{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "c88056d1-bf3c-477e-9b36-50e7398f9058",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "186faae4-de4a-4df5-b914-203a4e6296b0",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv(\"step5.csv\", encoding='utf-8-sig', sep=';')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e672cf21-1f91-424c-b944-5d0a12ac69fd",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from sklearn.metrics import classification_report, confusion_matrix\n",
    "\n",
    "classification_label_set = {\n",
    "    \"auto\": 0,\n",
    "    \"Business and Industry\": 1,\n",
    "    \"Crime and Justice\": 2,\n",
    "    \"Disaster and Emergency News\": 3,\n",
    "    \"Economics and Finance\": 4,\n",
    "    \"Education\": 5,\n",
    "    \"Entertainment and Culture\": 6,\n",
    "    \"Environment and Climate\": 7,\n",
    "    \"Family and Relationships\": 8,\n",
    "    \"Fashion\": 9,\n",
    "    \"Food and Drink\": 10,\n",
    "    \"Health and Medicine\": 11,\n",
    "    \"Transportation and Infrastructure\": 12,\n",
    "    \"Mental Health and Wellness\": 13,\n",
    "    \"Politics and Government\": 14,\n",
    "    \"Religion\": 15,\n",
    "    \"Sports\": 16,\n",
    "    \"Travel and Leisure\": 17,\n",
    "    \"Technology and Science\": 18\n",
    "}\n",
    "\n",
    "\n",
    "classification_label_dict = {idx: label for idx, label in enumerate(classification_label_set)}\n",
    "classification_label_dict_to_index = {v: k for k, v in classification_label_dict.items()}\n",
    "\n",
    "\n",
    "ner_label_set = [\"PAD\",\"O\",\n",
    "    \"B-ORG\", \"I-ORG\", \"B-PERSON\", \"I-PERSON\", \"B-CARDINAL\", \"I-CARDINAL\",\n",
    "    \"B-GPE\", \"I-GPE\", \"B-DATE\", \"I-DATE\", \"B-ORDINAL\", \"I-ORDINAL\",\n",
    "    \"B-PERCENT\", \"I-PERCENT\", \"B-LOC\", \"I-LOC\", \"B-NORP\", \"I-NORP\",\n",
    "    \"B-MONEY\", \"I-MONEY\", \"B-TIME\", \"I-TIME\", \"B-EVENT\", \"I-EVENT\",\n",
    "    \"B-PRODUCT\", \"I-PRODUCT\", \"B-FAC\", \"I-FAC\", \"B-QUANTITY\", \"I-QUANTITY\"\n",
    "]\n",
    "\n",
    "\n",
    "ner_label_dict = {label: idx for idx, label in enumerate(ner_label_set)}\n",
    "\n",
    "ner_label_dict_reverse = {idx: label for label, idx in ner_label_dict.items()}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "ced20ed4-b3ca-4401-9221-82bc6e42cfd9",
   "metadata": {},
   "outputs": [],
   "source": [
    "from transformers import AutoTokenizer\n",
    "model_name = \"nlpaueb/bert-base-greek-uncased-v1\"\n",
    "tokenizer = AutoTokenizer.from_pretrained(model_name)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "6360e73b-ddea-4c9b-ac30-366e5f843498",
   "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>url</th>\n",
       "      <th>title</th>\n",
       "      <th>text</th>\n",
       "      <th>label</th>\n",
       "      <th>clean_text</th>\n",
       "      <th>ner_results</th>\n",
       "      <th>ner_sequence</th>\n",
       "      <th>words</th>\n",
       "      <th>entities</th>\n",
       "      <th>entities_parsed</th>\n",
       "      <th>entities_numbers</th>\n",
       "      <th>label_numbers</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>https://www.sport-fm.gr/article/o-roueda-sti-S...</td>\n",
       "      <td>Ο πρώην της Ferrari πήγε στην Sauber</td>\n",
       "      <td>Τον πρώην επικεφαλής στρατηγικής και αθλητικό ...</td>\n",
       "      <td>auto</td>\n",
       "      <td>Ο πρώην της Ferrari πήγε στην Sauber Τον πρώην...</td>\n",
       "      <td>[{'text': 'ferrari', 'type': 'ORG', 'start_pos...</td>\n",
       "      <td>[{'word': 'Ο', 'entity': 'O'}, {'word': 'πρώην...</td>\n",
       "      <td>['Ο', 'πρώην', 'της', 'Ferrari', 'πήγε', 'στην...</td>\n",
       "      <td>['O', 'O', 'O', 'B-ORG', 'O', 'O', 'B-ORG', 'O...</td>\n",
       "      <td>['O', 'O', 'O', 'B-ORG', 'O', 'O', 'B-ORG', 'O...</td>\n",
       "      <td>[1, 1, 1, 2, 1, 1, 2, 1, 1, 1, 1, 1, 1, 1, 1, ...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                 url  \\\n",
       "0  https://www.sport-fm.gr/article/o-roueda-sti-S...   \n",
       "\n",
       "                                  title  \\\n",
       "0  Ο πρώην της Ferrari πήγε στην Sauber   \n",
       "\n",
       "                                                text label  \\\n",
       "0  Τον πρώην επικεφαλής στρατηγικής και αθλητικό ...  auto   \n",
       "\n",
       "                                          clean_text  \\\n",
       "0  Ο πρώην της Ferrari πήγε στην Sauber Τον πρώην...   \n",
       "\n",
       "                                         ner_results  \\\n",
       "0  [{'text': 'ferrari', 'type': 'ORG', 'start_pos...   \n",
       "\n",
       "                                        ner_sequence  \\\n",
       "0  [{'word': 'Ο', 'entity': 'O'}, {'word': 'πρώην...   \n",
       "\n",
       "                                               words  \\\n",
       "0  ['Ο', 'πρώην', 'της', 'Ferrari', 'πήγε', 'στην...   \n",
       "\n",
       "                                            entities  \\\n",
       "0  ['O', 'O', 'O', 'B-ORG', 'O', 'O', 'B-ORG', 'O...   \n",
       "\n",
       "                                     entities_parsed  \\\n",
       "0  ['O', 'O', 'O', 'B-ORG', 'O', 'O', 'B-ORG', 'O...   \n",
       "\n",
       "                                    entities_numbers  label_numbers  \n",
       "0  [1, 1, 1, 2, 1, 1, 2, 1, 1, 1, 1, 1, 1, 1, 1, ...              0  "
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "7eb5ce65-3cbf-4ba4-9521-7d13ba3687d0",
   "metadata": {},
   "outputs": [],
   "source": [
    "df['entities_numbers'] = df['entities_numbers'].apply(ast.literal_eval)\n",
    "df['labels'] = df['label_numbers']\n",
    "df['ner_labels'] = df['entities_numbers']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "68d22607-e2d3-4228-b35b-97842b5e64e3",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import random\n",
    "import torch\n",
    "\n",
    "def set_seed(seed=42):\n",
    "    random.seed(seed)\n",
    "    np.random.seed(seed)\n",
    "    torch.manual_seed(seed)\n",
    "    torch.cuda.manual_seed_all(seed)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "21ee5eb1-8832-4afe-80aa-27989e824853",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "--- Sample Inspection ---\n",
      "Tokens:\n",
      "['[CLS]', 'ανεμος', 'σαι', '##νθ', 'στο', 'μεξικο', 'κερδισμενο', '##ς', 'ο', 'νορ', '##ις', 'κυριαρχια', 'της', 'ferrari', 'κυριαρχο', '##ς', 'σαι', '##νθ', 'και', 'κυριαρχη', 'ferrari', 'στο', 'μεξικο', '.', 'ο', 'ισπανος', ',', 'ο', 'οποιος', 'ετοιμαζεται', 'για', 'το', 'νεο', 'κεφαλαιο', 'της', 'καριερας', 'του', ',', 'τερματισε', 'πρωτος', 'στο', '19', '##ο', 'grand', 'prix', 'της', 'σεζον', ',', 'σημειωνοντας', 'τη', '2', '##η', 'νικη', 'του', '(', 'επειτα', 'απο', 'την', 'αυστραλια', ')', ',', 'στη', '2', '##η', 'νικη', 'της', 'ferrari', 'επειτα', 'απο', 'εκεινη', 'του', 'λε', '##κλε', '##ρ', 'που', 'τερματισε', '3', '##ος', '.', 'κερδισμενο', '##ς', 'βγηκε', 'ο', 'νορ', '##ις', ',', 'ο', 'οποιος', 'μειωσε', 'στους', '47', 'την', 'αποσταση', 'του', 'απο', 'τον', 'πρωτοπορο', 'φερ', '##στα', '##πεν', '.', 'ο', '27', '##χρονος', 'τερματισε', 'στην', '6', '##η', 'θεση', 'λογω', 'ποινη', '##ς', '20', 'δευτερολεπτ', '##ων', '.', 'διοτι', 'σε', 'κουρσα', 'με', 'αντιπαλο', 'τον', 'νορ', '##ις', 'τον', 'ωθησε', 'εκτος', 'πιστα', '##ς', 'και', 'επειτα', 'βγηκε', 'και', 'ο', 'ιδιος', ',', 'αποκτωντας', 'πλεονεκτημα', '.', 'ο', 'σαι', '##νθ', 'ζει', 'τις', 'τελευταιες', 'του', 'στιγμες', 'ως', 'μελος', 'της', 'ferrari', 'και', 'θελει', 'να', 'την', 'αποχαιρετησ', '##ει', 'με', 'τον', 'καλυτερο', 'δυνατο', 'τροπο', '##ενα', 'λαθος', 'του', 'λε', '##κλε', '##ρ', 'πεντε', 'στροφες', 'πριν', 'το', 'φιναλε', 'του', 'στερ', '##ησε', 'τη', '2', '##η', 'θεση', 'και', 'επιπλεον', 'τρεις', 'ποντους', '.', 'παροτι', 'σχεδον', 'σε', 'ολοκληρο', 'τον', 'αγωνα', 'ειχε', 'το', 'πανω', 'χερι', ',', 'αλλα', 'το', 'σφαλμα', 'λιγο', 'πριν', 'το', 'τελος', 'του', 'κοστισε', 'ακριβα', '.', 'παραλληλα', ',', 'στους', 'κατασκευαστες', 'η', 'διαφορα', 'ειναι', 'στους', '29', 'ποντους', ',', 'με', 'τη', 'mclaren', 'να', 'βρισκεται', 'πρωτη', 'με', '56', '##6', ',', 'ενω', 'η', 'ferrari', 'εφτασε', 'τους', '53', '##7', '.', '[SEP]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]', '[PAD]']\n",
      "\n",
      "Transformed Word IDs:\n",
      "[-100, 1, 2, 2, 3, 4, 5, 5, 6, 7, 7, 8, 9, 10, 11, 11, 12, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 36, 37, 38, 39, 40, 41, 42, 43, 44, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 55, 56, 57, 58, 59, 60, 61, 62, 63, 63, 63, 64, 65, 66, 66, 67, 68, 68, 69, 70, 71, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 84, 84, 85, 86, 87, 87, 88, 89, 90, 90, 91, 92, 93, 93, 94, 95, 95, 96, 97, 98, 99, 100, 101, 102, 103, 103, 104, 105, 106, 107, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 133, 134, 135, 136, 137, 138, 138, 139, 140, 141, 141, 141, 142, 143, 144, 145, 146, 147, 148, 148, 149, 150, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 197, 198, 199, 200, 201, 202, 203, 204, 204, 205, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100]\n",
      "\n",
      "Aligned NER labels:\n",
      "[-100, 1, 4, -100, 1, 8, 1, -100, 1, 4, -100, 1, 1, 2, 1, -100, 4, -100, 1, 1, 2, 1, 8, 1, 1, 18, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 12, 1, 24, -100, 25, 25, 1, 1, 1, 1, 1, 12, -100, 1, 1, 1, 1, 1, 1, 2, 1, 1, 1, 12, -100, 1, 1, 2, 1, 1, 1, 1, 4, -100, -100, 1, 1, 12, -100, 1, 1, -100, 1, 1, 4, -100, 1, 1, 1, 1, 1, 6, 1, 1, 1, 1, 1, 1, 4, -100, -100, 1, 1, 1, -100, 1, 1, 12, -100, 1, 1, 1, -100, 22, 23, -100, 1, 1, 1, 1, 1, 1, 1, 4, -100, 1, 1, 1, 1, -100, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 4, -100, 1, 1, 1, 1, 1, 1, 1, 1, 2, 1, 1, 1, 1, 1, -100, 1, 1, 1, 1, 1, -100, 1, 1, 4, -100, -100, 22, 23, 1, 1, 1, 1, 1, -100, 1, 12, -100, 1, 1, 1, 6, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 6, 1, 1, 1, 1, 2, 1, 1, 12, 1, 6, -100, 1, 1, 1, 2, 1, 1, 6, -100, 1, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100, -100]\n"
     ]
    }
   ],
   "source": [
    "import torch\n",
    "from torch.utils.data import DataLoader, Dataset\n",
    "import ast\n",
    "import pandas as pd\n",
    "from sklearn.model_selection import train_test_split\n",
    "from transformers import AutoTokenizer\n",
    "\n",
    "ner_label_dict = {label: idx for idx, label in enumerate(ner_label_set)}\n",
    "ner_label_dict_reverse = {v: k for k, v in ner_label_dict.items()}\n",
    "\n",
    "def chunk_list(items, chunk_size):\n",
    "    \"\"\"Splits the list 'items' into chunks of size 'chunk_size'.\"\"\"\n",
    "    return [items[i:i + chunk_size] for i in range(0, len(items), chunk_size)]\n",
    "\n",
    "class NewsNERChunkedDataset(Dataset):\n",
    "    def __init__(self, dataframe, tokenizer, word_chunk_size=412, target_token_length=512):\n",
    "        self.tokenizer = tokenizer\n",
    "        self.word_chunk_size = word_chunk_size\n",
    "        self.target_token_length = target_token_length\n",
    "        \n",
    "        self.samples = []\n",
    "        for _, row in dataframe.iterrows():\n",
    "            words = row['words']\n",
    "            if isinstance(words, str):\n",
    "                words = ast.literal_eval(words)\n",
    "            \n",
    "            # Classification label\n",
    "            label_class = int(row['labels'])\n",
    "            \n",
    "            # NER labels\n",
    "            ner_labels_val = row['ner_labels']\n",
    "            if isinstance(ner_labels_val, str):\n",
    "                ner_labels_val = ast.literal_eval(ner_labels_val)\n",
    "            \n",
    "            # Split into chunks\n",
    "            word_chunks = chunk_list(words, self.word_chunk_size)\n",
    "            ner_label_chunks = chunk_list(ner_labels_val, self.word_chunk_size)\n",
    "            \n",
    "            for word_chunk, ner_chunk in zip(word_chunks, ner_label_chunks):\n",
    "                encoding = self.tokenizer(\n",
    "                    word_chunk,\n",
    "                    is_split_into_words=True,\n",
    "                    truncation=True,\n",
    "                    max_length=self.target_token_length,\n",
    "                    padding='max_length',\n",
    "                    return_offsets_mapping=True\n",
    "                )\n",
    "                \n",
    "                raw_word_ids = encoding.word_ids() \n",
    "                transformed_word_ids = []\n",
    "                for w_id in raw_word_ids:\n",
    "                    if w_id is None:\n",
    "                        transformed_word_ids.append(-100)      \n",
    "                    else:\n",
    "                        transformed_word_ids.append(w_id + 1)  \n",
    "\n",
    "                aligned_labels = []\n",
    "                previous_word_idx = None\n",
    "                for w_idx in encoding.word_ids():\n",
    "                    if w_idx is None:\n",
    "                        aligned_labels.append(-100) \n",
    "                    elif w_idx != previous_word_idx:\n",
    "                        if w_idx < len(ner_chunk):\n",
    "                            aligned_labels.append(ner_chunk[w_idx])\n",
    "                        else:\n",
    "                            aligned_labels.append(ner_label_dict[\"O\"])\n",
    "                        previous_word_idx = w_idx\n",
    "                    else:\n",
    "                        # Subword of the *same* word -> ignore for NER\n",
    "                        aligned_labels.append(-100)\n",
    "            \n",
    "                sample = {\n",
    "                    'input_ids': torch.tensor(encoding['input_ids']),\n",
    "                    'attention_mask': torch.tensor(encoding['attention_mask']),\n",
    "                    'labels_class': torch.tensor(label_class),\n",
    "                    'labels_ner': torch.tensor(aligned_labels),\n",
    "                    'word_ids': torch.tensor(transformed_word_ids)  \n",
    "                }\n",
    "                self.samples.append(sample)\n",
    "\n",
    "    def __len__(self):\n",
    "        return len(self.samples)\n",
    "    \n",
    "    def __getitem__(self, idx):\n",
    "        return self.samples[idx]\n",
    "\n",
    "\n",
    "train_df, temp_df = train_test_split(\n",
    "    df, \n",
    "    test_size=0.3, \n",
    "    random_state=42, \n",
    "    stratify=df['labels']\n",
    ")\n",
    "\n",
    "val_df, test_df = train_test_split(\n",
    "    temp_df, \n",
    "    test_size=0.5, \n",
    "    random_state=42, \n",
    "    stratify=temp_df['labels']\n",
    ")\n",
    "# -------------------------------\n",
    "# Tokenizer\n",
    "tokenizer = AutoTokenizer.from_pretrained(model_name)\n",
    "\n",
    "# Datasets\n",
    "train_dataset = NewsNERChunkedDataset(train_df, tokenizer)\n",
    "val_dataset   = NewsNERChunkedDataset(val_df, tokenizer)\n",
    "test_dataset  = NewsNERChunkedDataset(test_df, tokenizer)\n",
    "\n",
    "# sample\n",
    "sample = train_dataset[0]\n",
    "ids = sample['input_ids'].tolist()\n",
    "tokens = tokenizer.convert_ids_to_tokens(ids)\n",
    "wids = sample['word_ids'].tolist()\n",
    "nerlabs = sample['labels_ner'].tolist()\n",
    "\n",
    "print(\"\\n--- Sample Inspection ---\")\n",
    "print(\"Tokens:\")\n",
    "print(tokens)\n",
    "print(\"\\nTransformed Word IDs:\")\n",
    "print(wids)\n",
    "print(\"\\nAligned NER labels:\")\n",
    "print(nerlabs)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "1bbd67c1-6922-48c7-9a60-1036f736cb99",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "40a3dd23-cae5-4f09-935a-61fc13dbbcd3",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b0eb64f8-88be-4369-aa2d-9652fd253e25",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.21"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}