kleister-nda/main.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import lzma\n",
    "\n",
    "# RTead file with lzma\n",
    "NDAs = []\n",
    "\n",
    "with lzma.open('train/in.tsv.xz') as f:\n",
    "    for line in f:\n",
    "        NDAs.append(line.decode('utf-8'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Read expected information\n",
    "expected = []\n",
    "\n",
    "with open('train/expected.tsv') as f:\n",
    "    for line in f:\n",
    "        expected.append(line.replace('\\n', '').split(' '))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import re\n",
    "\n",
    "months = {'01': 'January', '02': 'February', '03': 'March', \n",
    "    '04': 'April', '05': 'May', '06': 'June',\n",
    "    '07': 'July', '08': 'August', '09': 'September',\n",
    "    '10': 'October', '11': 'November', '12': 'December'}\n",
    "\n",
    "def dayToWord(day):\n",
    "    day = int(day)\n",
    "    if day > 3:\n",
    "        return str(day) + 'th'\n",
    "    elif day == 3:\n",
    "        return str(day) + 'rd'\n",
    "    elif day == 2 :\n",
    "        return str(day) + 'nd'\n",
    "    else: return str(day) + 'st'\n",
    "\n",
    "def numToWord(number):\n",
    "    try:\n",
    "        number = int(number)\n",
    "        d = {1 : 'one', 2 : 'two', 3 : 'three', 4 : 'four', 5 : 'five',\n",
    "            6 : 'six', 7 : 'seven', 8 : 'eight', 9 : 'nine', 10 : 'ten',\n",
    "            11 : 'eleven', 12 : 'twelve', 13 : 'thirteen', 14 : 'fourteen',\n",
    "            15 : 'fifteen', 16 : 'sixteen', 17 : 'seventeen', 18 : 'eighteen',\n",
    "            19 : 'nineteen', 20 : 'twenty',\n",
    "            30 : 'thirty', 40 : 'forty', 50 : 'fifty', 60 : 'sixty',\n",
    "            70 : 'seventy', 80 : 'eighty', 90 : 'ninety' }\n",
    "        if number < 20:\n",
    "            return d[number]\n",
    "        else:\n",
    "            if number % 10 == 0: return d[number]\n",
    "            else: return d[number // 10 * 10] + '-' + d[number % 10]\n",
    "    except:\n",
    "        return number\n",
    "\n",
    "def labelJurisdiction(text, jurisdiction):\n",
    "    jurisdictions = []\n",
    "    jurisdiction = jurisdiction.replace('_', ' ')\n",
    "    for match in re.finditer(jurisdiction, text):\n",
    "        tup = (match.start(), match.end(), 'jurisdiction')\n",
    "        jurisdictions.append(tup)\n",
    "    return jurisdictions\n",
    "\n",
    "def labelEffectiveDate(text, date):\n",
    "    dates = []\n",
    "    year, month, day = date.split('-')\n",
    "    \n",
    "    dateFormats = [month + '/' + day + '/' + year,\n",
    "        month + '/' + day + '/' + year[-2:], \n",
    "        month[1] + '/' + day + '/' + year, \n",
    "        month[1] + '/' + day[1] + '/' + year, \n",
    "        month[1] + '/' + day + '/' + year[-2:], \n",
    "        month[1] + '/' + day[1] + '/' + year[-2:],\n",
    "        dayToWord(day) + ' of ' + months[month] + ', ' + year,\n",
    "        dayToWord(day) + ' day of ' + months[month] + ', ' + year,\n",
    "        months[month] + ' ' + day + ', ' + year ]\n",
    "\n",
    "    for format in dateFormats:\n",
    "        for match in re.finditer(format, text, flags=re.IGNORECASE):\n",
    "            tup = (match.start(), match.end(), 'effective_date')\n",
    "            dates.append(tup)\n",
    "\n",
    "    return dates\n",
    "\n",
    "def labelParties(text, party):\n",
    "    parties = []\n",
    "    if 'Inc' in party:\n",
    "        regular = ''\n",
    "        for word in party.split('_'):\n",
    "            regular += word + '(.*)'\n",
    "        party = regular\n",
    "    party = party.replace('_', ' ')\n",
    "    for match in re.finditer(party, text, flags=re.IGNORECASE):\n",
    "        tup = (match.start(), match.end(), 'party')\n",
    "        parties.append(tup)\n",
    "    return parties\n",
    "\n",
    "def labelTerms(text, term):\n",
    "    terms = []\n",
    "    term = term.split('_')\n",
    "    number = numToWord(term[0])\n",
    "    units = term[1]\n",
    "    for match in re.finditer(number + ' ' + units, text, flags=re.IGNORECASE):\n",
    "        tup = (match.start(), match.end(), 'term')\n",
    "        terms.append(tup)\n",
    "    return terms"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "expectEntities = []\n",
    "\n",
    "for expect in expected:\n",
    "    # expect = expect.split()\n",
    "    entities = []\n",
    "    for e in expect:\n",
    "        label, entity = e.split('=')\n",
    "        entities.append((label, entity))\n",
    "    expectEntities.append(entities)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "trainData =[]\n",
    "\n",
    "for i in range(len(expectEntities)):\n",
    "    listOfEntities = []\n",
    "    for entity in expectEntities[i]:\n",
    "        if entity[0] == 'effective_date':\n",
    "            listOfEntities.append(labelEffectiveDate(NDAs[i], entity[1]))\n",
    "        elif entity[0] == 'jurisdiction':\n",
    "            listOfEntities.append(labelJurisdiction(NDAs[i], entity[1]))\n",
    "        elif entity[0] == 'party':\n",
    "            listOfEntities.append(labelParties(NDAs[i], entity[1]))\n",
    "        else: listOfEntities.append(labelTerms(NDAs[i], entity[1]))\n",
    "    listOfEntities = [item for sublist in listOfEntities for item in sublist]\n",
    "    trainData.append((NDAs[i], {'entities': listOfEntities}))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Created blank \"en\" model\n"
     ]
    }
   ],
   "source": [
    "import spacy\n",
    "from spacy.tokens import DocBin\n",
    "\n",
    "model = None\n",
    "nIter = 100\n",
    "\n",
    "if model is not None:\n",
    "    nlp = spacy.load(model)\n",
    "    print('Loaded model')\n",
    "else:\n",
    "    nlp = spacy.blank('en')\n",
    "    print('Created blank \"en\" model')\n",
    "\n",
    "if 'ner' not in nlp.pipe_names:\n",
    "    # ner = nlp.create_pipe('ner')\n",
    "    ner = nlp.add_pipe('ner', last=True)\n",
    "else:\n",
    "    ner = nlp.get_pipe('ner')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "for data in trainData:\n",
    "    for ent in data[1].get('entities'):\n",
    "        ner.add_label(ent[2])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "otherPipes = [pipe for pipe in nlp.pipe_names if pipe != 'ner']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  1%|          | 3/254 [00:00<01:11,  3.49it/s]/Library/Frameworks/Python.framework/Versions/3.9/lib/python3.9/site-packages/spacy/training/iob_utils.py:141: UserWarning: [W030] Some entities could not be aligned in the text \"03efbda01358533c167ca9b1e6d72051.pdf\teffective_dat...\" with entities \"[(7513, 7521, 'effective_date'), (15032, 15040, 'e...\". Use `spacy.training.offsets_to_biluo_tags(nlp.make_doc(text), entities)` to check the alignment. Misaligned entities ('-') will be ignored during training.\n",
      "  warnings.warn(\n",
      "  2%|▏         | 4/254 [00:01<02:28,  1.68it/s]/Library/Frameworks/Python.framework/Versions/3.9/lib/python3.9/site-packages/spacy/training/iob_utils.py:141: UserWarning: [W030] Some entities could not be aligned in the text \"03fd0e629b617da00c54794a8a78b24d.pdf\teffective_dat...\" with entities \"[(287, 300, 'effective_date'), (25276, 25289, 'eff...\". Use `spacy.training.offsets_to_biluo_tags(nlp.make_doc(text), entities)` to check the alignment. Misaligned entities ('-') will be ignored during training.\n",
      "  warnings.warn(\n",
      "  2%|▏         | 6/254 [00:04<04:11,  1.01s/it]/Library/Frameworks/Python.framework/Versions/3.9/lib/python3.9/site-packages/spacy/training/iob_utils.py:141: UserWarning: [W030] Some entities could not be aligned in the text \"04bf0791804e8487c91ab84eaa47a335.pdf\teffective_dat...\" with entities \"[(198, 216, 'effective_date'), (22663, 22681, 'eff...\". Use `spacy.training.offsets_to_biluo_tags(nlp.make_doc(text), entities)` to check the alignment. Misaligned entities ('-') will be ignored during training.\n",
      "  warnings.warn(\n",
      "  3%|▎         | 8/254 [00:07<04:37,  1.13s/it]/Library/Frameworks/Python.framework/Versions/3.9/lib/python3.9/site-packages/spacy/training/iob_utils.py:141: UserWarning: [W030] Some entities could not be aligned in the text \"0587275477c6ad6d0d72419383e04b88.pdf\teffective_dat...\" with entities \"[(4528, 4536, 'jurisdiction'), (4604, 4612, 'juris...\". Use `spacy.training.offsets_to_biluo_tags(nlp.make_doc(text), entities)` to check the alignment. Misaligned entities ('-') will be ignored during training.\n",
      "  warnings.warn(\n",
      "  4%|▎         | 9/254 [00:12<09:04,  2.22s/it]/Library/Frameworks/Python.framework/Versions/3.9/lib/python3.9/site-packages/spacy/training/iob_utils.py:141: UserWarning: [W030] Some entities could not be aligned in the text \"05947711a24a5b7ce401911d31e19c91.pdf\teffective_dat...\" with entities \"[(18271, 18279, 'jurisdiction'), (18507, 18515, 'j...\". Use `spacy.training.offsets_to_biluo_tags(nlp.make_doc(text), entities)` to check the alignment. Misaligned entities ('-') will be ignored during training.\n",
      "  warnings.warn(\n",
      "  6%|▌         | 14/254 [00:18<04:18,  1.08s/it]/Library/Frameworks/Python.framework/Versions/3.9/lib/python3.9/site-packages/spacy/training/iob_utils.py:141: UserWarning: [W030] Some entities could not be aligned in the text \"0859334b76224ff82c1312ae7b2b5da1.pdf\teffective_dat...\" with entities \"[(279, 296, 'effective_date'), (22981, 22998, 'eff...\". Use `spacy.training.offsets_to_biluo_tags(nlp.make_doc(text), entities)` to check the alignment. Misaligned entities ('-') will be ignored during training.\n",
      "  warnings.warn(\n",
      "  7%|▋         | 17/254 [00:20<03:29,  1.13it/s]/Library/Frameworks/Python.framework/Versions/3.9/lib/python3.9/site-packages/spacy/training/iob_utils.py:141: UserWarning: [W030] Some entities could not be aligned in the text \"0c3ab1d0c8bb3b1c2f7a64f3ab584368.pdf\teffective_dat...\" with entities \"[(243, 259, 'effective_date'), (35225, 35241, 'eff...\". Use `spacy.training.offsets_to_biluo_tags(nlp.make_doc(text), entities)` to check the alignment. Misaligned entities ('-') will be ignored during training.\n",
      "  warnings.warn(\n",
      "  7%|▋         | 18/254 [00:23<04:38,  1.18s/it]/Library/Frameworks/Python.framework/Versions/3.9/lib/python3.9/site-packages/spacy/training/iob_utils.py:141: UserWarning: [W030] Some entities could not be aligned in the text \"0c7b90701575b147c4ac245ca478ee7c.pdf\teffective_dat...\" with entities \"[(10058, 10065, 'jurisdiction'), (10252, 10259, 'j...\". Use `spacy.training.offsets_to_biluo_tags(nlp.make_doc(text), entities)` to check the alignment. Misaligned entities ('-') will be ignored during training.\n",
      "  warnings.warn(\n",
      "  7%|▋         | 19/254 [00:25<05:25,  1.39s/it]"
     ]
    }
   ],
   "source": [
    "# import random\n",
    "from tqdm import tqdm\n",
    "\n",
    "from spacy.training.example import Example\n",
    "\n",
    "with nlp.disable_pipes(*otherPipes):\n",
    "    optimizer = nlp.begin_training()\n",
    "    for itn in range(nIter):\n",
    "        # random.shuffle(trainData)\n",
    "        losses = {}\n",
    "        for text, annotations in tqdm(trainData):\n",
    "            try:\n",
    "                doc = nlp.make_doc(text)\n",
    "                example = Example.from_dict(doc, annotations)\n",
    "                nlp.update([example], drop=0.5, sgd=optimizer, losses=losses)\n",
    "            except:\n",
    "                pass\n",
    "        print(losses)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "interpreter": {
   "hash": "aee8b7b246df8f9039afb4144a1f6fd8d2ca17a180786b69acc140d282b71a49"
  },
  "kernelspec": {
   "display_name": "Python 3.9.2 64-bit",
   "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": "3.9.2"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}