSI ad</td>\n",
       "      <td>TV channel Five has been criticised for sendi...</td>\n",
       "      <td>[-0.0036573493853211403, 0.015478764660656452,...</td>\n",
       "      <td>tech</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>692</th>\n",
       "      <td>entertainment</td>\n",
       "      <td>183.txt</td>\n",
       "      <td>Star Trek fans fight to save show</td>\n",
       "      <td>Star Trek fans have taken out a full-page ad ...</td>\n",
       "      <td>[0.007987278513610363, 0.014516561292111874, -...</td>\n",
       "      <td>tech</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>711</th>\n",
       "      <td>entertainment</td>\n",
       "      <td>202.txt</td>\n",
       "      <td>BBC 'should allow more scrutiny'</td>\n",
       "      <td>MPs have urged the BBC to give watchdogs more...</td>\n",
       "      <td>[0.0011142775183543563, 0.010890933685004711, ...</td>\n",
       "      <td>politics</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>775</th>\n",
       "      <td>entertainment</td>\n",
       "      <td>266.txt</td>\n",
       "      <td>Help for indies in download sales</td>\n",
       "      <td>A campaign has been launched to help independ...</td>\n",
       "      <td>[-0.0023760353215038776, 0.008679348975419998,...</td>\n",
       "      <td>tech</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>822</th>\n",
       "      <td>entertainment</td>\n",
       "      <td>313.txt</td>\n",
       "      <td>BBC to pour £9m into new comedy</td>\n",
       "      <td>The BBC is to invest £9m in developing new co...</td>\n",
       "      <td>[-0.0013941482175141573, 0.012636896222829819,...</td>\n",
       "      <td>business</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>851</th>\n",
       "      <td>entertainment</td>\n",
       "      <td>342.txt</td>\n",
       "      <td>US composer recreates Bach score</td>\n",
       "      <td>A US musicologist has recreated a lost musica...</td>\n",
       "      <td>[-0.011926768347620964, 0.011977385729551315, ...</td>\n",
       "      <td>tech</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>893</th>\n",
       "      <td>entertainment</td>\n",
       "      <td>384.txt</td>\n",
       "      <td>Robots march to US cinema summit</td>\n",
       "      <td>Animated movie Robots has opened at the top o...</td>\n",
       "      <td>[0.007351818960160017, 0.016657356172800064, -...</td>\n",
       "      <td>tech</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>899</th>\n",
       "      <td>politics</td>\n",
       "      <td>004.txt</td>\n",
       "      <td>Labour chooses Manchester</td>\n",
       "      <td>The Labour Party will hold its 2006 autumn co...</td>\n",
       "      <td>[0.007821072824299335, 0.00648058857768774, 0....</td>\n",
       "      <td>business</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>901</th>\n",
       "      <td>politics</td>\n",
       "      <td>006.txt</td>\n",
       "      <td>'Errors' doomed first Dome sale</td>\n",
       "      <td>The initial attempt to sell the Millennium Do...</td>\n",
       "      <td>[0.0006757106748409569, 0.007997625507414341, ...</td>\n",
       "      <td>business</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>918</th>\n",
       "      <td>politics</td>\n",
       "      <td>023.txt</td>\n",
       "      <td>E-University 'disgraceful waste'</td>\n",
       "      <td>A failed government scheme to offer UK univer...</td>\n",
       "      <td>[-7.362403266597539e-05, 0.006662579253315926,...</td>\n",
       "      <td>tech</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>971</th>\n",
       "      <td>politics</td>\n",
       "      <td>076.txt</td>\n",
       "      <td>Report attacks defence spending</td>\n",
       "      <td>The Ministry of Defence has been criticised o...</td>\n",
       "      <td>[0.0015996992588043213, 0.007126652169972658, ...</td>\n",
       "      <td>business</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>985</th>\n",
       "      <td>politics</td>\n",
       "      <td>090.txt</td>\n",
       "      <td>MSPs hear renewed climate warning</td>\n",
       "      <td>Climate change could be completely out of con...</td>\n",
       "      <td>[0.012153446674346924, 0.007975501008331776, -...</td>\n",
       "      <td>business</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1004</th>\n",
       "      <td>politics</td>\n",
       "      <td>109.txt</td>\n",
       "      <td>UK firms 'embracing e-commerce'</td>\n",
       "      <td>UK firms are embracing internet trading oppor...</td>\n",
       "      <td>[-0.0006202863878570497, 0.004257924854755402,...</td>\n",
       "      <td>business</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1008</th>\n",
       "      <td>politics</td>\n",
       "      <td>113.txt</td>\n",
       "      <td>MP's shock at long lost daughter</td>\n",
       "      <td>Labour MP Stephen Pound has told of his shock...</td>\n",
       "      <td>[0.002527164528146386, 0.006853112950921059, -...</td>\n",
       "      <td>entertainment</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1034</th>\n",
       "      <td>politics</td>\n",
       "      <td>139.txt</td>\n",
       "      <td>Straw praises Kashmir moves</td>\n",
       "      <td>The UK has welcomed the decision by India and...</td>\n",
       "      <td>[0.00722165871411562, 0.006721210666000843, -0...</td>\n",
       "      <td>business</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1035</th>\n",
       "      <td>politics</td>\n",
       "      <td>140.txt</td>\n",
       "      <td>UK helps raped Rwandan women</td>\n",
       "      <td>Britain is to give a £4m grant to help women ...</td>\n",
       "      <td>[0.004300092346966267, 0.013324585743248463, -...</td>\n",
       "      <td>business</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1073</th>\n",
       "      <td>politics</td>\n",
       "      <td>178.txt</td>\n",
       "      <td>Hatfield executives go on trial</td>\n",
       "      <td>Engineering firm Balfour Beatty and five rail...</td>\n",
       "      <td>[0.0065396688878536224, 0.009236570447683334, ...</td>\n",
       "      <td>entertainment</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1089</th>\n",
       "      <td>politics</td>\n",
       "      <td>194.txt</td>\n",
       "      <td>Cherie accused of attacking Bush</td>\n",
       "      <td>Cherie Blair has been accused of criticising ...</td>\n",
       "      <td>[0.004710556473582983, 0.0016932026483118534, ...</td>\n",
       "      <td>entertainment</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1225</th>\n",
       "      <td>politics</td>\n",
       "      <td>330.txt</td>\n",
       "      <td>Manchester wins Labour conference</td>\n",
       "      <td>The Labour Party will hold its 2006 autumn co...</td>\n",
       "      <td>[0.007821072824299335, 0.00648058857768774, 0....</td>\n",
       "      <td>business</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1288</th>\n",
       "      <td>politics</td>\n",
       "      <td>393.txt</td>\n",
       "      <td>Parliament's record of scandal</td>\n",
       "      <td>In a locked room at the heart of Parliament t...</td>\n",
       "      <td>[-0.004831457510590553, 0.009346305392682552, ...</td>\n",
       "      <td>tech</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1308</th>\n",
       "      <td>politics</td>\n",
       "      <td>413.txt</td>\n",
       "      <td>Sainsbury's Labour election gift</td>\n",
       "      <td>Science Minister Lord Sainsbury has made a £2...</td>\n",
       "      <td>[0.002463462296873331, 0.007179685402661562, -...</td>\n",
       "      <td>business</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1316</th>\n",
       "      <td>sport</td>\n",
       "      <td>004.txt</td>\n",
       "      <td>IAAF launches fight against drugs</td>\n",
       "      <td>The IAAF - athletics' world governing body - ...</td>\n",
       "      <td>[0.009765633381903172, 0.01343011949211359, -0...</td>\n",
       "      <td>business</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1339</th>\n",
       "      <td>sport</td>\n",
       "      <td>027.txt</td>\n",
       "      <td>Holmes feted with further honour</td>\n",
       "      <td>Double Olympic champion Kelly Holmes has been...</td>\n",
       "      <td>[0.010477458126842976, 0.0017678732983767986, ...</td>\n",
       "      <td>entertainment</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1443</th>\n",
       "      <td>sport</td>\n",
       "      <td>131.txt</td>\n",
       "      <td>Newry to fight cup exit in courts</td>\n",
       "      <td>Newry City are expected to discuss legal aven...</td>\n",
       "      <td>[-0.007764059584587812, 0.0037425938062369823,...</td>\n",
       "      <td>politics</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1675</th>\n",
       "      <td>sport</td>\n",
       "      <td>363.txt</td>\n",
       "      <td>Calder fears for Scottish rugby</td>\n",
       "      <td>Former Scotland international Finlay Calder f...</td>\n",
       "      <td>[0.0005596342380158603, 0.005513169337064028, ...</td>\n",
       "      <td>politics</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1824</th>\n",
       "      <td>tech</td>\n",
       "      <td>001.txt</td>\n",
       "      <td>Ink helps drive democracy in Asia</td>\n",
       "      <td>The Kyrgyz Republic, a small, mountainous sta...</td>\n",
       "      <td>[0.007163866423070431, -0.0017118602991104126,...</td>\n",
       "      <td>business</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1848</th>\n",
       "      <td>tech</td>\n",
       "      <td>025.txt</td>\n",
       "      <td>Sony PSP console hits US in March</td>\n",
       "      <td>US gamers will be able to buy Sony's PlayStat...</td>\n",
       "      <td>[0.0021356961224228144, 0.0038372818380594254,...</td>\n",
       "      <td>entertainment</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1851</th>\n",
       "      <td>tech</td>\n",
       "      <td>028.txt</td>\n",
       "      <td>Piero gives rugby perspective</td>\n",
       "      <td>BBC Sport unveils its new analysis tool Piero...</td>\n",
       "      <td>[0.00012840930139645934, 0.00610811822116375, ...</td>\n",
       "      <td>sport</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1858</th>\n",
       "      <td>tech</td>\n",
       "      <td>035.txt</td>\n",
       "      <td>Iran jails blogger for 14 years</td>\n",
       "      <td>An Iranian weblogger has been jailed for 14 y...</td>\n",
       "      <td>[-0.0033451421186327934, 0.01058292854577303, ...</td>\n",
       "      <td>politics</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1946</th>\n",
       "      <td>tech</td>\n",
       "      <td>123.txt</td>\n",
       "      <td>Attack prompts Bush site block</td>\n",
       "      <td>The official re-election site of President Ge...</td>\n",
       "      <td>[0.008621099404990673, 0.0067411684431135654, ...</td>\n",
       "      <td>politics</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2011</th>\n",
       "      <td>tech</td>\n",
       "      <td>188.txt</td>\n",
       "      <td>Call for action on internet scam</td>\n",
       "      <td>Phone companies are not doing enough to warn ...</td>\n",
       "      <td>[0.005809388123452663, 0.010312298312783241, -...</td>\n",
       "      <td>business</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2046</th>\n",
       "      <td>tech</td>\n",
       "      <td>223.txt</td>\n",
       "      <td>Satellite mapping aids Darfur relief</td>\n",
       "      <td>Aid workers trying to house, feed and clothe ...</td>\n",
       "      <td>[0.010239184834063053, 0.016440916806459427, -...</td>\n",
       "      <td>business</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2054</th>\n",
       "      <td>tech</td>\n",
       "      <td>231.txt</td>\n",
       "      <td>Parents face video game lessons</td>\n",
       "      <td>Ways of ensuring that parents know which vide...</td>\n",
       "      <td>[0.007026379927992821, 0.007754994556307793, -...</td>\n",
       "      <td>politics</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2140</th>\n",
       "      <td>tech</td>\n",
       "      <td>317.txt</td>\n",
       "      <td>Sony PSP console hits US in March</td>\n",
       "      <td>US gamers will be able to buy Sony's PlayStat...</td>\n",
       "      <td>[0.0021356961224228144, 0.0038372818380594254,...</td>\n",
       "      <td>entertainment</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           category filename                                 title  \\\n",
       "105        business  106.txt     Golden rule 'intact' says ex-aide   \n",
       "212        business  213.txt     US bank boss hails 'genius' Smith   \n",
       "226        business  227.txt       Call to save manufacturing jobs   \n",
       "235        business  236.txt       Card fraudsters 'targeting web'   \n",
       "240        business  241.txt      G7 backs Africa debt relief plan   \n",
       "241        business  242.txt       Q&A: Malcolm Glazer and Man Utd   \n",
       "254        business  255.txt     Call to overhaul UK state pension   \n",
       "292        business  293.txt     Glazer makes new Man Utd approach   \n",
       "387        business  388.txt     Venezuela identifies 'idle' farms   \n",
       "412        business  413.txt      EC calls truce in deficit battle   \n",
       "451        business  452.txt     McDonald's boss Bell dies aged 44   \n",
       "456        business  457.txt           Germany calls for EU reform   \n",
       "459        business  460.txt      Jobs go at Oracle after takeover   \n",
       "518   entertainment  009.txt      Da Vinci Code is 'lousy history'   \n",
       "657   entertainment  148.txt     Pupils to get anti-piracy lessons   \n",
       "682   entertainment  173.txt       UK TV channel rapped for CSI ad   \n",
       "692   entertainment  183.txt     Star Trek fans fight to save show   \n",
       "711   entertainment  202.txt      BBC 'should allow more scrutiny'   \n",
       "775   entertainment  266.txt     Help for indies in download sales   \n",
       "822   entertainment  313.txt       BBC to pour £9m into new comedy   \n",
       "851   entertainment  342.txt      US composer recreates Bach score   \n",
       "893   entertainment  384.txt      Robots march to US cinema summit   \n",
       "899        politics  004.txt             Labour chooses Manchester   \n",
       "901        politics  006.txt       'Errors' doomed first Dome sale   \n",
       "918        politics  023.txt      E-University 'disgraceful waste'   \n",
       "971        politics  076.txt       Report attacks defence spending   \n",
       "985        politics  090.txt     MSPs hear renewed climate warning   \n",
       "1004       politics  109.txt       UK firms 'embracing e-commerce'   \n",
       "1008       politics  113.txt      MP's shock at long lost daughter   \n",
       "1034       politics  139.txt           Straw praises Kashmir moves   \n",
       "1035       politics  140.txt          UK helps raped Rwandan women   \n",
       "1073       politics  178.txt       Hatfield executives go on trial   \n",
       "1089       politics  194.txt      Cherie accused of attacking Bush   \n",
       "1225       politics  330.txt     Manchester wins Labour conference   \n",
       "1288       politics  393.txt        Parliament's record of scandal   \n",
       "1308       politics  413.txt      Sainsbury's Labour election gift   \n",
       "1316          sport  004.txt     IAAF launches fight against drugs   \n",
       "1339          sport  027.txt      Holmes feted with further honour   \n",
       "1443          sport  131.txt     Newry to fight cup exit in courts   \n",
       "1675          sport  363.txt       Calder fears for Scottish rugby   \n",
       "1824           tech  001.txt     Ink helps drive democracy in Asia   \n",
       "1848           tech  025.txt     Sony PSP console hits US in March   \n",
       "1851           tech  028.txt         Piero gives rugby perspective   \n",
       "1858           tech  035.txt       Iran jails blogger for 14 years   \n",
       "1946           tech  123.txt        Attack prompts Bush site block   \n",
       "2011           tech  188.txt      Call for action on internet scam   \n",
       "2046           tech  223.txt  Satellite mapping aids Darfur relief   \n",
       "2054           tech  231.txt       Parents face video game lessons   \n",
       "2140           tech  317.txt     Sony PSP console hits US in March   \n",
       "\n",
       "                                                content  \\\n",
       "105    Chancellor Gordon Brown will meet his golden ...   \n",
       "212    US Federal Reserve chairman Alan Greenspan ha...   \n",
       "226    The Trades Union Congress (TUC) is calling on...   \n",
       "235    New safeguards on credit and debit card payme...   \n",
       "240    G7 finance ministers have backed plans to wri...   \n",
       "241    The battle for control of Manchester United h...   \n",
       "254    The UK pension system has been branded inadeq...   \n",
       "292    Malcolm Glazer has made a fresh approach to b...   \n",
       "387    Venezuelan authorities have identified more t...   \n",
       "412    The European Commission (EC) has called a tru...   \n",
       "451    Charlie Bell, the straight-talking former hea...   \n",
       "456    German Chancellor Gerhard Schroeder has calle...   \n",
       "459    Oracle has announced it is cutting about 5,00...   \n",
       "518    The plot of an international bestseller that ...   \n",
       "657    Lessons on music piracy and copyright issues ...   \n",
       "682    TV channel Five has been criticised for sendi...   \n",
       "692    Star Trek fans have taken out a full-page ad ...   \n",
       "711    MPs have urged the BBC to give watchdogs more...   \n",
       "775    A campaign has been launched to help independ...   \n",
       "822    The BBC is to invest £9m in developing new co...   \n",
       "851    A US musicologist has recreated a lost musica...   \n",
       "893    Animated movie Robots has opened at the top o...   \n",
       "899    The Labour Party will hold its 2006 autumn co...   \n",
       "901    The initial attempt to sell the Millennium Do...   \n",
       "918    A failed government scheme to offer UK univer...   \n",
       "971    The Ministry of Defence has been criticised o...   \n",
       "985    Climate change could be completely out of con...   \n",
       "1004   UK firms are embracing internet trading oppor...   \n",
       "1008   Labour MP Stephen Pound has told of his shock...   \n",
       "1034   The UK has welcomed the decision by India and...   \n",
       "1035   Britain is to give a £4m grant to help women ...   \n",
       "1073   Engineering firm Balfour Beatty and five rail...   \n",
       "1089   Cherie Blair has been accused of criticising ...   \n",
       "1225   The Labour Party will hold its 2006 autumn co...   \n",
       "1288   In a locked room at the heart of Parliament t...   \n",
       "1308   Science Minister Lord Sainsbury has made a £2...   \n",
       "1316   The IAAF - athletics' world governing body - ...   \n",
       "1339   Double Olympic champion Kelly Holmes has been...   \n",
       "1443   Newry City are expected to discuss legal aven...   \n",
       "1675   Former Scotland international Finlay Calder f...   \n",
       "1824   The Kyrgyz Republic, a small, mountainous sta...   \n",
       "1848   US gamers will be able to buy Sony's PlayStat...   \n",
       "1851   BBC Sport unveils its new analysis tool Piero...   \n",
       "1858   An Iranian weblogger has been jailed for 14 y...   \n",
       "1946   The official re-election site of President Ge...   \n",
       "2011   Phone companies are not doing enough to warn ...   \n",
       "2046   Aid workers trying to house, feed and clothe ...   \n",
       "2054   Ways of ensuring that parents know which vide...   \n",
       "2140   US gamers will be able to buy Sony's PlayStat...   \n",
       "\n",
       "                                              embedding     prediction  \n",
       "105   [0.006964212749153376, 0.0015263919485732913, ...       politics  \n",
       "212   [0.0021459858398884535, 0.0036910325288772583,...  entertainment  \n",
       "226   [0.0012122669722884893, 0.0076951221562922, -0...       politics  \n",
       "235   [7.788832590449601e-05, 0.016654228791594505, ...           tech  \n",
       "240   [-0.0027972061652690172, 0.013444329611957073,...       politics  \n",
       "241   [0.002724364399909973, 0.004842962138354778, -...          sport  \n",
       "254   [0.004907403141260147, 0.0119895339012146, -0....       politics  \n",
       "292   [-0.0023372862488031387, 0.006699692457914352,...          sport  \n",
       "387   [0.006353371310979128, 0.014112972654402256, -...       politics  \n",
       "412   [0.003549534361809492, 0.005724851507693529, -...       politics  \n",
       "451   [0.004609732422977686, 0.006802281830459833, -...  entertainment  \n",
       "456   [-0.0018439812120050192, 0.009731155820190907,...       politics  \n",
       "459   [0.0036041021812707186, 0.009482146240770817, ...           tech  \n",
       "518   [-0.002073341980576515, 0.007545760832726955, ...       politics  \n",
       "657   [-0.004451509099453688, 0.014561662450432777, ...           tech  \n",
       "682   [-0.0036573493853211403, 0.015478764660656452,...           tech  \n",
       "692   [0.007987278513610363, 0.014516561292111874, -...           tech  \n",
       "711   [0.0011142775183543563, 0.010890933685004711, ...       politics  \n",
       "775   [-0.0023760353215038776, 0.008679348975419998,...           tech  \n",
       "822   [-0.0013941482175141573, 0.012636896222829819,...       business  \n",
       "851   [-0.011926768347620964, 0.011977385729551315, ...           tech  \n",
       "893   [0.007351818960160017, 0.016657356172800064, -...           tech  \n",
       "899   [0.007821072824299335, 0.00648058857768774, 0....       business  \n",
       "901   [0.0006757106748409569, 0.007997625507414341, ...       business  \n",
       "918   [-7.362403266597539e-05, 0.006662579253315926,...           tech  \n",
       "971   [0.0015996992588043213, 0.007126652169972658, ...       business  \n",
       "985   [0.012153446674346924, 0.007975501008331776, -...       business  \n",
       "1004  [-0.0006202863878570497, 0.004257924854755402,...       business  \n",
       "1008  [0.002527164528146386, 0.006853112950921059, -...  entertainment  \n",
       "1034  [0.00722165871411562, 0.006721210666000843, -0...       business  \n",
       "1035  [0.004300092346966267, 0.013324585743248463, -...       business  \n",
       "1073  [0.0065396688878536224, 0.009236570447683334, ...  entertainment  \n",
       "1089  [0.004710556473582983, 0.0016932026483118534, ...  entertainment  \n",
       "1225  [0.007821072824299335, 0.00648058857768774, 0....       business  \n",
       "1288  [-0.004831457510590553, 0.009346305392682552, ...           tech  \n",
       "1308  [0.002463462296873331, 0.007179685402661562, -...       business  \n",
       "1316  [0.009765633381903172, 0.01343011949211359, -0...       business  \n",
       "1339  [0.010477458126842976, 0.0017678732983767986, ...  entertainment  \n",
       "1443  [-0.007764059584587812, 0.0037425938062369823,...       politics  \n",
       "1675  [0.0005596342380158603, 0.005513169337064028, ...       politics  \n",
       "1824  [0.007163866423070431, -0.0017118602991104126,...       business  \n",
       "1848  [0.0021356961224228144, 0.0038372818380594254,...  entertainment  \n",
       "1851  [0.00012840930139645934, 0.00610811822116375, ...          sport  \n",
       "1858  [-0.0033451421186327934, 0.01058292854577303, ...       politics  \n",
       "1946  [0.008621099404990673, 0.0067411684431135654, ...       politics  \n",
       "2011  [0.005809388123452663, 0.010312298312783241, -...       business  \n",
       "2046  [0.010239184834063053, 0.016440916806459427, -...       business  \n",
       "2054  [0.007026379927992821, 0.007754994556307793, -...       politics  \n",
       "2140  [0.0021356961224228144, 0.0038372818380594254,...  entertainment  "
      ]
     },
     "execution_count": 100,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_wrong_predictions = df_test_result[df_test_result['category'] != df_test_result['prediction']]\n",
    "df_wrong_predictions"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "azureml_py38",
   "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": "3.8.5"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "6d65a8c07f5b6469e0fc613f182488c0dccce05038bbda39e5ac9075c0454d11"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}