{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "(label-propagation-page)=\n",
    "# Propagation of NER labels\n",
    "\n",
    "In the context of Named Entity Recognition (NER), typical datasets contain the text tokens and the NER labels for each of the tokens. For example:\n",
    "\n",
    "````{margin}\n",
    "```{note}\n",
    "`B-P` is short for \"Beginning-Place\"\n",
    "and `I-P` is short for \"Inside-Place\"\n",
    "whereas `O` means \"Other\".\n",
    "See [IOB Tagging](https://en.wikipedia.org/wiki/Inside%E2%80%93outside%E2%80%93beginning_(tagging)) for more details\n",
    "```\n",
    "````\n",
    "    NER Labels: B-P I-P  O  O\n",
    "          Text: New York is big\n"
   ]
  },
  {
   "source": [
    "Now, imagine we have obtained a noisy version of the grouth truth text through the OCR process, for example. The problem becomes: how can we label the noisy tokens?\n",
    "\n",
    "\n",
    "        NER Labels:  B-P I-P  O  O\n",
    "           GT Text:  New York is big\n",
    "        Noisy Text:  New Yo rkis big\n",
    "        NER Labels:   ?  ?   ?    ?\n",
    "\n",
    "We can utilize text alignment and **propagate** the NER labels onto the noisy tokens. We will demonstrate how in the rest of this document.\n"
   ],
   "cell_type": "markdown",
   "metadata": {}
  },
  {
   "source": [
    "## Tokenization\n",
    "\n",
    "To ensure consistent interpretation of the text alignment results, we need to first tokenize the grouth truth and the OCR'ed (nosiy) text."
   ],
   "cell_type": "markdown",
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from genalog.text import ner_label\n",
    "from genalog.text import preprocess\n",
    "\n",
    "gt_txt = \"New York is big\"\n",
    "ocr_txt = \"New Yo rkis big\"\n",
    "\n",
    "# Input to the method\n",
    "gt_labels = [\"B-P\", \"I-P\", \"O\", \"O\"]\n",
    "gt_tokens = preprocess.tokenize(gt_txt) # tokenize into list of tokens\n",
    "ocr_tokens = preprocess.tokenize(ocr_txt)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['B-P', 'I-P', 'O', 'O']\n",
      "['New', 'York', 'is', 'big']\n",
      "['New', 'Yo', 'rkis', 'big']\n"
     ]
    }
   ],
   "source": [
    "# Inputs to the method\n",
    "print(gt_labels)\n",
    "print(gt_tokens)\n",
    "print(ocr_tokens)"
   ]
  },
  {
   "source": [
    "## Label Propagation\n",
    "\n",
    "We then can run label propagation to obtain the NER labels for the OCR'ed (noisy) tokens."
   ],
   "cell_type": "markdown",
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Method returns a tuple of 4 elements (gt_tokens, gt_labels, ocr_tokens, ocr_labels, gap_char)\n",
    "ocr_labels, aligned_gt, aligned_ocr, gap_char = ner_label.propagate_label_to_ocr(gt_labels, gt_tokens, ocr_tokens)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "OCR labels:           ['B-P', 'I-P', 'I-P', 'O']\n",
      "Aligned ground truth: New Yo@rk is big\n",
      "Alinged OCR text:     New Yo rk@is big\n"
     ]
    }
   ],
   "source": [
    "# Outputs\n",
    "print(f\"OCR labels:           {ocr_labels}\")\n",
    "print(f\"Aligned ground truth: {aligned_gt}\")\n",
    "print(f\"Alinged OCR text:     {aligned_ocr}\")"
   ]
  },
  {
   "source": [
    "## Display Result After Propagation"
   ],
   "cell_type": "markdown",
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "B-P I-P  O  O   \n",
      "New York is big \n",
      "New Yo@rk is big\n",
      "||||||.||.||||||\n",
      "New Yo rk@is big\n",
      "New Yo  rkis big \n",
      "B-P I-P I-P  O   \n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(ner_label.format_label_propagation(gt_tokens, gt_labels, ocr_tokens, ocr_labels, aligned_gt, aligned_ocr))"
   ]
  },
  {
   "source": [
    "## Final Results\n",
    "\n",
    "Formatting the OCR tokens and their NER labels."
   ],
   "cell_type": "markdown",
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "B-P I-P I-P  O   \n",
      "New Yo  rkis big \n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Format tokens and labels\n",
    "print(ner_label.format_labels(ocr_tokens, ocr_labels))"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}