2-3-Language-Identifier-Using-Word-Bigrams.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Language Identifier Using Word Bigrams"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Based on [Language Identifier by asif31iqbal](https://github.com/asif31iqbal/language-identifier)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 0. Importing libraries and creating helper tokenize method"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pickle\n",
    "import string\n",
    "import os\n",
    "from nltk import ngrams, FreqDist, word_tokenize\n",
    "from numpy import arange\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def ultimate_tokenize(sentence):\n",
    "    # Remove punctuation and digits\n",
    "    sentence = sentence.translate(str.maketrans('', '', string.punctuation + string.digits))\n",
    "    return word_tokenize(sentence.lower())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. Understanding the process"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['oh', 'then', 'i', 'see', 'queen', 'mab', 'hath', 'been', 'with', 'you']"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "simple_example_text = 'Oh, then, I see Queen Mab hath been with you.'\n",
    "\n",
    "simple_example_tokens_words = ultimate_tokenize(simple_example_text)\n",
    "simple_example_tokens_words"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['o', 'h']"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "simple_example_tokens_chars = list(simple_example_tokens_words[0])\n",
    "simple_example_tokens_chars"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('oh',),\n",
       " ('then',),\n",
       " ('i',),\n",
       " ('see',),\n",
       " ('queen',),\n",
       " ('mab',),\n",
       " ('hath',),\n",
       " ('been',),\n",
       " ('with',),\n",
       " ('you',)]"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "simple_example_tokens_words_unigrams = list(ngrams(simple_example_tokens_words, 1))\n",
    "simple_example_tokens_words_unigrams"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('_', 'oh'),\n",
       " ('oh', 'then'),\n",
       " ('then', 'i'),\n",
       " ('i', 'see'),\n",
       " ('see', 'queen'),\n",
       " ('queen', 'mab'),\n",
       " ('mab', 'hath'),\n",
       " ('hath', 'been'),\n",
       " ('been', 'with'),\n",
       " ('with', 'you'),\n",
       " ('you', '_')]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "simple_example_tokens_words_bigrams = list(ngrams(simple_example_tokens_words, 2, pad_left=True, pad_right=True, left_pad_symbol='_', right_pad_symbol='_'))\n",
    "simple_example_tokens_words_bigrams"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "FreqDist({('been',): 1,\n",
       "          ('hath',): 1,\n",
       "          ('i',): 1,\n",
       "          ('mab',): 1,\n",
       "          ('oh',): 1,\n",
       "          ('queen',): 1,\n",
       "          ('see',): 1,\n",
       "          ('then',): 1,\n",
       "          ('with',): 1,\n",
       "          ('you',): 1})"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "fdist = FreqDist(simple_example_tokens_words_unigrams)\n",
    "fdist"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'been': 1,\n",
       " 'hath': 1,\n",
       " 'i': 1,\n",
       " 'mab': 1,\n",
       " 'oh': 1,\n",
       " 'queen': 1,\n",
       " 'see': 1,\n",
       " 'then': 1,\n",
       " 'with': 1,\n",
       " 'you': 1}"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "unigram_dict = dict()\n",
    "for k, v in fdist.items():\n",
    "        unigram_dict[' '.join(k)] = v\n",
    "unigram_dict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "\"approval of the minutes of the previous sitting the minutes of yesterday 's sitting have been distri\""
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "file = 'ngram_langid_files/LangId.train.English.txt'\n",
    "with open(file, encoding='utf8') as f:\n",
    "        content = f.read().lower()\n",
    "content.replace('\\n', '')[:100]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'approval': 3,\n",
       " 'of': 2769,\n",
       " 'the': 5699,\n",
       " 'minutes': 11,\n",
       " 'previous': 13,\n",
       " 'sitting': 11,\n",
       " 'yesterday': 15,\n",
       " 's': 244,\n",
       " 'have': 491,\n",
       " 'been': 193,\n",
       " 'distributed': 4,\n",
       " 'are': 571,\n",
       " 'there': 229,\n",
       " 'any': 98,\n",
       " 'comments': 9,\n",
       " 'mr': 371,\n",
       " 'president': 270,\n",
       " 'on': 865,\n",
       " 'monday': 11,\n",
       " 'i': 721,\n",
       " 'made': 61,\n",
       " 'a': 1343,\n",
       " 'point': 70,\n",
       " 'order': 76,\n",
       " 'about': 93,\n",
       " 'nicole': 1,\n",
       " 'fontaine': 4,\n",
       " 'reported': 2,\n",
       " 'in': 1668,\n",
       " 'british': 16,\n",
       " 'press': 13,\n",
       " 'regarding': 23,\n",
       " 'her': 51,\n",
       " 'recent': 12,\n",
       " 'visit': 6,\n",
       " 'with': 449,\n",
       " 'majesty': 1,\n",
       " 'queen': 3,\n",
       " 'elizabeth': 1,\n",
       " 'ii': 1,\n",
       " 'labour': 14,\n",
       " 'member': 111,\n",
       " 'this': 887,\n",
       " 'house': 49,\n",
       " 'miller': 2,\n",
       " 'repeated': 6,\n",
       " 'what': 132,\n",
       " 'were': 82,\n",
       " 'purported': 1,\n",
       " 'to': 2490,\n",
       " 'be': 620,\n",
       " 'remarks': 1,\n",
       " 'not': 525,\n",
       " 'once': 27,\n",
       " 'but': 240,\n",
       " 'three': 21,\n",
       " 'times': 11,\n",
       " 'tuesday': 8,\n",
       " 'and': 2040,\n",
       " 'wednesday': 2,\n",
       " 'he': 72,\n",
       " 'sought': 5,\n",
       " 'drag': 1,\n",
       " 'into': 78,\n",
       " 'political': 51,\n",
       " 'controversy': 2,\n",
       " 'use': 45,\n",
       " 'name': 16,\n",
       " 'score': 2,\n",
       " 'cheap': 7,\n",
       " 'points': 23,\n",
       " 'it': 694,\n",
       " 'may': 59,\n",
       " 'that': 1205,\n",
       " 'was': 163,\n",
       " 'unfamiliar': 1,\n",
       " 'rules': 25,\n",
       " 'protocol': 3,\n",
       " 'which': 496,\n",
       " 'obtain': 6,\n",
       " 'such': 102,\n",
       " 'matters': 26,\n",
       " 'has': 339,\n",
       " 'no': 116,\n",
       " 'excuse': 3,\n",
       " 'knows': 8,\n",
       " 'perfectly': 8,\n",
       " 'well': 79,\n",
       " 'ours': 3,\n",
       " 'is': 1303,\n",
       " 'non': 38,\n",
       " 'head': 10,\n",
       " 'state': 39,\n",
       " 'for': 821,\n",
       " 'nearly': 2,\n",
       " 'fifty': 3,\n",
       " 'years': 69,\n",
       " 'she': 32,\n",
       " 'scrupulously': 1,\n",
       " 'avoided': 3,\n",
       " 'engaging': 1,\n",
       " 'controversial': 2,\n",
       " 'issues': 26,\n",
       " 'can': 220,\n",
       " 'come': 30,\n",
       " 'set': 31,\n",
       " 'record': 6,\n",
       " 'straight': 2,\n",
       " 'his': 57,\n",
       " 'behaviour': 15,\n",
       " 'disgrace': 1,\n",
       " 'scandal': 3,\n",
       " 'am': 80,\n",
       " 'proud': 2,\n",
       " 'today': 64,\n",
       " 'ashamed': 1,\n",
       " 'share': 14,\n",
       " 'my': 109,\n",
       " 'nationality': 3,\n",
       " 'man': 6,\n",
       " 'shame': 2,\n",
       " 'respect': 39,\n",
       " 'decency': 1,\n",
       " 'or': 177,\n",
       " 'honour': 3,\n",
       " 'at': 274,\n",
       " 'all': 222,\n",
       " 'wish': 40,\n",
       " 'speak': 9,\n",
       " 'capacity': 12,\n",
       " 'as': 552,\n",
       " 'chairman': 9,\n",
       " 'delegation': 8,\n",
       " 'relations': 15,\n",
       " 'countries': 66,\n",
       " 'central': 13,\n",
       " 'america': 14,\n",
       " 'mexico': 3,\n",
       " 'you': 117,\n",
       " 'will': 277,\n",
       " 'aware': 24,\n",
       " 'elections': 4,\n",
       " 'held': 10,\n",
       " 'last': 56,\n",
       " 'sunday': 1,\n",
       " 'chair': 1,\n",
       " 'met': 5,\n",
       " 'discussed': 6,\n",
       " 'report': 286,\n",
       " 'drawn': 6,\n",
       " 'up': 134,\n",
       " 'by': 453,\n",
       " 'our': 208,\n",
       " 'fellow': 18,\n",
       " 'members': 60,\n",
       " 'who': 153,\n",
       " 'had': 53,\n",
       " 'act': 13,\n",
       " 'observers': 4,\n",
       " 'these': 162,\n",
       " 'behalf': 87,\n",
       " 'parliament': 189,\n",
       " 'would': 273,\n",
       " 'ask': 41,\n",
       " 'see': 45,\n",
       " 'if': 155,\n",
       " 'could': 64,\n",
       " 'forward': 22,\n",
       " 'resolution': 119,\n",
       " 'complete': 10,\n",
       " 'text': 34,\n",
       " 'receive': 16,\n",
       " 'mexican': 1,\n",
       " 'authorities': 40,\n",
       " 'ladies': 33,\n",
       " 'gentlemen': 33,\n",
       " 'want': 48,\n",
       " 'draw': 12,\n",
       " 'your': 26,\n",
       " 'attention': 31,\n",
       " 'something': 25,\n",
       " 'really': 30,\n",
       " 'annoying': 1,\n",
       " 'connection': 10,\n",
       " 'x': 49,\n",
       " 'texts': 7,\n",
       " 'adopted': 34,\n",
       " 'also': 221,\n",
       " 'form': 19,\n",
       " 'part': 64,\n",
       " 'pointed': 11,\n",
       " 'out': 105,\n",
       " 'during': 60,\n",
       " 'vote': 91,\n",
       " 'directive': 77,\n",
       " 'combat': 6,\n",
       " 'money': 16,\n",
       " 'laundering': 1,\n",
       " 'english': 9,\n",
       " 'translation': 3,\n",
       " 'article': 21,\n",
       " 'needs': 37,\n",
       " 'corrected': 2,\n",
       " 'bring': 17,\n",
       " 'line': 17,\n",
       " 'amendment': 38,\n",
       " 'because': 118,\n",
       " 'wrong': 11,\n",
       " 'middle': 3,\n",
       " 'second': 25,\n",
       " 'paragraph': 8,\n",
       " 'version': 6,\n",
       " 'only': 113,\n",
       " 'reference': 9,\n",
       " 'representation': 3,\n",
       " 'legal': 32,\n",
       " 'proceedings': 8,\n",
       " 'consultants': 1,\n",
       " 'german': 3,\n",
       " 'key': 14,\n",
       " 'since': 40,\n",
       " 'council': 85,\n",
       " 'working': 49,\n",
       " 'group': 131,\n",
       " 'basis': 28,\n",
       " 'quite': 28,\n",
       " 'crucial': 10,\n",
       " 'correct': 13,\n",
       " 'already': 41,\n",
       " 'complained': 2,\n",
       " 'committee': 134,\n",
       " 'vis': 4,\n",
       " 'language': 4,\n",
       " 'services': 46,\n",
       " 'time': 123,\n",
       " 'decided': 10,\n",
       " 'table': 2,\n",
       " 'an': 266,\n",
       " 'original': 5,\n",
       " 'drafted': 4,\n",
       " 'rapporteur': 49,\n",
       " 'situation': 67,\n",
       " 'remedied': 2,\n",
       " 'third': 12,\n",
       " 'complaining': 2,\n",
       " 'do': 149,\n",
       " 'think': 41,\n",
       " 'should': 222,\n",
       " 'work': 118,\n",
       " 'more': 176,\n",
       " 'carefully': 8,\n",
       " 'questions': 12,\n",
       " 'intolerable': 4,\n",
       " 'despite': 14,\n",
       " 'criticism': 9,\n",
       " 'got': 11,\n",
       " 'again': 36,\n",
       " 'get': 18,\n",
       " 'bottom': 2,\n",
       " 'matter': 57,\n",
       " 'sort': 11,\n",
       " 'most': 63,\n",
       " 'important': 91,\n",
       " 'aspect': 11,\n",
       " 'please': 7,\n",
       " 'we': 864,\n",
       " 'received': 15,\n",
       " 'communication': 18,\n",
       " 'yellow': 1,\n",
       " 'paper': 6,\n",
       " 'from': 226,\n",
       " 'vice': 18,\n",
       " 'some': 85,\n",
       " 'nuts': 3,\n",
       " 'attached': 3,\n",
       " 'note': 17,\n",
       " 'supporting': 7,\n",
       " 'export': 3,\n",
       " 'products': 1,\n",
       " 'vanuatu': 1,\n",
       " 'developing': 8,\n",
       " 'country': 36,\n",
       " 'pacific': 3,\n",
       " 'very': 129,\n",
       " 'good': 68,\n",
       " 'idea': 12,\n",
       " 'right': 80,\n",
       " 'happy': 8,\n",
       " 'support': 98,\n",
       " 'rather': 31,\n",
       " 'puzzles': 1,\n",
       " 'me': 58,\n",
       " 'explicitly': 1,\n",
       " 'stated': 9,\n",
       " 'illegal': 11,\n",
       " 'import': 2,\n",
       " 'european': 389,\n",
       " 'union': 120,\n",
       " 'so': 163,\n",
       " 'must': 228,\n",
       " 'explain': 6,\n",
       " 'how': 49,\n",
       " 'came': 3,\n",
       " 'illegally': 1,\n",
       " 'imported': 2,\n",
       " 'delivered': 1,\n",
       " 'them': 83,\n",
       " 'us': 100,\n",
       " 'itself': 20,\n",
       " 'contradiction': 3,\n",
       " 'clarify': 4,\n",
       " 'laughter': 1,\n",
       " 'applause': 3,\n",
       " 'approved': 11,\n",
       " 'welcome': 12,\n",
       " 'petitions': 71,\n",
       " 'ombudsman': 110,\n",
       " 'next': 29,\n",
       " 'item': 17,\n",
       " 'joint': 25,\n",
       " 'debate': 58,\n",
       " 'following': 20,\n",
       " 'reports': 31,\n",
       " 'tabled': 14,\n",
       " 'mrs': 117,\n",
       " 'lambert': 14,\n",
       " 'deliberations': 3,\n",
       " 'parliamentary': 13,\n",
       " 'year': 67,\n",
       " 'thors': 12,\n",
       " 'annual': 12,\n",
       " 'activities': 19,\n",
       " 'glad': 3,\n",
       " 'opportunity': 19,\n",
       " 'address': 17,\n",
       " 'concerning': 15,\n",
       " 'results': 15,\n",
       " 'covers': 2,\n",
       " 'first': 60,\n",
       " 'mandate': 10,\n",
       " 'office': 17,\n",
       " 'new': 69,\n",
       " 'complaints': 32,\n",
       " 'compared': 4,\n",
       " 'increase': 33,\n",
       " 'number': 67,\n",
       " 'inquiries': 6,\n",
       " 'opened': 3,\n",
       " 'rose': 2,\n",
       " 'over': 52,\n",
       " 'becoming': 15,\n",
       " 'better': 25,\n",
       " 'known': 19,\n",
       " 'attracting': 1,\n",
       " 'higher': 3,\n",
       " 'proportion': 7,\n",
       " 'serious': 36,\n",
       " 'require': 8,\n",
       " 'full': 25,\n",
       " 'investigation': 9,\n",
       " 'interested': 9,\n",
       " 'know': 53,\n",
       " 'same': 46,\n",
       " 'trend': 6,\n",
       " 'continued': 6,\n",
       " 'present': 38,\n",
       " 'six': 9,\n",
       " 'months': 14,\n",
       " 'period': 24,\n",
       " 'resource': 2,\n",
       " 'intensive': 1,\n",
       " 'therefore': 98,\n",
       " 'great': 47,\n",
       " 'challenge': 3,\n",
       " 'deal': 21,\n",
       " 'whilst': 3,\n",
       " 'maintaining': 4,\n",
       " 'quality': 16,\n",
       " 'meeting': 7,\n",
       " 'target': 2,\n",
       " 'closing': 2,\n",
       " 'within': 60,\n",
       " 'one': 131,\n",
       " 'hope': 36,\n",
       " 'seeking': 5,\n",
       " 'resources': 16,\n",
       " 'needed': 10,\n",
       " 'carry': 7,\n",
       " 'effectively': 13,\n",
       " 'high': 36,\n",
       " 'lay': 4,\n",
       " 'outside': 14,\n",
       " 'even': 64,\n",
       " 'increased': 7,\n",
       " 'slightly': 2,\n",
       " 'probably': 11,\n",
       " 'growing': 13,\n",
       " 'e': 6,\n",
       " 'mail': 6,\n",
       " 'gladly': 1,\n",
       " 'accept': 16,\n",
       " 'many': 74,\n",
       " 'contain': 6,\n",
       " 'presented': 15,\n",
       " 'allegations': 2,\n",
       " 'maladministration': 8,\n",
       " 'however': 83,\n",
       " 'directed': 6,\n",
       " 'against': 82,\n",
       " 'states': 144,\n",
       " 'campaign': 8,\n",
       " 'election': 3,\n",
       " 'critical': 12,\n",
       " 'voices': 2,\n",
       " 'suggested': 5,\n",
       " 'broadened': 1,\n",
       " 'include': 16,\n",
       " 'cases': 32,\n",
       " 'where': 58,\n",
       " 'question': 30,\n",
       " 'law': 40,\n",
       " 'involved': 21,\n",
       " 'sure': 18,\n",
       " 'whether': 28,\n",
       " 'they': 190,\n",
       " 'understood': 2,\n",
       " 'mean': 17,\n",
       " 'carrying': 6,\n",
       " 'tasks': 3,\n",
       " 'currently': 20,\n",
       " 'belonging': 1,\n",
       " 'commission': 274,\n",
       " 'guardian': 1,\n",
       " 'treaty': 24,\n",
       " 'democratic': 9,\n",
       " 'forum': 2,\n",
       " 'citizens': 92,\n",
       " 'study': 7,\n",
       " 'especially': 29,\n",
       " 'involve': 5,\n",
       " 'rights': 81,\n",
       " 'under': 47,\n",
       " 'managed': 8,\n",
       " 'advise': 1,\n",
       " 'complainant': 3,\n",
       " 'transfer': 2,\n",
       " 'complaint': 8,\n",
       " 'competent': 3,\n",
       " 'body': 16,\n",
       " 'half': 5,\n",
       " 'transferred': 2,\n",
       " 'dealt': 7,\n",
       " 'advised': 2,\n",
       " 'further': 38,\n",
       " 'complainants': 1,\n",
       " 'their': 190,\n",
       " 'petition': 10,\n",
       " 'national': 69,\n",
       " 'regional': 21,\n",
       " 'local': 13,\n",
       " 'concerned': 40,\n",
       " 'every': 32,\n",
       " 'judicial': 4,\n",
       " 'assist': 2,\n",
       " 'conflicts': 12,\n",
       " 'administration': 35,\n",
       " 'believer': 1,\n",
       " 'subsidiarity': 7,\n",
       " 'remain': 14,\n",
       " 'convinced': 9,\n",
       " 'effective': 28,\n",
       " 'way': 86,\n",
       " 'achieve': 21,\n",
       " 'result': 27,\n",
       " 'through': 29,\n",
       " 'development': 37,\n",
       " 'network': 4,\n",
       " 'cooperation': 45,\n",
       " 'ombudsmen': 11,\n",
       " 'similar': 13,\n",
       " 'bodies': 27,\n",
       " 'committees': 11,\n",
       " 'organised': 4,\n",
       " 'jointly': 7,\n",
       " 'french': 16,\n",
       " 'stasi': 1,\n",
       " 'seminar': 3,\n",
       " 'paris': 6,\n",
       " 'september': 8,\n",
       " 'perry': 2,\n",
       " 'represented': 5,\n",
       " 'planned': 4,\n",
       " 'take': 102,\n",
       " 'place': 31,\n",
       " 'brussels': 8,\n",
       " 'belgian': 6,\n",
       " 'presidency': 15,\n",
       " 'belgium': 3,\n",
       " 'federal': 7,\n",
       " 'invited': 3,\n",
       " 'florence': 1,\n",
       " 'gemelli': 3,\n",
       " 'attended': 3,\n",
       " 'shown': 12,\n",
       " 'positive': 17,\n",
       " 'attitude': 13,\n",
       " 'towards': 25,\n",
       " 'strongly': 10,\n",
       " 'believe': 61,\n",
       " 'equal': 6,\n",
       " 'terms': 49,\n",
       " 'best': 15,\n",
       " 'beginning': 8,\n",
       " 'doubts': 9,\n",
       " 'expressed': 12,\n",
       " 'sufficient': 11,\n",
       " 'powers': 8,\n",
       " 'anything': 10,\n",
       " 'citizen': 12,\n",
       " 'unjustified': 1,\n",
       " 'community': 65,\n",
       " 'institutions': 47,\n",
       " 'responded': 2,\n",
       " 'properly': 8,\n",
       " 'themselves': 16,\n",
       " 'settled': 1,\n",
       " 'satisfaction': 5,\n",
       " 'example': 32,\n",
       " 'replied': 3,\n",
       " 'unanswered': 1,\n",
       " 'letters': 1,\n",
       " 'payments': 2,\n",
       " 'due': 17,\n",
       " 'reversed': 3,\n",
       " 'unfavourable': 1,\n",
       " 'decisions': 22,\n",
       " 'indicates': 4,\n",
       " 'spirit': 6,\n",
       " 'responsiveness': 1,\n",
       " 'flexibility': 6,\n",
       " 'superior': 1,\n",
       " 'those': 93,\n",
       " 'administrations': 9,\n",
       " 'experience': 14,\n",
       " 'sixty': 1,\n",
       " 'two': 49,\n",
       " 'settlements': 1,\n",
       " 'occurred': 5,\n",
       " 'reducing': 8,\n",
       " 'scope': 5,\n",
       " 'friendly': 11,\n",
       " 'solutions': 13,\n",
       " 'proposed': 19,\n",
       " 'did': 25,\n",
       " 'solution': 20,\n",
       " 'case': 54,\n",
       " 'agreed': 12,\n",
       " 'paid': 8,\n",
       " 'amount': 6,\n",
       " 'him': 20,\n",
       " 'coordinator': 1,\n",
       " 'monitoring': 35,\n",
       " 'palestinian': 1,\n",
       " 'institution': 21,\n",
       " 'ultimate': 2,\n",
       " 'weapon': 3,\n",
       " 'draft': 19,\n",
       " 'recommendation': 14,\n",
       " 'followed': 5,\n",
       " 'necessary': 48,\n",
       " 'special': 18,\n",
       " 'five': 13,\n",
       " 'ago': 19,\n",
       " 'people': 126,\n",
       " 'said': 38,\n",
       " 'pay': 11,\n",
       " 'much': 59,\n",
       " 'heed': 1,\n",
       " 'might': 18,\n",
       " 'then': 36,\n",
       " 'things': 18,\n",
       " 'clearly': 28,\n",
       " 'changed': 6,\n",
       " 'amsterdam': 3,\n",
       " 'fall': 3,\n",
       " 'santer': 1,\n",
       " 'ten': 10,\n",
       " 'recommendations': 3,\n",
       " 'seven': 4,\n",
       " 'accepted': 9,\n",
       " 'remains': 6,\n",
       " 'ongoing': 2,\n",
       " 'europol': 3,\n",
       " 'comply': 5,\n",
       " 'approve': 4,\n",
       " 'public': 68,\n",
       " 'access': 28,\n",
       " 'documents': 20,\n",
       " 'reached': 4,\n",
       " 'level': 44,\n",
       " 'expect': 6,\n",
       " 'consists': 4,\n",
       " 'persons': 3,\n",
       " 'fairly': 3,\n",
       " 'complex': 6,\n",
       " 'environment': 19,\n",
       " 'achieving': 6,\n",
       " 'above': 10,\n",
       " 'like': 144,\n",
       " 'thank': 35,\n",
       " 'another': 30,\n",
       " 'constructive': 4,\n",
       " 'its': 146,\n",
       " 'responsible': 22,\n",
       " 'loyola': 1,\n",
       " 'de': 23,\n",
       " 'palacio': 13,\n",
       " 'continuing': 4,\n",
       " 'predecessor': 2,\n",
       " 'commitment': 4,\n",
       " 'honest': 2,\n",
       " 'express': 21,\n",
       " 'warm': 4,\n",
       " 'appreciation': 3,\n",
       " 'helpful': 2,\n",
       " 'cooperative': 1,\n",
       " 'comprehensive': 6,\n",
       " 'presenting': 3,\n",
       " 'morning': 11,\n",
       " 'dual': 3,\n",
       " 'reflects': 6,\n",
       " 'mediator': 1,\n",
       " 'transnationally': 1,\n",
       " 'abducted': 2,\n",
       " 'children': 40,\n",
       " 'banotti': 1,\n",
       " 'personal': 6,\n",
       " 'soederman': 19,\n",
       " 'secretariat': 3,\n",
       " 'least': 30,\n",
       " 'statistical': 3,\n",
       " 'break': 9,\n",
       " 'down': 26,\n",
       " 'provided': 17,\n",
       " 'confess': 1,\n",
       " 'own': 41,\n",
       " 'urge': 17,\n",
       " 'familiar': 3,\n",
       " 'read': 5,\n",
       " 'providing': 3,\n",
       " 'service': 29,\n",
       " 'represent': 6,\n",
       " 'website': 3,\n",
       " 'mentioned': 14,\n",
       " 'europe': 96,\n",
       " 'direct': 9,\n",
       " 'initiative': 11,\n",
       " 'provide': 34,\n",
       " 'useful': 5,\n",
       " 'information': 46,\n",
       " 'possibly': 8,\n",
       " 'avoid': 8,\n",
       " 'need': 125,\n",
       " 'indeed': 29,\n",
       " 'confirm': 7,\n",
       " 'appropriate': 15,\n",
       " 'vehicle': 3,\n",
       " 'become': 21,\n",
       " 'importance': 25,\n",
       " 'relevance': 1,\n",
       " 'aim': 13,\n",
       " 'individuals': 5,\n",
       " 'civil': 23,\n",
       " 'society': 21,\n",
       " 'generally': 5,\n",
       " 'individual': 17,\n",
       " 'resident': 3,\n",
       " 'feel': 20,\n",
       " 'principles': 11,\n",
       " 'being': 62,\n",
       " 'applied': 13,\n",
       " 'whatever': 5,\n",
       " 'considered': 12,\n",
       " 'admissible': 2,\n",
       " 'mention': 10,\n",
       " 'just': 69,\n",
       " 'demonstrate': 3,\n",
       " 'range': 4,\n",
       " 'topics': 3,\n",
       " 'potential': 6,\n",
       " 'effectiveness': 5,\n",
       " 'mechanism': 10,\n",
       " 'petitioner': 1,\n",
       " 'succeeded': 1,\n",
       " 'getting': 1,\n",
       " 'teaching': 2,\n",
       " 'recognised': 8,\n",
       " 'presumably': 2,\n",
       " 'affecting': 4,\n",
       " 'salary': 5,\n",
       " 'promotion': 8,\n",
       " 'prospects': 1,\n",
       " 'spanish': 3,\n",
       " 'acting': 4,\n",
       " 'environmental': 30,\n",
       " 'association': 1,\n",
       " 'resulted': 4,\n",
       " 'infringement': 3,\n",
       " 'brought': 7,\n",
       " 'relation': 17,\n",
       " 'habitats': 1,\n",
       " 'frequent': 7,\n",
       " 'cause': 21,\n",
       " 'lasting': 3,\n",
       " 'effect': 8,\n",
       " 'valuable': 9,\n",
       " 'feedback': 2,\n",
       " 'other': 100,\n",
       " 'legislation': 22,\n",
       " 'works': 5,\n",
       " 'benefit': 10,\n",
       " 'without': 33,\n",
       " 'numerous': 5,\n",
       " 'officials': 6,\n",
       " 'investigate': 3,\n",
       " 'instigate': 1,\n",
       " 'thanks': 8,\n",
       " 'rapid': 3,\n",
       " 'equally': 5,\n",
       " 'committed': 4,\n",
       " 'response': 12,\n",
       " 'replying': 1,\n",
       " 'requests': 4,\n",
       " 'explanation': 11,\n",
       " 'dealing': 5,\n",
       " 'taken': 54,\n",
       " 'reply': 11,\n",
       " 'letter': 8,\n",
       " 'unusual': 2,\n",
       " 'why': 51,\n",
       " 'recommending': 2,\n",
       " 'relevant': 11,\n",
       " 'permanent': 2,\n",
       " 'meetings': 2,\n",
       " 'develop': 6,\n",
       " 'greater': 32,\n",
       " 'understanding': 4,\n",
       " 'problems': 49,\n",
       " 'face': 9,\n",
       " 'respond': 8,\n",
       " 'here': 53,\n",
       " 'improve': 27,\n",
       " 'overall': 9,\n",
       " 'mechanisms': 6,\n",
       " 'ideas': 8,\n",
       " 'others': 28,\n",
       " 'appear': 7,\n",
       " 'kessler': 1,\n",
       " 'proposing': 4,\n",
       " 'database': 1,\n",
       " 'petitioners': 3,\n",
       " 'track': 2,\n",
       " 'progress': 31,\n",
       " 'lie': 2,\n",
       " 'trust': 4,\n",
       " 'bear': 5,\n",
       " 'mind': 10,\n",
       " 'proposal': 38,\n",
       " 'various': 22,\n",
       " 'call': 30,\n",
       " 'often': 25,\n",
       " 'opinion': 36,\n",
       " 'certain': 44,\n",
       " 'areas': 36,\n",
       " 'concern': 34,\n",
       " 'arisen': 1,\n",
       " 'preparing': 4,\n",
       " 'particular': 45,\n",
       " 'social': 84,\n",
       " 'security': 28,\n",
       " 'consumer': 14,\n",
       " 'protection': 45,\n",
       " 'institute': 1,\n",
       " 'style': 2,\n",
       " 'system': 45,\n",
       " 'specific': 36,\n",
       " 'keen': 2,\n",
       " 'establish': 14,\n",
       " 'links': 4,\n",
       " 'view': 55,\n",
       " 'exchanging': 1,\n",
       " 'consider': 23,\n",
       " 'empowering': 1,\n",
       " 'tool': 2,\n",
       " 'democracy': 20,\n",
       " 'remote': 5,\n",
       " 'anonymous': 1,\n",
       " 'value': 8,\n",
       " 'whole': 32,\n",
       " 'commend': 3,\n",
       " 'commissioner': 56,\n",
       " 'characterised': 2,\n",
       " 'crises': 3,\n",
       " 'eu': 52,\n",
       " 'immense': 1,\n",
       " 'distrust': 2,\n",
       " 'integration': 9,\n",
       " 'legitimate': 7,\n",
       " 'actively': 2,\n",
       " 'participate': 3,\n",
       " 'dialogue': 22,\n",
       " 'real': 30,\n",
       " 'opinions': 3,\n",
       " 'seriously': 11,\n",
       " 'task': 14,\n",
       " 'course': 61,\n",
       " 'according': 23,\n",
       " 'request': 20,\n",
       " 'gave': 7,\n",
       " 'definition': 8,\n",
       " 'term': 10,\n",
       " 'occur': 4,\n",
       " 'when': 87,\n",
       " 'fails': 5,\n",
       " 'accordance': 7,\n",
       " 'rule': 14,\n",
       " 'principle': 36,\n",
       " 'binding': 6,\n",
       " 'upon': 18,\n",
       " 'apply': 13,\n",
       " 'now': 85,\n",
       " 'future': 37,\n",
       " 'activity': 8,\n",
       " 'stresses': 4,\n",
       " 'involves': 7,\n",
       " 'inspect': 2,\n",
       " 'interpreted': 3,\n",
       " 'irrespective': 2,\n",
       " 'charter': 13,\n",
       " 'fundamental': 25,\n",
       " 'expression': 6,\n",
       " 'common': 29,\n",
       " 'constitutional': 5,\n",
       " 'traditions': 3,\n",
       " 'shall': 15,\n",
       " 'respected': 6,\n",
       " 'observes': 1,\n",
       " 'ought': 12,\n",
       " 'play': 13,\n",
       " 'implementing': 9,\n",
       " 'clear': 34,\n",
       " 'fact': 79,\n",
       " 'give': 38,\n",
       " 'opposite': 6,\n",
       " 'both': 46,\n",
       " 'code': 6,\n",
       " 'administrative': 18,\n",
       " 'heard': 5,\n",
       " 'prompt': 3,\n",
       " 'answers': 1,\n",
       " 'replies': 3,\n",
       " 'supported': 12,\n",
       " 'reasons': 17,\n",
       " 'aspects': 17,\n",
       " 'prerequisites': 1,\n",
       " 'active': 12,\n",
       " 'citizenship': 4,\n",
       " 'regain': 2,\n",
       " 'confidence': 10,\n",
       " 'make': 72,\n",
       " 'instead': 7,\n",
       " 'passive': 1,\n",
       " 'create': 19,\n",
       " 'conditions': 29,\n",
       " 'whereas': 6,\n",
       " 'undermine': 5,\n",
       " 'categories': 2,\n",
       " 'incomplete': 3,\n",
       " 'refused': 5,\n",
       " 'delays': 50,\n",
       " 'discrimination': 15,\n",
       " 'lack': 26,\n",
       " 'defence': 8,\n",
       " 'conclusions': 13,\n",
       " 'ever': 19,\n",
       " 'increasing': 8,\n",
       " 'nature': 9,\n",
       " 'renowned': 3,\n",
       " 'wider': 4,\n",
       " 'lot': 12,\n",
       " 'praise': 3,\n",
       " 'completed': 2,\n",
       " 'few': 41,\n",
       " 'having': 16,\n",
       " 'allows': 3,\n",
       " 'tirelessly': 1,\n",
       " 'transparency': 22,\n",
       " 'longer': 16,\n",
       " 'possible': 45,\n",
       " 'data': 7,\n",
       " 'secrecy': 5,\n",
       " 'used': 29,\n",
       " 'prevent': 12,\n",
       " 'scrutiny': 2,\n",
       " 'step': 14,\n",
       " 'developed': 10,\n",
       " 'endeavoured': 2,\n",
       " 'gained': 2,\n",
       " 'corresponding': 1,\n",
       " 'candidate': 8,\n",
       " 'substantial': 3,\n",
       " 'reason': 30,\n",
       " 'making': 26,\n",
       " 'financial': 33,\n",
       " 'available': 20,\n",
       " 'investigations': 5,\n",
       " 'requires': 11,\n",
       " 'independent': 12,\n",
       " 'decision': 23,\n",
       " 'big': 4,\n",
       " 'difficult': 18,\n",
       " 'receives': 1,\n",
       " 'speakers': 6,\n",
       " 'rightly': 9,\n",
       " 'restricted': 6,\n",
       " 'draftsman': 1,\n",
       " 'residence': 3,\n",
       " 'area': 38,\n",
       " 'free': 17,\n",
       " 'movement': 15,\n",
       " 'basic': 14,\n",
       " 'freedoms': 4,\n",
       " 'achieved': 10,\n",
       " 'maastricht': 4,\n",
       " 'general': 41,\n",
       " 'adequate': 14,\n",
       " 'means': 47,\n",
       " 'subsistence': 1,\n",
       " 'health': 52,\n",
       " 'insurance': 3,\n",
       " 'freedom': 15,\n",
       " 'practice': 20,\n",
       " 'comes': 8,\n",
       " 'slow': 6,\n",
       " 'transpose': 2,\n",
       " 'b': 56,\n",
       " 'arise': 2,\n",
       " 'recognition': 6,\n",
       " 'diplomas': 2,\n",
       " 'qualifications': 3,\n",
       " 'professions': 1,\n",
       " 'c': 19,\n",
       " 'standardised': 1,\n",
       " 'eg': 1,\n",
       " 'complicated': 3,\n",
       " 'forms': 4,\n",
       " 'refund': 1,\n",
       " 'medical': 3,\n",
       " 'costs': 17,\n",
       " 'analysis': 4,\n",
       " 'between': 91,\n",
       " 'shows': 4,\n",
       " 'link': 11,\n",
       " 'before': 38,\n",
       " 'court': 9,\n",
       " 'justice': 14,\n",
       " 'unfortunately': 17,\n",
       " 'ignores': 1,\n",
       " 'communications': 5,\n",
       " 'violations': 1,\n",
       " 'negative': 8,\n",
       " 'devote': 2,\n",
       " 'complain': 3,\n",
       " 'invention': 1,\n",
       " 'put': 53,\n",
       " 'concerns': 14,\n",
       " 'roman': 1,\n",
       " 'emperor': 1,\n",
       " 'always': 26,\n",
       " 'danger': 4,\n",
       " 'abolished': 5,\n",
       " 'infallible': 1,\n",
       " 'discrepancy': 1,\n",
       " 'provisions': 12,\n",
       " 'gauge': 2,\n",
       " 'legislator': 2,\n",
       " 'levels': 10,\n",
       " 'population': 19,\n",
       " 'hamsters': 1,\n",
       " 'industrial': 5,\n",
       " 'ban': 11,\n",
       " 'silicon': 1,\n",
       " 'implants': 1,\n",
       " 'introduction': 8,\n",
       " 'heavy': 1,\n",
       " 'goods': 9,\n",
       " 'tax': 12,\n",
       " 'turn': 17,\n",
       " 'congratulates': 2,\n",
       " 'accomplished': 1,\n",
       " ...}"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "with open('ngram_langid_files/English.unigram.pickle', 'rb') as handle:\n",
    "    unigram_english_dict = pickle.load(handle)\n",
    "unigram_english_dict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'_ approval': 1,\n",
       " 'approval of': 3,\n",
       " 'of the': 906,\n",
       " 'the minutes': 6,\n",
       " 'minutes of': 2,\n",
       " 'the previous': 9,\n",
       " 'previous sitting': 2,\n",
       " 'sitting the': 1,\n",
       " 'of yesterday': 2,\n",
       " 'yesterday s': 5,\n",
       " 's sitting': 1,\n",
       " 'sitting have': 1,\n",
       " 'have been': 65,\n",
       " 'been distributed': 1,\n",
       " 'distributed are': 1,\n",
       " 'are there': 5,\n",
       " 'there any': 1,\n",
       " 'any comments': 1,\n",
       " 'comments mr': 1,\n",
       " 'mr president': 192,\n",
       " 'president on': 8,\n",
       " 'on monday': 10,\n",
       " 'monday i': 1,\n",
       " 'i made': 1,\n",
       " 'made a': 3,\n",
       " 'a point': 8,\n",
       " 'point of': 17,\n",
       " 'of order': 6,\n",
       " 'order about': 1,\n",
       " 'about president': 1,\n",
       " 'president nicole': 1,\n",
       " 'nicole fontaine': 1,\n",
       " 'fontaine s': 1,\n",
       " 's reported': 1,\n",
       " 'reported comments': 1,\n",
       " 'comments in': 1,\n",
       " 'in the': 442,\n",
       " 'the british': 9,\n",
       " 'british press': 1,\n",
       " 'press regarding': 1,\n",
       " 'regarding her': 1,\n",
       " 'her recent': 1,\n",
       " 'recent visit': 2,\n",
       " 'visit with': 1,\n",
       " 'with her': 2,\n",
       " 'her majesty': 1,\n",
       " 'majesty queen': 1,\n",
       " 'queen elizabeth': 1,\n",
       " 'elizabeth ii': 1,\n",
       " 'ii a': 1,\n",
       " 'a british': 1,\n",
       " 'british labour': 1,\n",
       " 'labour member': 1,\n",
       " 'member of': 11,\n",
       " 'of this': 79,\n",
       " 'this house': 32,\n",
       " 'house mr': 4,\n",
       " 'mr miller': 2,\n",
       " 'miller repeated': 1,\n",
       " 'repeated what': 1,\n",
       " 'what were': 1,\n",
       " 'were purported': 1,\n",
       " 'purported to': 1,\n",
       " 'to be': 176,\n",
       " 'be the': 18,\n",
       " 'the queen': 2,\n",
       " 'queen s': 1,\n",
       " 's remarks': 1,\n",
       " 'remarks not': 1,\n",
       " 'not once': 1,\n",
       " 'once but': 1,\n",
       " 'but three': 1,\n",
       " 'three times': 1,\n",
       " 'times on': 1,\n",
       " 'monday on': 1,\n",
       " 'on tuesday': 6,\n",
       " 'tuesday and': 1,\n",
       " 'and on': 6,\n",
       " 'on wednesday': 2,\n",
       " 'wednesday he': 1,\n",
       " 'he sought': 1,\n",
       " 'sought to': 1,\n",
       " 'to drag': 1,\n",
       " 'drag the': 1,\n",
       " 'queen into': 1,\n",
       " 'into a': 10,\n",
       " 'a political': 7,\n",
       " 'political controversy': 1,\n",
       " 'controversy and': 1,\n",
       " 'and to': 36,\n",
       " 'to use': 11,\n",
       " 'use her': 1,\n",
       " 'her name': 1,\n",
       " 'name to': 2,\n",
       " 'to score': 1,\n",
       " 'score cheap': 1,\n",
       " 'cheap political': 1,\n",
       " 'political points': 1,\n",
       " 'points mr': 1,\n",
       " 'president it': 4,\n",
       " 'it may': 9,\n",
       " 'may be': 16,\n",
       " 'be that': 4,\n",
       " 'that president': 1,\n",
       " 'president fontaine': 1,\n",
       " 'fontaine was': 1,\n",
       " 'was unfamiliar': 1,\n",
       " 'unfamiliar with': 1,\n",
       " 'with the': 148,\n",
       " 'british rules': 1,\n",
       " 'rules of': 11,\n",
       " 'of protocol': 1,\n",
       " 'protocol which': 1,\n",
       " 'which obtain': 1,\n",
       " 'obtain in': 1,\n",
       " 'in such': 8,\n",
       " 'such matters': 1,\n",
       " 'matters but': 1,\n",
       " 'but mr': 1,\n",
       " 'miller has': 1,\n",
       " 'has no': 4,\n",
       " 'no such': 2,\n",
       " 'such excuse': 1,\n",
       " 'excuse he': 1,\n",
       " 'he knows': 3,\n",
       " 'knows perfectly': 3,\n",
       " 'perfectly well': 3,\n",
       " 'well that': 5,\n",
       " 'that ours': 1,\n",
       " 'ours is': 2,\n",
       " 'is a': 120,\n",
       " 'a non': 8,\n",
       " 'non political': 1,\n",
       " 'political head': 1,\n",
       " 'head of': 9,\n",
       " 'of state': 3,\n",
       " 'state he': 1,\n",
       " 'that for': 3,\n",
       " 'for nearly': 1,\n",
       " 'nearly fifty': 1,\n",
       " 'fifty years': 2,\n",
       " 'years she': 1,\n",
       " 'she has': 10,\n",
       " 'has scrupulously': 1,\n",
       " 'scrupulously avoided': 1,\n",
       " 'avoided engaging': 1,\n",
       " 'engaging in': 1,\n",
       " 'in controversial': 1,\n",
       " 'controversial political': 1,\n",
       " 'political issues': 2,\n",
       " 'issues he': 1,\n",
       " 'that she': 7,\n",
       " 'she can': 1,\n",
       " 'can not': 50,\n",
       " 'not come': 3,\n",
       " 'come to': 5,\n",
       " 'to this': 31,\n",
       " 'house to': 5,\n",
       " 'to set': 5,\n",
       " 'set the': 2,\n",
       " 'the record': 3,\n",
       " 'record straight': 1,\n",
       " 'straight his': 1,\n",
       " 'his behaviour': 1,\n",
       " 'behaviour is': 2,\n",
       " 'a disgrace': 1,\n",
       " 'disgrace and': 1,\n",
       " 'and a': 18,\n",
       " 'a scandal': 1,\n",
       " 'scandal mr': 1,\n",
       " 'president i': 68,\n",
       " 'i am': 76,\n",
       " 'am proud': 1,\n",
       " 'proud to': 1,\n",
       " 'be british': 1,\n",
       " 'british but': 1,\n",
       " 'but today': 1,\n",
       " 'today i': 2,\n",
       " 'am ashamed': 1,\n",
       " 'ashamed to': 1,\n",
       " 'to share': 2,\n",
       " 'share my': 1,\n",
       " 'my nationality': 1,\n",
       " 'nationality with': 1,\n",
       " 'with that': 2,\n",
       " 'that man': 1,\n",
       " 'man has': 1,\n",
       " 'has he': 3,\n",
       " 'he no': 3,\n",
       " 'no shame': 1,\n",
       " 'shame has': 1,\n",
       " 'no respect': 1,\n",
       " 'respect has': 1,\n",
       " 'no decency': 1,\n",
       " 'decency or': 1,\n",
       " 'or honour': 1,\n",
       " 'honour at': 1,\n",
       " 'at all': 11,\n",
       " 'all mr': 2,\n",
       " 'i wish': 18,\n",
       " 'wish to': 29,\n",
       " 'use this': 3,\n",
       " 'this point': 10,\n",
       " 'order to': 56,\n",
       " 'to speak': 6,\n",
       " 'speak in': 1,\n",
       " 'in my': 24,\n",
       " 'my capacity': 1,\n",
       " 'capacity as': 3,\n",
       " 'as chairman': 1,\n",
       " 'chairman of': 7,\n",
       " 'the delegation': 3,\n",
       " 'delegation for': 1,\n",
       " 'for relations': 1,\n",
       " 'relations with': 6,\n",
       " 'the countries': 6,\n",
       " 'countries of': 5,\n",
       " 'of central': 2,\n",
       " 'central america': 2,\n",
       " 'america and': 3,\n",
       " 'and mexico': 1,\n",
       " 'mexico as': 1,\n",
       " 'as you': 13,\n",
       " 'you will': 6,\n",
       " 'will be': 56,\n",
       " 'be aware': 3,\n",
       " 'aware elections': 1,\n",
       " 'elections were': 1,\n",
       " 'were held': 1,\n",
       " 'held in': 3,\n",
       " 'in mexico': 1,\n",
       " 'mexico last': 1,\n",
       " 'last sunday': 1,\n",
       " 'sunday yesterday': 1,\n",
       " 'yesterday the': 4,\n",
       " 'delegation which': 1,\n",
       " 'which i': 17,\n",
       " 'i chair': 1,\n",
       " 'chair met': 1,\n",
       " 'met and': 1,\n",
       " 'and discussed': 1,\n",
       " 'discussed a': 1,\n",
       " 'a report': 9,\n",
       " 'report drawn': 2,\n",
       " 'drawn up': 4,\n",
       " 'up by': 4,\n",
       " 'by our': 6,\n",
       " 'our fellow': 7,\n",
       " 'fellow members': 8,\n",
       " 'members who': 6,\n",
       " 'who had': 1,\n",
       " 'had been': 4,\n",
       " 'been to': 1,\n",
       " 'to mexico': 1,\n",
       " 'mexico to': 1,\n",
       " 'to act': 6,\n",
       " 'act as': 4,\n",
       " 'as observers': 1,\n",
       " 'observers at': 1,\n",
       " 'at these': 1,\n",
       " 'these elections': 1,\n",
       " 'elections on': 1,\n",
       " 'on behalf': 85,\n",
       " 'behalf of': 85,\n",
       " 'this parliament': 15,\n",
       " 'parliament mr': 3,\n",
       " 'i would': 147,\n",
       " 'would ask': 20,\n",
       " 'ask you': 16,\n",
       " 'you to': 16,\n",
       " 'to see': 24,\n",
       " 'see if': 1,\n",
       " 'if the': 24,\n",
       " 'the president': 12,\n",
       " 'president of': 10,\n",
       " 'parliament could': 1,\n",
       " 'could forward': 1,\n",
       " 'forward our': 1,\n",
       " 'our resolution': 7,\n",
       " 'resolution the': 3,\n",
       " 'the complete': 1,\n",
       " 'complete text': 1,\n",
       " 'text of': 5,\n",
       " 'of which': 13,\n",
       " 'which you': 3,\n",
       " 'will receive': 1,\n",
       " 'receive today': 1,\n",
       " 'today to': 1,\n",
       " 'to the': 331,\n",
       " 'the mexican': 1,\n",
       " 'mexican authorities': 1,\n",
       " 'authorities mr': 1,\n",
       " 'president ladies': 16,\n",
       " 'ladies and': 33,\n",
       " 'and gentlemen': 33,\n",
       " 'gentlemen i': 9,\n",
       " 'i want': 7,\n",
       " 'want to': 30,\n",
       " 'to draw': 7,\n",
       " 'draw your': 3,\n",
       " 'your attention': 7,\n",
       " 'attention to': 13,\n",
       " 'to something': 2,\n",
       " 'something really': 1,\n",
       " 'really annoying': 1,\n",
       " 'annoying in': 1,\n",
       " 'in connection': 3,\n",
       " 'connection with': 4,\n",
       " 'minutes the': 1,\n",
       " 'the x': 2,\n",
       " 'x texts': 1,\n",
       " 'texts adopted': 1,\n",
       " 'adopted x': 1,\n",
       " 'x also': 2,\n",
       " 'also form': 1,\n",
       " 'form part': 1,\n",
       " 'part of': 33,\n",
       " 'minutes i': 1,\n",
       " 'i pointed': 1,\n",
       " 'pointed out': 11,\n",
       " 'out yesterday': 1,\n",
       " 'yesterday during': 1,\n",
       " 'during the': 23,\n",
       " 'the vote': 35,\n",
       " 'vote on': 21,\n",
       " 'on the': 299,\n",
       " 'the directive': 41,\n",
       " 'directive to': 1,\n",
       " 'to combat': 6,\n",
       " 'combat money': 1,\n",
       " 'money laundering': 1,\n",
       " 'laundering the': 1,\n",
       " 'the english': 6,\n",
       " 'english translation': 3,\n",
       " 'translation of': 1,\n",
       " 'of article': 3,\n",
       " 'article needs': 1,\n",
       " 'needs to': 15,\n",
       " 'be corrected': 2,\n",
       " 'corrected to': 1,\n",
       " 'to bring': 13,\n",
       " 'bring it': 2,\n",
       " 'it in': 8,\n",
       " 'in line': 6,\n",
       " 'line with': 7,\n",
       " 'with amendment': 1,\n",
       " 'amendment no': 11,\n",
       " 'no because': 1,\n",
       " 'because the': 22,\n",
       " 'translation is': 1,\n",
       " 'is wrong': 3,\n",
       " 'wrong in': 2,\n",
       " 'the middle': 3,\n",
       " 'middle of': 2,\n",
       " 'the second': 7,\n",
       " 'second paragraph': 1,\n",
       " 'paragraph of': 4,\n",
       " 'english version': 1,\n",
       " 'version there': 1,\n",
       " 'there is': 89,\n",
       " 'is only': 6,\n",
       " 'only a': 9,\n",
       " 'a reference': 2,\n",
       " 'reference to': 5,\n",
       " 'to representation': 1,\n",
       " 'representation in': 1,\n",
       " 'in legal': 1,\n",
       " 'legal proceedings': 3,\n",
       " 'proceedings but': 1,\n",
       " 'but no': 2,\n",
       " 'no reference': 2,\n",
       " 'to legal': 1,\n",
       " 'legal consultants': 1,\n",
       " 'consultants as': 1,\n",
       " 'as there': 5,\n",
       " 'is in': 21,\n",
       " 'the german': 3,\n",
       " 'german text': 2,\n",
       " 'text that': 2,\n",
       " 'that is': 75,\n",
       " 'a key': 3,\n",
       " 'key point': 1,\n",
       " 'point in': 6,\n",
       " 'in this': 99,\n",
       " 'this directive': 7,\n",
       " 'directive and': 12,\n",
       " 'and since': 3,\n",
       " 'since the': 10,\n",
       " 'the council': 57,\n",
       " 'council working': 1,\n",
       " 'working group': 3,\n",
       " 'group is': 6,\n",
       " 'is working': 2,\n",
       " 'working on': 1,\n",
       " 'the basis': 16,\n",
       " 'basis of': 14,\n",
       " 'english text': 1,\n",
       " 'text it': 1,\n",
       " 'it is': 289,\n",
       " 'is quite': 6,\n",
       " 'quite crucial': 1,\n",
       " 'crucial to': 4,\n",
       " 'to correct': 2,\n",
       " 'correct this': 2,\n",
       " 'this text': 10,\n",
       " 'text i': 1,\n",
       " 'i had': 4,\n",
       " 'had already': 1,\n",
       " 'already complained': 1,\n",
       " 'complained about': 2,\n",
       " 'about this': 8,\n",
       " 'this in': 10,\n",
       " 'in committee': 5,\n",
       " 'committee vis': 1,\n",
       " 'vis a': 2,\n",
       " 'a vis': 2,\n",
       " 'vis the': 2,\n",
       " 'the language': 1,\n",
       " 'language services': 1,\n",
       " 'services at': 2,\n",
       " 'at the': 92,\n",
       " 'the time': 15,\n",
       " 'time the': 8,\n",
       " 'the committee': 86,\n",
       " 'committee decided': 1,\n",
       " 'decided not': 1,\n",
       " 'not to': 16,\n",
       " 'to table': 1,\n",
       " 'table an': 1,\n",
       " 'an amendment': 5,\n",
       " 'amendment because': 1,\n",
       " 'text is': 2,\n",
       " 'is the': 90,\n",
       " 'the original': 4,\n",
       " 'original version': 1,\n",
       " 'version as': 1,\n",
       " 'as drafted': 1,\n",
       " 'drafted by': 2,\n",
       " 'by the': 139,\n",
       " 'the rapporteur': 40,\n",
       " 'rapporteur i': 2,\n",
       " 'i complained': 1,\n",
       " 'this yesterday': 1,\n",
       " 'the situation': 36,\n",
       " 'situation was': 2,\n",
       " 'was not': 12,\n",
       " 'not remedied': 1,\n",
       " 'remedied and': 1,\n",
       " 'and today': 2,\n",
       " 'today is': 7,\n",
       " 'the third': 4,\n",
       " 'third time': 1,\n",
       " 'time i': 4,\n",
       " 'am complaining': 1,\n",
       " 'complaining about': 1,\n",
       " 'about it': 3,\n",
       " 'it i': 2,\n",
       " 'i do': 28,\n",
       " 'do think': 2,\n",
       " 'think the': 2,\n",
       " 'the services': 5,\n",
       " 'services of': 1,\n",
       " 'house should': 1,\n",
       " 'should work': 2,\n",
       " 'work more': 1,\n",
       " 'more carefully': 1,\n",
       " 'carefully on': 1,\n",
       " 'on such': 2,\n",
       " 'such key': 1,\n",
       " 'key questions': 1,\n",
       " 'questions and': 2,\n",
       " 'and it': 30,\n",
       " 'it really': 4,\n",
       " 'really is': 5,\n",
       " 'is intolerable': 1,\n",
       " 'intolerable that': 1,\n",
       " 'that despite': 2,\n",
       " 'despite my': 1,\n",
       " 'my criticism': 1,\n",
       " 'criticism yesterday': 1,\n",
       " 'minutes have': 1,\n",
       " 'have got': 1,\n",
       " 'got the': 2,\n",
       " 'the text': 14,\n",
       " 'text wrong': 1,\n",
       " 'wrong again': 1,\n",
       " 'again i': 1,\n",
       " 'you get': 1,\n",
       " 'get to': 3,\n",
       " 'the bottom': 2,\n",
       " 'bottom of': 2,\n",
       " 'this matter': 18,\n",
       " 'matter to': 1,\n",
       " 'to sort': 2,\n",
       " 'sort out': 2,\n",
       " 'out this': 3,\n",
       " 'this most': 2,\n",
       " 'most important': 12,\n",
       " 'important aspect': 1,\n",
       " 'aspect of': 3,\n",
       " 'to please': 1,\n",
       " 'please correct': 1,\n",
       " 'translation mr': 1,\n",
       " 'president yesterday': 1,\n",
       " 'yesterday we': 1,\n",
       " 'we received': 3,\n",
       " 'received a': 2,\n",
       " 'a communication': 2,\n",
       " 'communication on': 4,\n",
       " 'on yellow': 1,\n",
       " 'yellow paper': 1,\n",
       " 'paper from': 1,\n",
       " 'from the': 65,\n",
       " 'the vice': 5,\n",
       " 'vice president': 15,\n",
       " 'president with': 2,\n",
       " 'with some': 3,\n",
       " 'some nuts': 1,\n",
       " 'nuts attached': 1,\n",
       " 'attached to': 2,\n",
       " 'to it': 2,\n",
       " 'it this': 3,\n",
       " 'this note': 1,\n",
       " 'note was': 1,\n",
       " 'was about': 1,\n",
       " 'about supporting': 1,\n",
       " 'supporting the': 4,\n",
       " 'the export': 1,\n",
       " 'export of': 1,\n",
       " 'of products': 1,\n",
       " 'products from': 1,\n",
       " 'from vanuatu': 1,\n",
       " 'vanuatu a': 1,\n",
       " 'a developing': 1,\n",
       " 'developing country': 1,\n",
       " 'country in': 4,\n",
       " 'the pacific': 2,\n",
       " 'pacific i': 1,\n",
       " 'i think': 22,\n",
       " 'think that': 17,\n",
       " 'a very': 34,\n",
       " 'very good': 9,\n",
       " 'good idea': 4,\n",
       " 'idea and': 2,\n",
       " 'and quite': 1,\n",
       " 'quite right': 2,\n",
       " 'right and': 2,\n",
       " 'and i': 64,\n",
       " 'am quite': 1,\n",
       " 'quite happy': 1,\n",
       " 'happy to': 2,\n",
       " 'to support': 14,\n",
       " 'support it': 4,\n",
       " 'it but': 4,\n",
       " 'but what': 4,\n",
       " 'what rather': 1,\n",
       " 'rather puzzles': 1,\n",
       " 'puzzles me': 1,\n",
       " 'me is': 2,\n",
       " 'is that': 33,\n",
       " 'that in': 34,\n",
       " 'in it': 5,\n",
       " 'it the': 10,\n",
       " 'president quite': 1,\n",
       " 'quite explicitly': 1,\n",
       " 'explicitly stated': 1,\n",
       " 'stated that': 2,\n",
       " 'that it': 48,\n",
       " 'is illegal': 1,\n",
       " 'illegal to': 1,\n",
       " 'to import': 1,\n",
       " 'import these': 1,\n",
       " 'these nuts': 2,\n",
       " 'nuts into': 1,\n",
       " 'into the': 19,\n",
       " 'the european': 238,\n",
       " 'european union': 94,\n",
       " 'union so': 1,\n",
       " 'so i': 14,\n",
       " 'i must': 14,\n",
       " 'must ask': 2,\n",
       " 'ask this': 1,\n",
       " 'this vice': 1,\n",
       " 'president to': 6,\n",
       " 'to explain': 4,\n",
       " 'explain how': 1,\n",
       " 'how these': 2,\n",
       " 'nuts came': 1,\n",
       " 'came to': 2,\n",
       " 'be illegally': 1,\n",
       " 'illegally imported': 1,\n",
       " 'imported since': 1,\n",
       " 'since it': 5,\n",
       " 'it was': 34,\n",
       " 'was he': 1,\n",
       " 'he who': 1,\n",
       " 'who delivered': 1,\n",
       " 'delivered them': 1,\n",
       " 'them to': 17,\n",
       " 'to us': 15,\n",
       " 'us that': 10,\n",
       " 'in itself': 3,\n",
       " 'itself is': 2,\n",
       " 'a contradiction': 1,\n",
       " 'contradiction so': 1,\n",
       " 'so would': 1,\n",
       " 'would he': 2,\n",
       " 'he please': 1,\n",
       " 'please clarify': 1,\n",
       " 'clarify the': 1,\n",
       " 'the matter': 17,\n",
       " 'matter laughter': 1,\n",
       " 'laughter and': 1,\n",
       " 'and applause': 1,\n",
       " 'applause the': 1,\n",
       " 'minutes were': 1,\n",
       " 'were approved': 1,\n",
       " 'approved welcome': 1,\n",
       " 'welcome committee': 1,\n",
       " 'committee on': 73,\n",
       " 'on petitions': 35,\n",
       " 'petitions european': 1,\n",
       " 'european ombudsman': 16,\n",
       " 'ombudsman the': 4,\n",
       " 'the next': 25,\n",
       " 'next item': 12,\n",
       " 'item is': 11,\n",
       " 'the joint': 16,\n",
       " 'joint debate': 8,\n",
       " 'debate on': 15,\n",
       " 'the following': 10,\n",
       " 'following reports': 1,\n",
       " 'reports tabled': 1,\n",
       " 'tabled on': 1,\n",
       " 'petitions a': 2,\n",
       " 'a by': 18,\n",
       " 'by mrs': 30,\n",
       " 'mrs lambert': 12,\n",
       " 'lambert on': 4,\n",
       " 'the deliberations': 2,\n",
       " 'deliberations of': 2,\n",
       " 'petitions during': 2,\n",
       " 'the parliamentary': 5,\n",
       " 'parliamentary year': 2,\n",
       " 'year a': 2,\n",
       " 'mrs thors': 10,\n",
       " 'thors on': 2,\n",
       " 'the annual': 8,\n",
       " 'annual report': 10,\n",
       " 'report on': 28,\n",
       " 'the activities': 5,\n",
       " 'activities of': 5,\n",
       " 'ombudsman in': 6,\n",
       " 'in mr': 2,\n",
       " 'am glad': 2,\n",
       " 'glad to': 1,\n",
       " 'to have': 27,\n",
       " 'have the': 23,\n",
       " 'the opportunity': 8,\n",
       " 'opportunity to': 17,\n",
       " 'to address': 11,\n",
       " 'address the': 6,\n",
       " 'european parliament': 58,\n",
       " 'parliament concerning': 1,\n",
       " 'concerning the': 4,\n",
       " 'the results': 9,\n",
       " 'results of': 6,\n",
       " 'of my': 11,\n",
       " 'my work': 1,\n",
       " 'work as': 6,\n",
       " 'as european': 1,\n",
       " 'the year': 13,\n",
       " 'year the': 10,\n",
       " 'report for': 4,\n",
       " 'for covers': 1,\n",
       " 'covers the': 1,\n",
       " 'the last': 20,\n",
       " 'last year': 16,\n",
       " 'year of': 3,\n",
       " 'the first': 18,\n",
       " 'first mandate': 1,\n",
       " 'mandate of': 2,\n",
       " 'ombudsman last': 2,\n",
       " 'the ombudsman': 83,\n",
       " 'ombudsman s': 25,\n",
       " 's office': 6,\n",
       " 'office received': 1,\n",
       " 'received new': 1,\n",
       " 'new complaints': 1,\n",
       " 'complaints compared': 1,\n",
       " 'compared to': 4,\n",
       " 'to in': 4,\n",
       " 'in an': 17,\n",
       " 'an increase': 5,\n",
       " 'increase of': 3,\n",
       " 'of nearly': 1,\n",
       " 'nearly the': 1,\n",
       " 'the number': 19,\n",
       " 'number of': 65,\n",
       " 'of inquiries': 1,\n",
       " 'inquiries opened': 1,\n",
       " 'opened rose': 1,\n",
       " 'rose from': 1,\n",
       " 'from in': 3,\n",
       " 'in to': 5,\n",
       " 'of over': 1,\n",
       " 'over as': 1,\n",
       " 'as well': 27,\n",
       " 'well as': 21,\n",
       " 'as becoming': 1,\n",
       " 'becoming better': 1,\n",
       " 'better known': 2,\n",
       " 'known the': 1,\n",
       " 'ombudsman is': 7,\n",
       " 'is also': 29,\n",
       " 'also attracting': 1,\n",
       " 'attracting a': 1,\n",
       " 'a higher': 2,\n",
       " 'higher proportion': 1,\n",
       " 'proportion of': 5,\n",
       " 'of serious': 2,\n",
       " 'serious complaints': 1,\n",
       " 'complaints which': 1,\n",
       " 'which require': 1,\n",
       " 'require full': 1,\n",
       " 'full investigation': 1,\n",
       " 'investigation you': 1,\n",
       " 'you may': 2,\n",
       " 'be interested': 2,\n",
       " 'interested to': 2,\n",
       " 'to know': 6,\n",
       " 'know that': 17,\n",
       " 'that the': 208,\n",
       " 'the same': 46,\n",
       " 'same trend': 2,\n",
       " 'trend has': 1,\n",
       " 'has continued': 2,\n",
       " 'continued to': 4,\n",
       " 'the present': 16,\n",
       " 'present year': 1,\n",
       " 'year during': 1,\n",
       " 'first six': 1,\n",
       " 'six months': 3,\n",
       " 'months of': 2,\n",
       " 'this year': 11,\n",
       " 'of new': 5,\n",
       " 'new inquiries': 1,\n",
       " 'inquiries rose': 1,\n",
       " 'rose by': 1,\n",
       " 'by over': 1,\n",
       " 'over the': 23,\n",
       " 'same period': 2,\n",
       " 'period in': 4,\n",
       " 'in inquiries': 1,\n",
       " 'inquiries are': 1,\n",
       " 'are the': 15,\n",
       " 'the most': 28,\n",
       " 'most resource': 1,\n",
       " 'resource intensive': 1,\n",
       " 'intensive aspect': 1,\n",
       " 's work': 5,\n",
       " 'work it': 2,\n",
       " 'is therefore': 16,\n",
       " 'therefore a': 4,\n",
       " 'a great': 16,\n",
       " 'great challenge': 1,\n",
       " 'challenge to': 1,\n",
       " 'to deal': 11,\n",
       " 'deal with': 14,\n",
       " 'with this': 15,\n",
       " 'this increase': 1,\n",
       " 'increase whilst': 1,\n",
       " 'whilst maintaining': 1,\n",
       " 'maintaining the': 3,\n",
       " 'the quality': 7,\n",
       " 'quality of': 10,\n",
       " 'of our': 40,\n",
       " 'our work': 6,\n",
       " 'work and': 3,\n",
       " 'and meeting': 1,\n",
       " 'meeting our': 1,\n",
       " 'our target': 1,\n",
       " 'target of': 1,\n",
       " 'of closing': 1,\n",
       " 'closing most': 1,\n",
       " 'most inquiries': 1,\n",
       " 'inquiries within': 1,\n",
       " 'within one': 1,\n",
       " 'one year': 2,\n",
       " 'year i': 4,\n",
       " 'do hope': 1,\n",
       " 'hope that': 20,\n",
       " 'parliament will': 7,\n",
       " 'will support': 4,\n",
       " 'support the': 20,\n",
       " 'in seeking': 1,\n",
       " 'seeking the': 1,\n",
       " 'the resources': 5,\n",
       " 'resources needed': 1,\n",
       " 'needed to': 4,\n",
       " 'to carry': 2,\n",
       " 'carry out': 5,\n",
       " 'out the': 9,\n",
       " 'the work': 31,\n",
       " 'work effectively': 3,\n",
       " 'effectively during': 1,\n",
       " 'during we': 1,\n",
       " 'we continued': 1,\n",
       " 'to receive': 6,\n",
       " 'receive a': 3,\n",
       " 'a quite': 1,\n",
       " 'quite high': 1,\n",
       " 'high proportion': 1,\n",
       " 'of complaints': 8,\n",
       " 'complaints over': 1,\n",
       " 'over which': 1,\n",
       " 'which lay': 1,\n",
       " 'lay outside': 1,\n",
       " 'outside the': 7,\n",
       " 'the mandate': 4,\n",
       " 'the proportion': 1,\n",
       " 'the complaints': 7,\n",
       " 'complaints outside': 1,\n",
       " 'mandate even': 1,\n",
       " 'even increased': 1,\n",
       " 'increased slightly': 1,\n",
       " 'slightly probably': 1,\n",
       " 'probably because': 1,\n",
       " 'because we': 15,\n",
       " 'a growing': 2,\n",
       " 'growing number': 2,\n",
       " 'complaints by': 2,\n",
       " 'by e': 5,\n",
       " 'e mail': 6,\n",
       " 'mail we': 1,\n",
       " 'we gladly': 1,\n",
       " 'gladly accept': 1,\n",
       " 'accept complaints': 1,\n",
       " 'mail and': 1,\n",
       " 'and many': 9,\n",
       " 'many of': 8,\n",
       " 'of them': 11,\n",
       " 'them contain': 1,\n",
       " 'contain well': 1,\n",
       " 'well presented': 1,\n",
       " 'presented allegations': 1,\n",
       " 'allegations of': 1,\n",
       " 'of maladministration': 3,\n",
       " 'maladministration however': 1,\n",
       " 'however in': 2,\n",
       " 'in all': 10,\n",
       " 'all more': 1,\n",
       " 'more of': 2,\n",
       " 'the e': 1,\n",
       " 'mail complaints': 1,\n",
       " 'complaints are': 2,\n",
       " 'are directed': 1,\n",
       " 'directed against': 2,\n",
       " 'against authorities': 1,\n",
       " 'authorities of': 2,\n",
       " 'of member': 3,\n",
       " 'member states': 87,\n",
       " 'states during': 3,\n",
       " 'the campaign': 3,\n",
       " 'campaign for': 1,\n",
       " 'for the': 188,\n",
       " 'the election': 1,\n",
       " 'election of': 1,\n",
       " 'of a': 73,\n",
       " 'a european': 17,\n",
       " 'year some': 2,\n",
       " 'some critical': 1,\n",
       " 'critical voices': 1,\n",
       " 'voices suggested': 1,\n",
       " 'suggested that': 1,\n",
       " 's mandate': 1,\n",
       " 'mandate should': 1,\n",
       " 'should be': 58,\n",
       " 'be broadened': 1,\n",
       " 'broadened to': 1,\n",
       " 'to include': 4,\n",
       " 'include complaints': 1,\n",
       " 'complaints against': 3,\n",
       " 'against member': 1,\n",
       " 'states in': 8,\n",
       " 'in cases': 3,\n",
       " 'cases where': 2,\n",
       " 'where a': 1,\n",
       " 'a question': 4,\n",
       " 'question of': 13,\n",
       " 'of european': 25,\n",
       " 'union law': 3,\n",
       " 'law is': 2,\n",
       " 'is involved': 4,\n",
       " 'involved i': 2,\n",
       " 'am not': 8,\n",
       " 'not sure': 2,\n",
       " 'sure whether': 2,\n",
       " 'whether they': 3,\n",
       " 'they understood': 1,\n",
       " 'understood that': 1,\n",
       " 'that this': 51,\n",
       " 'this would': 8,\n",
       " 'would mean': 4,\n",
       " 'mean the': 2,\n",
       " 'ombudsman carrying': 1,\n",
       " 'carrying out': 6,\n",
       " 'out tasks': 1,\n",
       " 'tasks currently': 1,\n",
       " 'currently belonging': 1,\n",
       " 'belonging to': 1,\n",
       " 'european commission': 15,\n",
       " 'commission as': 5,\n",
       " 'as the': 73,\n",
       " 'the guardian': 1,\n",
       " 'guardian of': 1,\n",
       " 'the treaty': 13,\n",
       " 'treaty and': 3,\n",
       " 'parliament as': 2,\n",
       " 'the democratic': 2,\n",
       " 'democratic forum': 1,\n",
       " 'forum to': 1,\n",
       " 'to which': 11,\n",
       " 'which citizens': 1,\n",
       " 'citizens may': 2,\n",
       " 'may address': 1,\n",
       " 'address petitions': 1,\n",
       " 'petitions we': 2,\n",
       " 'we study': 1,\n",
       " 'study carefully': 1,\n",
       " 'carefully all': 1,\n",
       " 'all the': 51,\n",
       " 'complaints we': 2,\n",
       " 'we receive': 4,\n",
       " 'receive especially': 1,\n",
       " 'especially if': 2,\n",
       " 'if they': 15,\n",
       " 'they involve': 1,\n",
       " 'involve rights': 1,\n",
       " 'rights under': 2,\n",
       " 'under european': 2,\n",
       " 'law in': 4,\n",
       " 'in we': 2,\n",
       " 'we managed': 1,\n",
       " 'managed to': 6,\n",
       " 'to advise': 1,\n",
       " 'advise the': 1,\n",
       " 'the complainant': 3,\n",
       " 'complainant or': 1,\n",
       " 'or transfer': 1,\n",
       " 'transfer the': 1,\n",
       " 'the complaint': 2,\n",
       " 'complaint to': 2,\n",
       " 'to a': 44,\n",
       " 'a competent': 1,\n",
       " 'competent body': 1,\n",
       " 'body in': 1,\n",
       " 'in about': 1,\n",
       " 'about half': 1,\n",
       " 'half of': 1,\n",
       " 'the cases': 2,\n",
       " 'cases of': 5,\n",
       " 'of these': 23,\n",
       " 'these we': 1,\n",
       " 'we transferred': 1,\n",
       " 'transferred cases': 1,\n",
       " 'cases to': 1,\n",
       " 'parliament to': 15,\n",
       " 'be dealt': 5,\n",
       " 'dealt with': 7,\n",
       " 'with as': 2,\n",
       " 'as petitions': 2,\n",
       " 'petitions and': 9,\n",
       " 'and advised': 1,\n",
       " 'advised a': 1,\n",
       " 'a further': 4,\n",
       " 'further complainants': 1,\n",
       " 'complainants of': 1,\n",
       " 'of their': 26,\n",
       " 'their rights': 5,\n",
       " 'rights to': 3,\n",
       " 'to petition': 2,\n",
       " 'petition the': 2,\n",
       " 'parliament in': 6,\n",
       " 'in many': 6,\n",
       " 'many cases': 5,\n",
       " 'cases complaints': 1,\n",
       " 'against national': 1,\n",
       " 'national regional': 1,\n",
       " 'regional or': 1,\n",
       " 'or local': 1,\n",
       " 'local authorities': 3,\n",
       " 'authorities could': 1,\n",
       " 'could be': 13,\n",
       " 'with effectively': 1,\n",
       " 'effectively by': 1,\n",
       " 'by an': 7,\n",
       " 'an ombudsman': 5,\n",
       " 'the member': 44,\n",
       " 'states concerned': 1,\n",
       " ...}"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "with open('ngram_langid_files/English.bigram.pickle', 'rb') as handle:\n",
    "    bigram_english_dict = pickle.load(handle)\n",
    "bigram_english_dict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "906"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "bigram_english_dict.get('of the')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('the', 5699),\n",
       " ('of', 2769),\n",
       " ('to', 2490),\n",
       " ('and', 2040),\n",
       " ('in', 1668),\n",
       " ('a', 1343),\n",
       " ('is', 1303),\n",
       " ('that', 1205),\n",
       " ('this', 887),\n",
       " ('on', 865)]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import operator\n",
    "english_unigram_freqs = sorted(unigram_english_dict.items(), key=operator.itemgetter(1), reverse=True)\n",
    "english_unigram_freqs[:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAnIAAAHgCAYAAAAhaIBAAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4wLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvFvnyVgAAIABJREFUeJzs3XucVXW9//HXBwZMT+INMG6GJiggFxFFPaYZAoqpXEzxcjKxPFZ2LLrIOZalpaLl8S4eUovMxLIUU1MIxZ9hykURFS+YYIAGomgp3ka/vz/WmmkYZmA2DLNnwev5eOzH3vu7Lvvz3bOHefNd67t2pJSQJElS8bQodwGSJEnaMAY5SZKkgjLISZIkFZRBTpIkqaAMcpIkSQVlkJMkSSoog5yksouIz0VEiohv12qfExFvbcR+V0bEUxtfYfFExE/z93RAuWupEhF75TVdXe5apM2FQU6Flf9BKOX2xWZQ89YR8c2ImBQR8yPig7y20evZLiLiyxExNyLejog3IuJPETG4xNf/XAPepw0OTpKkplVR7gKkjXBeHW3fALYDrgDeqLVs3iavaP12Av43f/wKsBzo1IDtJgD/CSwGrgP+DRgNTI2IU1NKvyixjoXAr+tZ9n6J+9qURgFblbsINZrngR7AqnIXIm0uDHIqrJTSD2u35aNu2wGXp5QWN3FJDfE6cDjweEppRUT8FPjWujbIR93+E3gaOCCl9M+8/TJgDnB1RNybUvp7CXU8X9f719yklF4qdw1qPCml94Fny12HtDnx0Kq2SBHRMyJ+HRGvRMT7EbE0Im6MiK51rFt9rlFEnJ4fEn0nIv4eEf8XEW0b+roppdUppftSSitKKPcr+f15VSEu39dzwM/IRuf+o4T9laTqPLWIaB0RP4yIFyPivYh4KSJ+FBF1/ocwIk6LiCci4t38vboxItqVct5bXetGRIv8MPOj+Tlw70TE3yLinogYXs9+2kTEFfnP+b2IeD4izirhPfhD/hnoUKv993n7lFrt7SPio4i4p1b7NhFxbkQ8ndf9ZkQ8EBHH1PGa1eeTRUSv/LVezfc7oMZ6B+SH2d/KD7nfGxH9G9q3Gvup9+cSEWfmtRxbq31lRDzV0Pc31nGOXN7HO/M+vBURD0XEYQ147R0j4qr8M1AZ+XmWEbFLRJwfEY9ExPIav+e/jIjd11VbRPSIiCkRsSr/Gd0dEd3z9TpExM/zz/S7EfGXiDiwjv3tkL/+goj4Z0T8IyIWRsTNEdF7fT8PqaEckdMWJyI+DfwR2Bq4newwYy/gVOCYiPhMSunJOjb9PjAIuBW4GzgUOB04JCL2TynVPpTbWA4FEnBfHcv+CHwT+Czwk030+gAB/B7oB9wLvA0cBXwP2B74+horR5xP9n69CtwIvEU2Evn/GqGWy/PXWwjcku+7IzAQGA7cUWv9rYEHgG2BP+RtI4HLI6IipXRpA15zOvA5sp//ryALlMAh+fJDIqJlSunD/Plnyd6z6VU7iIiqOvYDngSuAtoAnwfuiIhzUkoX1vHavYBHyU4NuCnvx9v5Pg8j+yy2AH5Lduh9X+DP+a0pbPT7GxF9gYeAjwNTgGeA7mSf73vWsek2ZJ+p1vl6q4El+bIhwFjgfrKR69XAHsCJwNERMTD/z1BtewCPAI8BNwDdgKOBvSPi4Lyvfyf77LUHjic7xeFTKaXleX9akv3s98779UfgI2AXYDAwjewzIG28lJI3b5vNjewPWQK61rO8osY6x9RadlrePrdW+0/z9reBnrWWTcyXXbGB9Vbte3Q9y3fOl79Sz/Ku+fK/NvD1Ppev/zzww3puI2ttMyffZiawXY32NsBS4D1g+xrtewEfAsuAT9Rob0n2RzoBb9VT17freO23ajxvAbwDvABsVUf/2tZ6vjLf72011we65D/PvwPRgPetT76fn9do2ydvm5rfD6yx7Gd529412i6oUUvLGu2dyM6X/BDoV+t9TPntnHo+yy/lyw+rteycGtsOaOBnY433utayM/N9Hbsx72+NPl1daz+P5u3/Uav92Br9qO+17wQ+VkfNnwC2qaN9IPAu8Nta7TXf77NqLbs4b3+d7BzXmn36Sr7sRzXaDsjbbqrn57Z97XZv3jb05qFVbWkGAZ8EpqWU1jgcllK6AXgc6F/PoanrU0oLarWdQxYsTslHaBrbdvn9m/Usr2rfvsT9dgN+UM9tZD3bfCulVF1HSukfZKOTrclG6qqcTBa4Lk01zttL2WjV2SXWWZdENiHjw7UWpLSynvW/nlJ6r8Z6S8hGFncm+zysz5Nko4uDarRVPf5eredVj19jzQk2Y4BKsvexuvaU0jJgPNl7NqaO116cL69tENkIzz0ppT/VWvYTsiDdFDbq/Y2IPclGKeenlG5aY8cp3UYW8tblmymld9cqKqW/p5RW19H+KPAw2YhdXRYAV9Zqm1S1OfA/KaVUY9kv8/t+rO2dOl6/Mm260XttgQxy2tJUBbT761n+QH6/dx3LHqzdkFJ6lewf/u2A3Ta6utJFVSklbnd3SinquZ1cx/ofkYXc2qoOY+1Qo63qvVvr0F5K6VmykZQNklL6CJhMNvPxqcjO0RsSEduuY7OXU0qv1NFeV+31vW4i+2x0iYhuefNngYUppVlkn4FBAJGdZ7krMKPqD35+bt0nyEZO65rAUfV5rOtz91jN4FdD1We5rs/l+2SHB5vCxr6/VX2eWc/yh9ax7esppb/WtzAiRkbEH/Pz2aou9ZPITldoU8/n5rFaQQ3g5fx+Qe3QmFJ6m+w/VJ1r7oNsUseXI2JGRIyNiIER0WodfZE2iEFOW5qqEa66/vDUbK9rhGt5PdtUjTptV8/yjVE1AlbfvtvUWm9TeafmiEsNlfl9yxptVbXW937V195Q/0k2svch2WjYfcBrEXFbRNQ1+lPf6Eddta9L1flugyKiNXBQjbbpwL9HxMf418jc9Brbbsznrr7ZyOt7n0uZxbwxNvb93ZjPS719jIhzgN8BA8hC+GXA+WSXLXomX62uS9vU9btUuY5lVcurQ1r+u3IwcA3ZuX6XkgXrFZFNntq6vrqlUjnZQVuaqn+IP1HP8g611qtp53q2qdpXo4eplNLyiHgD2Dki2uSHM2uqGh16vrFfeyNU1bgz2TlctdX3PjZISukD4BLgkoj4BPBpssO5o4A9I6JvPSNYG6tq1OwwshG4f+NfYe1+sgkYB/GvIFdz1HdjPnf1jbZWrbu+z2VDfUT9fxNKPXRfipqfl7qs6/NS53uTB6VzyD5/A2ofco8SL6S9IfLR+jOBMyNiD+AzZOfTfYvss/OV+reWGs4ROW1pqg4Pfqae5VXtj9Wx7JDaDRHRDuhJ9kf1xY2srT4PkB1CHVrHsiPy+/oOFZdD1Xt8UO0F+flQDb5cy/rk50H9NqV0DDCLbIbnWpeWaKTXegH4G9lhucPIQkTVofgZZCOEh+XLl6UaMyLzQ49/Bz4VEV3q2P2h+X1dn7v6VK1b1+eyNbB/CfuC7CK9W+Wf6do25dd8VX1e/r2e5Wt9jhqgE9ls2gfrCHE7AE16+Y+U0nMppf8jG6X7gGx2tdQoDHLa0vyJ7I/x4RFxRM0FkV1MuD8wL6VU1x/UL0VEz1ptF5D9wfhlfv7WpjAhv/9BzXN68v/lf4lsduBNdW1YJr8iG935VkRUj6bkk0Eu3pgdR8THI6Ku4LIV/zpEt9YJ5o3ofmBHssO781JKrwHkJ68/TnYJm09Qd7D+Odnht4trToyJiI7AOPJZsSXUMp3sPLRh+WVIavoODfvGkJpm5fdfrtkYEUeRXX5jk0gpPUM2Y7ZPRKxxPcT82nGlBlLI3pdKYGB+uLtqf1sB15JdKmWTiYju+e9nbW3JRj3XmoQhbSgPrWqLklKqjIgvkF3X6Q8R8XuyS1n0Irsu2irgi/Vs/idgVkTcCqwgG0UZSHZY89yG1hARPyA7GR7+NdLxlYg4vOp1Ukq/qlHztIiYSHbNuvl5zVVf0bUtMCaV9q0OAN0j4ofrWH5JXTP+GiKl9GREjAf+B3gyIn5Ldq23I8hmuD5L6SGjyvbAjIj4K1nw+BvZtcQOJzvM/OuU0t82cN8NMZ3s89Gef81WrLns7BqPa/sx2TXETgB6RsR9ZD+/48i+uu0HKaW6JpTUKf8sjyG7jtw9EXEb2QzXAWSjWNPy12uo/wP+C7ggIvYju05fT7JRxjuAESXsq1T/STZpY1JEjCQ7dN0dOIbs2nRHkf3noEFSSu9FxHVkhzafjIi7yP7DNSi/n0n9I4CNYT/gpoiYRfaNLH8nO0Q8nGx0faP+QyPVZJDTFiel9GD+h+p7ZGFsONmlJX4JnJ9Squ8Q6Y/ILjp6JlloeJPsOnLfK/FyAkeRXYOspoPzG2Sh51e1lp8BzM3vv0I22jAbGJ9SmlbCa1epuvxIfa5mI0YNUkrnRMRismBwGtl7dTdZ0JnNv86LKtVrZAHxULJz49rl+1pI9sdxUv2bNor763kMawa5tUbkUkqrI+IzZKNlxwNnkV1G5XGy6xD+vtRiUkp/iohDyT6bx5B9Lv5CFuROpIQgl1Jamtd3MVl4G0R26Y9DyMLhJgtyKaXH8m9HuJBsNvBgsvdlGFlfjqL0z8y3yCaRfIHsd+Z1skuifI+1Ly/S2GaSncd5CHAk2X9AluftV6SU6gr60gaJtWdZS6op/vV9qPumlOaUu54ii+zrzJYD96eUNvkJ5yq+yL7+7GigS0ppabnrkZobz5GT1Ogi+67RlrXaWgNXkP27c3tZClOzFBGt6ppkkZ+fdxQw2xAn1c1Dq5I2hS8AYyPifrKv8WpHNiN4N7JDfz8rX2lqhrYDlkXEdOA5sokffcgOs66m1nf5SvoXg5ykTeHPZOewHUp2In8im1TyQ+An+bXgpCpvAdeTBbd/J5uQsAL4NXBhSunpMtYmNWueIydJklRQniMnSZJUUAY5SZKkgjLISZIkFZRBTpIkqaAMcpIkSQVlkJMkSSoog5wkSVJBGeQkSZIKyiAnSZJUUAY5SZKkgjLISZIkFZRBTpIkqaAMcpIkSQVlkJMkSSoog5wkSVJBGeQkSZIKyiAnSZJUUAY5SZKkgjLISZIkFZRBTpIkqaAMcpIkSQVlkJMkSSoog5wkSVJBGeQkSZIKyiAnSZJUUAY5SZKkgjLISZIkFZRBTpIkqaAMcpIkSQVlkJMkSSqoinIX0BTatm2bunbtWu4yJEmS1mvu3LkrU0rtGrLuFhHkunbtypw5c8pdhiRJ0npFxEsNXddDq5IkSQVlkJMkSSoog5wkSVJBGeQkSZIKyiDXxLp27Urv3r3p168fAwYMqG6/6qqr2GOPPejVqxff/e53AXj//fc59dRT6d27N3379mXGjBnV699666306dNnjfUlSdKWZYuYtdrcPPDAA7Rt23aN51OmTGH+/PlstdVWrFixAoCf/exnADz55JOsWLGCI444gtmzZ7Nq1Sq+853vMHfuXNq1a8cpp5zC9OnTGTRoUFn6I0mSysMRuWZgwoQJjBs3jq222gqA9u3bA7BgwYLqcNa+fXu233575syZw4svvkj37t1p1y67xMxhhx3G7373u/IUL0mSysYg18QigiFDhrDPPvswceJEAJ5//nkeeughBg4cyCGHHMLs2bMB6Nu3L1OmTKGyspJFixYxd+5clixZwu67786zzz7L4sWLqays5I477mDJkiXl7JYkSSoDD602sZkzZ9KxY0dWrFjB4MGD2XPPPamsrGTVqlU88sgjzJ49m+OOO44XX3yRMWPG8MwzzzBgwAA++clPcuCBB1JRUcEOO+zAhAkTOP7442nRogUHHnggL774Yrm7JkmSmphBrol17NgRyA6VjhgxglmzZtG5c2dGjhxJRLDffvvRokULVq5cSbt27bjsssuqtz3wwAPp1q0bAEcddRRHHXUUABMnTqRly5ZN3xlJklRWHlptQm+//Tb//Oc/qx9PnTqVvfbai+HDh3P//fcD2WHW999/n7Zt27J69WrefvttAKZNm0ZFRQU9e/YEqJ4QsWrVKq699lq+9KUvlaFHkiSpnByRa0LLly9nxIgRAFRWVnLiiSdy+OGH8/777zNmzBj22msvWrduzaRJk4gIVqxYwdChQ2nRogWdOnXipptuqt7XWWedxRNPPAHAueeeS/fu3cvSJ0mSVD6RUip3DZvcgAED0pw5c8pdhiRJ0npFxNyU0oD1r+mhVUmSpMIyyEmSJBWUQU6SJKmgnOzQiLqOu7vcJazX4vFHlrsESZLUSByRkyRJKiiDnCRJUkEZ5CRJkgrKICdJklRQBjlJkqSCMshJkiQVlEFOkiSpoAxykiRJBWWQkyRJKiiDnCRJUkEZ5CRJkgrKICdJklRQBjlJkqSCMshJkiQVlEFOkiSpoAxykiRJBWWQkyRJKiiDnCRJUkEZ5CRJkgrKICdJklRQBjlJkqSCMshJkiQVlEFOkiSpoAxykiRJBWWQkyRJKiiDnCRJUkEZ5CRJkgrKICdJklRQBjlJkqSCMshJkiQVlEFOkiSpoAxykiRJBWWQkyRJKiiDnCRJUkEZ5CRJkgrKICdJklRQBjlJkqSCMshJkiQVlEFOkiSpoAxykiRJBdWkQS4iFkfEkxExLyLm5G07RsS0iFiY3++Qt0dEXBkRL0TE/IjoX2M/p+TrL4yIU5qyD5IkSc1FOUbkDk0p9UspDcifjwOmp5S6AdPz5wBHAN3y2+nABMiCH/ADYCCwH/CDqvAnSZK0JWkOh1aPASbljycBw2u0/zJlHgG2j4gOwFBgWkrp9ZTSKmAacHhTFy1JklRuTR3kEjA1IuZGxOl5284ppVcA8vv2eXsnYEmNbZfmbfW1S5IkbVEqmvj1/j2l9HJEtAemRcSz61g36mhL62hfc+MsKJ4OsMsuu2xIrZIkSc1ak47IpZRezu9XALeTneO2PD9kSn6/Il99KdClxuadgZfX0V77tSamlAaklAa0a9eusbsiSZJUdk0W5CLi3yJi26rHwBDgKeBOoGrm6SnAlPzxncAX8tmr+wNv5ode7wOGRMQO+SSHIXmbJEnSFqUpD63uDNweEVWv++uU0r0RMRv4TUScBvwN+Hy+/j3AMOAFYDVwKkBK6fWI+BEwO1/v/JTS603XDUmSpOahyYJcSulFoG8d7a8Bg+poT8DX6tnXjcCNjV2jJElSkTSHy49IkiRpAxjkJEmSCsogJ0mSVFAGOUmSpIIyyEmSJBWUQU6SJKmgDHKSJEkFZZCTJEkqKIOcJElSQRnkJEmSCsogJ0mSVFAGOUmSpIIyyEmSJBWUQU6SJKmgDHKSJEkFZZCTJEkqKIOcJElSQRnkJEmSCsogJ0mSVFAGOUmSpIIyyEmSJBWUQU6SJKmgDHKSJEkFZZCTJEkqKIOcJElSQRnkJEmSCsogJ0mSVFAGOUmSpIIyyEmSJBWUQU6SJKmgDHKSJEkFZZCTJEkqKIOcJElSQRnkJEmSCsogJ0mSVFAGOUmSpIIyyEmSJBWUQU6SJKmgDHKSJEkFZZCTJEkqKIOcJElSQRnkJEmSCsogJ0mSVFAGOUmSpIIyyEmSJBWUQU6SJKmgDHKSJEkFZZCTJEkqKIOcJElSQRnkJEmSCsogJ0mSVFAGOUmSpIIyyEmSJBWUQU6SJKmgDHKSJEkF1eRBLiJaRsTjEXFX/nzXiHg0IhZGxK0R0Tpv3yp//kK+vGuNffx33v5cRAxt6j5IkiQ1B+UYkTsLeKbG84uBy1JK3YBVwGl5+2nAqpTS7sBl+XpERE9gNNALOBy4NiJaNlHtkiRJzUaTBrmI6AwcCVyfPw/gs8Bt+SqTgOH542Py5+TLB+XrHwNMTim9l1JaBLwA7Nc0PZAkSWo+mnpE7nLgu8BH+fOdgDdSSpX586VAp/xxJ2AJQL78zXz96vY6tqkWEadHxJyImPPqq682dj8kSZLKrsmCXER8DliRUppbs7mOVdN6lq1rm381pDQxpTQgpTSgXbt2JdcrSZLU3FU04Wv9O3B0RAwDPga0IRuh2z4iKvJRt87Ay/n6S4EuwNKIqAC2A16v0V6l5jaSJElbjCYbkUsp/XdKqXNKqSvZZIX7U0onAQ8Ax+arnQJMyR/fmT8nX35/Sinl7aPzWa27At2AWU3UDUmSpGajKUfk6nM2MDkifgw8DtyQt98A3BQRL5CNxI0GSCk9HRG/ARYAlcDXUkofNn3ZkiRJ5VWWIJdSmgHMyB+/SB2zTlNK7wKfr2f7C4ALNl2FkiRJzZ/f7CBJklRQBjlJkqSCMshJkiQVlEFOkiSpoAxykiRJBWWQkyRJKiiDnCRJUkEZ5CRJkgrKICdJklRQBjlJkqSCMshJkiQVlEFOkiSpoAxykiRJBWWQkyRJKiiDnCRJUkEZ5CRJkgrKICdJklRQBjlJkqSCMshJkiQVlEFOkiSpoAxykiRJBWWQkyRJKiiDnCRJUkEZ5CRJkgrKICdJklRQJQW5iNhxUxUiSZKk0pQ6IvdyREyOiMGbpBpJkiQ1WKlBbkS+zR8i4qWI+EFEfHIT1CVJkqT1KCnIpZT+mFI6DugI/C8wHPhrREyNiOMjovWmKFKSJElr26DJDiml11NKV6SU9ga+ARwM3EJ26PXHEbFNYxYpSZKktVVsyEYR0R74AnAq0BW4DbiBbKRuHLAfMKRxSpQkSVJdSgpyEXE0MAY4AngO+D/gppTSqhrrzAMeb8wiJUmStLZSR+RuBm4FDk4pPVrPOi8CF29UVZIkSVqvUoNch5TSW+taIaX0DvD9DS9JkiRJDVHqZIchEXFU7caIOCoiRjRSTZIkSWqAUoPc+cAHdbS/C/xo48uRJElSQ5Ua5D4FPFtH+0Jgt40vR5IkSQ1VapB7gyzM1dYN+OfGlyNJkqSGKjXI3QlcFhHVYS4idgcuzZdJkiSpiZQa5L4LrAaejYhFEbEIeAZ4B/hOYxcnSZKk+pV0+ZGU0psRcQBwONAPCOAx4L6UUtoE9UmSJKkeJX9FVx7Y/pjfJEmSVCYlB7mIGAB8FmhPrUOzKaWxjVSXJEmS1qPU71r9JtnEhsXAy0DNw6keWpUkSWpCpY7IfRMYm1K6fFMUI0mSpIYrddbqdniZEUmSpGah1CD3G2DIpihEkiRJpSn10OpfgR9FxP7Ak9T63tWU0pWNVZgkSZLWrdQg9zXgXWBQfqspAQY5SZKkJlLqBYG7bKpCJEmSVJpSz5GrFhE7RUQ0ZjGSJElquJKCXES0iogLI+INYDmwa95+UUScsSkKlCRJUt1KHZH7PjAKOA14r0b7XODUxipKkiRJ61dqkDsR+M+U0u+Aj2q0Pwns0WhVSZIkab1KDXKdyL6eq7aWrGfiRER8LCJmRcQTEfF0RJyXt+8aEY9GxMKIuDUiWuftW+XPX8iXd62xr//O25+LiKEl9kGSJGmzUGqQWwB8uo72zwOPr2fb94DPppT6Av2Aw/Pr0V0MXJZS6gasIjtsS36/KqW0O3BZvh4R0RMYDfQCDgeujYiWJfZDkiSp8EoNcucDV0bE2fm2IyPiZ8D/AD9a14Yp81b+tFV+S8Bngdvy9knA8PzxMflz8uWD8lmyxwCTU0rvpZQWAS8A+5XYD0mSpMIrKcillKYAJwFHkx1OvQDoDQxPKU1d3/YR0TIi5gErgGlk3xTxRkqpMl9lKdnhW/L7JfnrVgJvAjvVbK9jG0mSpC1Gqd/sQErpHuCeDXmxlNKHQL+I2B64HehR12r5fV3XqEvraF9DRJwOnA6wyy67bEi5kiRJzdoGXxB4Y6SU3gBmAPsD20dEVaDsDLycP14KdAHIl28HvF6zvY5tar7GxJTSgJTSgHbt2m2KbkiSJJVVqRcEXhURr9d3W8+27fKROCJia+Aw4BngAeDYfLVTgCn54zvz5+TL708ppbx9dD6rdVegGzCrlH6ocSxZsoRDDz2UHj160KtXL6644goAjj/+ePr160e/fv3o2rUr/fr1q95m/vz5HHDAAfTq1YvevXvz7rvvAnDrrbfSp08fevXqxXe/+92y9EeSpKIp9dDqt2s9bwXsTTZB4aL1bNsBmJTPMG0B/CaldFdELAAmR8SPyWa+3pCvfwNwU0S8QDYSNxogpfR0RPyGbAZtJfC1/JCtmlhFRQWXXnop/fv355///Cf77LMPgwcP5tZbb61e51vf+hbbbbcdAJWVlZx88sncdNNN9O3bl9dee41WrVrx2muv8Z3vfIe5c+fSrl07TjnlFKZPn86gQYPK1TVJkgqhpCCXUrqhrvaImAMcsp5t55OFvtrtL1LHrNOU0rtklzWpa18XkE20UBl16NCBDh06ALDtttvSo0cPli1bRs+ePQFIKfGb3/yG+++/H4CpU6fSp08f+vbtC8BOO+0EwIsvvkj37t2pOgR+2GGH8bvf/c4gJ0nSejTWOXLTyS4Loi3U4sWLefzxxxk4cGB120MPPcTOO+9Mt27dAHj++eeJCIYOHUr//v255JJLANh999159tlnWbx4MZWVldxxxx0sWbKkzteRJEn/UvKs1Xp8HnitkfalgnnrrbcYNWoUl19+OW3atKluv+WWWzjhhBOqn1dWVvLnP/+Z2bNns8022zBo0CD22WcfBg0axIQJEzj++ONp0aIFBx54IC+++GI5uiJJUqGUFOQi4nHWvNRHAJ8A2gFnNmJdKogPPviAUaNGcdJJJzFy5Mjq9srKSn7/+98zd+7c6rbOnTtzyCGH0LZtWwCGDRvGY489xqBBgzjqqKM46qijAJg4cSItW/plHZIkrU+ph1bvAu6ucbuT7Fy1viml6xq5NjVzKSVOO+00evTowdixY9dY9qc//Yk999yTzp07V7cNHTqU+fPns3r1aiorK3nwwQerz6dbsWIFAKtWreLaa6/lS1/6UtN1RJKkgip1ssP3N1UhKp6ZM2dy00030bt37+pLjFx44YUMGzaMyZMnr3FYFWCHHXZg7Nix7LvvvkQEw4YN48gjjwTgrLPO4oknngDg3HPPpXv37k3bGUmSCiiyS7Nt3gYMGJDmzJmzyV+n67i7N/lrbKzF448sdwmSJGkdImJuSmlAQ9Yt9Ry5D6jj67DqklJqXcq+JUmSVJpSZ60nTvInAAAgAElEQVR+CzgX+APwl7ztAOAo4IfAq41WmSRJktap1CD3WeCclNL/1WibGBFnAMNSSkc3XmmSJElal1JnrR5GdvHf2v4EeBl+SZKkJlTqiNxrwEjgklrtI4CVjVKRmoUiTNwAJ29IkrZspQa5HwLXR8Qh/Oscuf2Bw4HTG7EuSZIkrUep15H7eUQ8B5wFHEf2zQ4LgENSSjM3QX2SJEmqR8nftZpSehh4eBPUIkmSpBKUOtmBiGgXEd+IiCsjYqe8bf+I+GTjlydJkqT6lBTkImJv4DngNOAMYLt80RHAhY1bmtS0lixZwqGHHkqPHj3o1asXV1xxxRrLf/rTnxIRrFyZzetZtWoVI0aMoE+fPuy333489dRT1evee++97LHHHuy+++6MHz++SfshSdpylDoidylwbUqpN/BejfZ7gYMarSqpDCoqKrj00kt55plneOSRR7jmmmtYsGABkIW8adOmscsuu1Svf+GFF9KvXz/mz5/PL3/5S8466ywAPvzwQ772ta/xxz/+kQULFnDLLbdU70eSpMZUapDbB/h5He0vAztvfDlS+XTo0IH+/fsDsO2229KjRw+WLVsGwDe/+U0uueQSIqJ6/QULFjBoUHb5xD333JPFixezfPlyZs2axe67785uu+1G69atGT16NFOmTGn6DkmSNnulBrl3gTZ1tO+BX8+lzcjixYt5/PHHGThwIHfeeSedOnWib9++a6zTt29ffv/73wMwa9YsXnrpJZYuXcqyZcvo0qVL9XqdO3euDoSSJDWmUoPcH4BzI6JV/jxFxC7AeOD3jVqZVCZvvfUWo0aN4vLLL6eiooILLriA888/f631xo0bx6pVq+jXrx9XXXUVe++9NxUVFaSU1lq35kieJEmNpdTLj3yL7Hy4FcDWwIPAJ4BZwDmNW5rU9D744ANGjRrFSSedxMiRI3nyySdZtGhR9Wjc0qVL6d+/P7NmzeITn/gEP/95dqZBSoldd92VXXfdldWrV7NkyZLqfS5dupSOHTuWpT+SpM1bqRcEfjMiDgQGA/3JRvQeA+5LdQ1DSAWSUuK0006jR48ejB07FoDevXuzYsWK6nW6du3KnDlzaNu2LW+88QbbbLMNrVu35vrrr+fggw+mTZs27LvvvixcuJBFixbRqVMnJk+ezK9//etydUuStBlrcJDLD6fOAMaklKYCUzdVUVI5zJw5k5tuuonevXvTr18/IJuZOmzYsDrXf+aZZ/jCF75Ay5Yt6dmzJzfccAOQzX69+uqrGTp0KB9++CFjxoyhV69eTdYPSdKWo8FBLqX0QUR0Az7ahPVIZXPQQQfVeX5bTYsXL65+fMABB7Bw4cI61xs2bFi9AVCSpMZS6mSHm8guBixJkqQyK3WyQ2vgSxExGJgDvF1zYUppbGMVJkmSpHUrNcj1A+bnj3vWWuZkB0mSpCbUoCAXEX2Ap1JKn97E9UibRNdxd5e7hAZZPP7IcpcgSSqQhp4j9zjQtupJRNwdER02TUmSJElqiIYGudqXpT+Y7ILAkiRJKpNSZ61KkiSpmWhokEusPZnByQ2SJEll1NBZqwH8KiLey59/DPhZRKyuuVJK6ejGLE6SJEn1a2iQm1Tr+a8auxBJkiSVpkFBLqV06qYuRJIkSaVxsoMkSVJBGeQkSZIKyiAnSZJUUAY5SZKkgjLISZIkFZRBTpIkqaAMcpIkSQVlkJMkSSoog5y0mVqyZAmHHnooPXr0oFevXlxxxRUAvP766wwePJhu3boxePBgVq1atcZ2s2fPpmXLltx2223VbZMmTaJbt25069aNSZNqf9GLJKlcDHLSZqqiooJLL72UZ555hkceeYRrrrmGBQsWMH78eAYNGsTChQsZNGgQ48ePr97mww8/5Oyzz2bo0KHVba+//jrnnXcejz76KLNmzeK8885bK/xJksrDICdtpjp06ED//v0B2HbbbenRowfLli1jypQpnHLKKQCccsop3HHHHdXbXHXVVYwaNYr27dtXt913330MHjyYHXfckR122IHBgwdz7733Nm1nJEl1MshJW4DFixfz+OOPM3DgQJYvX06HDh2ALOytWLECgGXLlnH77bdzxhlnrLHtsmXL6NKlS/Xzzp07s2zZsqYrXpJUL4OctJl76623GDVqFJdffjlt2rSpd71vfOMbXHzxxbRs2XKN9pTSWutGRKPXKUkqXUW5C5C06XzwwQeMGjWKk046iZEjRwKw884788orr9ChQwdeeeWV6sOoc+bMYfTo0QCsXLmSe+65h4qKCjp37syMGTOq97l06VI+85nPNHVXJEl1cERO2kyllDjttNPo0aMHY8eOrW4/+uijq2eeTpo0iWOOOQaARYsWsXjxYhYvXsyxxx7Ltddey/Dhwxk6dChTp05l1apVrFq1iqlTp64xGUKSVD6OyEmbqZkzZ3LTTTfRu3dv+vXrB8CFF17IuHHjOO6447jhhhvYZZdd+O1vf7vO/ey44458//vfZ9999wXg3HPPZccdd9zk9UuS1s8gJ22mDjrooDrPbwOYPn36Orf9xS9+scbzMWPGMGbMmMYqTZLUSDy0KkmSVFAGOUmSpIIyyEmSJBVUkwW5iOgSEQ9ExDMR8XREnJW37xgR0yJiYX6/Q94eEXFlRLwQEfMjon+NfZ2Sr78wIk5pqj5IkiQ1J0052aES+FZK6bGI2BaYGxHTgC8C01NK4yNiHDAOOBs4AuiW3wYCE4CBEbEj8ANgAJDy/dyZUvLLH7XF6Dru7nKXsF6Lxx9Z7hIkabPXZCNyKaVXUkqP5Y//CTwDdAKOASblq00ChuePjwF+mTKPANtHRAdgKDAtpfR6Ht6mAYc3VT8kSZKai7KcIxcRXYG9gUeBnVNKr0AW9oCqb+vuBCypsdnSvK2+dkmSpC1Kkwe5iPg48DvgGymlf6xr1Tra0jraa7/O6RExJyLmvPrqqxtWrCRJUjPWpEEuIlqRhbibU0q/z5uX54dMye9X5O1LgS41Nu8MvLyO9jWklCamlAaklAa0a9eucTsiSZLUDDTlrNUAbgCeSSn9b41FdwJVM09PAabUaP9CPnt1f+DN/NDrfcCQiNghn+E6JG+TJEnaojTlrNV/B/4DeDIi5uVt/wOMB34TEacBfwM+ny+7BxgGvACsBk4FSCm9HhE/Ambn652fUnq9abogSZLUfDRZkEsp/Zm6z28DGFTH+gn4Wj37uhG4sfGqkyRJKh6/2UGSJKmgDHKSJEkFZZCTJEkqKIOcpEIYM2YM7du3Z6+99lqj/aqrrmKPPfagV69efPe7361unz9/PgcccAC9evWid+/evPvuuwDccsst9O7dmz59+nD44YezcuXKJu2HJDUmg5ykQvjiF7/Ivffeu0bbAw88wJQpU5g/fz5PP/003/72twGorKzk5JNP5rrrruPpp59mxowZtGrVisrKSs466yweeOAB5s+fT58+fbj66qvL0R1JahQGOUmFcPDBB7Pjjjuu0TZhwgTGjRvHVlttBUD79tk3/E2dOpU+ffrQt29fAHbaaSdatmxJSomUEm+//TYpJf7xj3/QsWPHpu2IJDUig5ykwnr++ed56KGHGDhwIIcccgizZ8+ubo8Ihg4dSv/+/bnkkksAaNWqFRMmTKB379507NiRBQsWcNppp5WzC5K0UQxykgqrsrKSVatW8cgjj/CTn/yE4447jpQSlZWV/PnPf+bmm2/mz3/+M7fffjvTp0/ngw8+YMKECTz++OO8/PLL9OnTh4suuqjc3ZCkDWaQk1RYnTt3ZuTIkUQE++23Hy1atGDlypV07tyZQw45hLZt27LNNtswbNgwHnvsMebNy75U5lOf+hQRwXHHHcfDDz9c5l5I0oYzyEkqrOHDh3P//fcD2eHU999/n7Zt2zJ06FDmz5/P6tWrqays5MEHH6Rnz5506tSJBQsW8OqrrwIwbdo0evToUc4uSNJGacrvWpWkDXbCCScwY8aM6hG38847jzFjxjBmzBj22msvWrduzaRJk4gIdthhB8aOHcu+++5LRDBs2DCOPPJIAH7wgx9w8MEH06pVKz75yU/yi1/8orwdk6SNYJCTVAi33HJLne2/+tWv6mw/+eSTOfnkk9dqP+OMMzjjjDMatTZJKhcPrUqSJBWUQU6SJKmgDHKSJEkFZZCTJEkqKCc7SCq7ruPuLncJ67V4/JHlLkGS1uKInCRJUkEZ5CRJkgrKICdJklRQBjlJkqSCMshJkiQVlEFOkiSpoAxykiRJBWWQk6QyGDNmDO3bt2evvfaqbvv+979Pnz596NevH0OGDOHll19eY5vZs2fTsmVLbrvtNgBeeukl9tlnH/r160evXr247rrrmrQPksrPICdJZfDFL36Re++9d42273znO8yfP5958+bxuc99jvPPP7962YcffsjZZ5/N0KFDq9s6dOjAww8/zLx583j00UcZP378WuFP0ubNICdJZXDwwQez4447rtHWpk2b6sdvv/02EVH9/KqrrmLUqFG0b9++uq1169ZstdVWALz33nt89NFHm7jq+pUywphS4r/+67/Yfffd6dOnD4899hjgCKO0IQxyktSMnHPOOXTp0oWbb765ekRu2bJl3H777Zxxxhlrrb9kyRL69OlDly5dOPvss+nYsWNTlwyUNsL4xz/+kYULF7Jw4UImTpzIV77yFcARRmlDGOQkqRm54IILWLJkCSeddBJXX301AN/4xje4+OKLadmy5Vrrd+nShfnz5/PCCy8wadIkli9f3tQlA6WNME6ZMoUvfOELRAT7778/b7zxBq+88kqzGmGUisIgJ0nN0Iknnsjvfvc7AObMmcPo0aPp2rUrt912G1/96le544471li/Y8eO9OrVi4ceeqgc5darvhHGLl26VK/TuXNnli1bBjSfEUapKAxyktRMLFy4sPrxnXfeyZ577gnAokWLWLx4MYsXL+bYY4/l2muvZfjw4SxdupR33nkHgFWrVjFz5kz22GOPstRen7pGGFNKa61XNVrXXEYYpaIwyElSGZxwwgkccMABPPfcc3Tu3JkbbriBcePGsddee9GnTx+mTp3KFVdcsc59PPPMMwwcOJC+fftyyCGH8O1vf5vevXs3UQ9KU3OEsXPnzixZsqR62dKlS9caeWuuI4xSc1NR7gIkaUt0yy23rNV22mmnrXe7X/ziF9WPBw8ezPz58xuzrEa1cOFCunXrBqw5wnj00Udz9dVXM3r0aB599FG22247OnTowNKlS9lpp53Yeuutq0cYx44d2+R1jxkzhrvuuov27dvz1FNPAdnEjT/84Q+0bt2aT33qU/z85z9n++23B+Ciiy7ihhtuoGXLllx55ZXVl4jp2rUr2267LS1btqSiooI5c+Y0eV+0+XNETpK00UoZYRw2bBi77bYbu+++O1/+8pe59tprgeYzwljXDNzBgwfz1FNPMX/+fLp3785FF10EwIIFC5g8eTJPP/009957L1/96lf58MMPq7d74IEHmDdvniFOm4wjcpKkjVbKCGNEcM0116zV3lxGGA8++GAWL168RtuQIUOqH++///7V364xZcoURo8ezVZbbcWuu+7K7rvvzqxZszjggAOasmRtwRyRkySpBDfeeCNHHHEEsO4ZuBHBkCFD2GeffZg4cWJZatXmzxE5SWpkXcfdXe4S1mvx+CPLXUIhXXDBBVRUVHDSSScB656BO3PmTDp27MiKFSsYPHgwe+65JwcffHCT1qvNnyNykiQ1wKRJk7jrrru4+eabq8PaumbgVt23b9+eESNGMGvWrKYvWps9g5wkSetx7733cvHFF3PnnXeyzTbbVLcfffTRTJ48mffee49FixaxcOFC9ttvP95++23++c9/Atm3WkydOnWN76GVGouHViVJ9SrCYWJo3EPFJ5xwAjNmzGDlypV07tyZ8847j4suuoj33nuPwYMHA9mEh+uuu45evXpx3HHH0bNnTyoqKrjmmmto2bIly5cvZ8SIEQBUVlZy4okncvjhhzdajVIVg5wkSTWUeo2/c845h3POOWeNtt12240nnnii0WuTavPQqiRJUkEZ5CRJkgrKICdJklRQBjlJkqSCcrKDJGmLUYRZuF6sWaVwRE6SJKmgDHKSJEkFZZCTJEkqKIOcJElSQRnkJEmSCsogJ0mSVFAGOUmSpIIyyEmStAW57LLL6NWrF3vttRcnnHAC7777LtOnT6d///7069ePgw46iBdeeAGAb37zm/Tr149+/frRvXt3tt9++zJXr9oMcpIkbSGWLVvGlVdeyZw5c3jqqaf48MMPmTx5Ml/5yle4+eabmTdvHieeeCI//vGPgSz0zZs3j3nz5vH1r3+dkSNHlrkHa6orlKaUOOecc+jevTs9evTgyiuvBGDGjBlst9121cH0/PPPL3P1jaPJvtkhIm4EPgesSCntlbftCNwKdAUWA8ellFZFRABXAMOA1cAXU0qP5ducAnwv3+2PU0qTmqoPkiQVXWVlJe+88w6tWrVi9erVdOzYkYjgH//4BwBvvvkmHTt2XGu7W265hfPOO6+py61XVShdsGABW2+9NccddxyTJ08mpcSSJUt49tlnadGiBStWrKje5tOf/jR33XVXGatufE35FV2/AK4GflmjbRwwPaU0PiLG5c/PBo4AuuW3gcAEYGAe/H4ADAASMDci7kwprWqyXkiSVFCdOnXi29/+Nrvssgtbb701Q4YMYciQIVx//fUMGzaMrbfemjZt2vDII4+ssd1LL73EokWL+OxnP1umyutWVyj93ve+x69//WtatMgOOrZv377MVW5aTXZoNaX0/4DXazUfA1SNqE0Chtdo/2XKPAJsHxEdgKHAtJTS63l4mwYcvumrlySp+FatWsWUKVNYtGgRL7/8Mm+//Ta/+tWvuOyyy7jnnntYunQpp556KmPHjl1ju8mTJ3PsscfSsmXLMlW+tpqhtEOHDmy33XYMGTKEv/71r9x6660MGDCAI444goULF1Zv85e//IW+fftyxBFH8PTTT5ex+sZT7nPkdk4pvQKQ31fF5k7AkhrrLc3b6mtfS0ScHhFzImLOq6++2uiFS5JUNH/605/YddddadeuHa1atWLkyJHMnDmTJ554goEDBwJw/PHH8/DDD6+x3eTJkznhhBPKUXK96gul7733Hh/72MeYM2cOX/7ylxkzZgwA/fv356WXXuKJJ57g61//OsOHD1/PKxRDuYNcfaKOtrSO9rUbU5qYUhqQUhrQrl27Ri1OkqQi2mWXXXjkkUdYvXo1KSWmT59Oz549efPNN3n++ecBmDZtGj169Kje5rnnnmPVqlUccMAB5Sq7TnWF0ocffpjOnTszatQoAEaMGMH8+fMBaNOmDR//+McBGDZsGB988AErV64sW/2NpSnPkavL8ojokFJ6JT90WnVG4lKgS431OgMv5+2fqdU+ownqlCSp8AYOHMixxx5L//79qaioYO+99+b000+vDj8tWrRghx124MYbb6ze5pZbbmH06NFk8xCbj5qhdOutt2b69OkMGDCANm3acP/99zNmzBgefPBBunfvDsDf//53dt55ZyKCWbNm8dFHH7HTTjuVuRcbr9xB7k7gFGB8fj+lRvuZETGZbLLDm3nYuw+4MCJ2yNcbAvx3E9csSVJhnXfeeWvNPh0xYgQjRoyoc/0f/vCHTVBV6eoLpe+88w4nnXQSl112GR//+Me5/vrrAbjtttuYMGECFRUVbL311kyePLnZhdMN0ZSXH7mFbDStbUQsJZt9Oh74TUScBvwN+Hy++j1klx55gezyI6cCpJRej4gfAbPz9c5PKdWeQCFJkrYAdYXSrbbairvvvnutdc8880zOPPPMpiqtyTRZkEsp1XeW5KA61k3A1+rZz43AjXUtkyRJ2pI018kOkiRJWo9ynyMnSZI2UNdxax9CbG4Wjz+y3CVs1hyRkyRJKihH5CRJUrPgCGPpHJGTJEkqKIOcJElSQRnkJEmSCsogJ0mSVFAGOUmSpIIyyEmSJBWUQU6SJKmgDHKSJEkFZZCTJEkqKIOcJElSQRnkJEmSCsogJ0mSVFAGOUmSpIIyyEmSJBWUQU6SJKmgDHKSJEkFZZCTJEkqKIOcJElSQRnkJEmSCsogJ0mSVFAGOUmSpIIyyEmSJBWUQU6SJKmgDHKSJEkFZZCTJEkqKIOcJElSQRnkJEmSCsogJ0mSVFAGOUmSpIIyyEmSJBWUQU6SJKmgDHKSJEkFZZCTJEkqKIOcJElSQRnkJEmSCsogJ0mSVFAGOUmSpIIyyEmSJBWUQU6SJKmgDHKSJEkFZZCTJEkqKIOcJElSQRnkJEmSCsogJ0mSVFAGOUmSpIIyyEmSJBWUQU6SJKmgDHKSJEkFZZCTJEkqKIOcJElSQRU2yEXE4RHxXES8EBHjyl2PJElSUytkkIuIlsA1wBFAT+CEiOhZ3qokSZKaViGDHLAf8EJK6cWU0vvAZOCYMtckSZLUpIoa5DoBS2o8X5q3SZIkbTEipVTuGkoWEZ8HhqaUvpQ//w9gv5TS12usczpwev50D+C5Ji9047UFVpa7iEZkf5q3zak/m1NfwP40d5tTfzanvkBx+/PJlFK7hqxYsakr2USWAl1qPO8MvFxzhZTSRGBiUxbV2CJiTkppQLnraCz2p3nbnPqzOfUF7E9ztzn1Z3PqC2x+/alLUQ+tzga6RcSuEdEaGA3cWeaaJEmSmlQhR+RSSpURcSZwH9ASuDGl9HSZy5IkSWpShQxyACmle4B7yl3HJlboQ8N1sD/N2+bUn82pL2B/mrvNqT+bU19g8+vPWgo52UGSJEnFPUdOkiRpi2eQK6OI2D4ivpo//kxE3FXumppCRPxXRDwTETeXu5aGqvmz2hJFxFvlrmF9IuLhcteg+hXx57Ox/0ZHxBcjouOmqW7DNLRPEXG935hUDAa58toe2BLDwVeBYSmlk8pdSAm21J9VYaSUDix3DapfQX8+G/t7/0WgWQU5GtinlNKXUkoLmqAebSSDXHmNBz4VEfOAnwAfj4jbIuLZiLg5IgIgIvaJiAcjYm5E3BcRHcpadQkiYmxEPJXfvhER1wG7AXdGxDfLXV8Jqn9WEfGT/PZURDwZEceXu7iGiIg78s/Q0/kFs4mItyLigoh4IiIeiYid8/ZdI+IvETE7In5U3sobpmrUMB9lmFHX71IR1fVzK6IaP58OEfH/8t+lpyLi0+WubR0a+m/0ufnvylMRMTEyxwIDgJvzvm5dxn7U1NA+zYiIARHRMiJ+UePfu2b973Ydf3O65keAfpb/Dk1tRj+LxpFS8lamG9AVeCp//BngTbKLG7cA/gIcBLQCHgba5esdT3a5lbLX34D+7QM8Cfwb8HHgaWBvYDHQttz1bcTPahQwjezSNzsDfwP+f3v3H+tVXcdx/PmCNH5EVGaLtVVqqJhXyfwRy/JCjFhbpYsma+micrFc0OZwlbBhf9R0WYLUdOCkkbjcrAVbWTAvID/EJH5Iuiw0DUwNuoqXH4Lw7o/P5xuH6/fe772Xyz33XF6P7bvv+X7u55zzPvd8z/m+z+f8+IwqO8YuLMN78vtQYDtwBhDA53P57cDsPLwMuD4P3wi0lR1/F5avLb/X3ZbKjq8311vZMZ3g+rkJuCUPDwZGlB1bJzE33EcX11EeXlLYplYBl5a9HD1cplWkRPTjwIrC+O8qexk6WbaOfnPeBMbmOg8CXy071t58uUWuf3k8InZGxFFgC2mDOw+4EFiRj6Bmkza6KrgS+G1E7IuINuA3QH8++u6qK4EHIuJIRLwMrAYuKzmmrpghaSvwGKlnlNHAIaB2jcwm0ncO4JPAA3l4SR/G2FvqbUtVVW+9VdmfgWmS5gJNEfF6yfF0R0ffq/GSNkp6EpgAfLSsAHug0bbyLHC2pLskTQb29nWA3dDRb85zEbEl1ynu5wYEJ3L9yxuF4SOk5/wJ+GtEjM2vpoiYVE543VbZ01kNVG65JDUDE4FxEXExsBkYAhyOfJjKse9cTZWfTVRvW6qcTtZbZUXEGuDTwC5giaTrSw6pO97yvZI0BPgFMCUimoCFVGsddbqtREQrcDGphe5GYFGfRdZ9He2bB8T+oCNO5Mr1OjCiQZ2/AWdKGgcg6TRJVTnaWwNcLWmYpOHANcCjJcfUU8V1tQa4Nl87cibpR+nx0iLrmpFAa0Tsl3Q+8IkG9deRur4DqNJNKQNNd9dbvyfpQ8ArEbEQuBe4pOSQOtOVfXQtadst6R3AlG6O39e6FZOk9wKDIuIhYA79e30NpN+cLhtQWWnVRMQeSeskbQcOAC/XqXMoXzQ7X9JI0jq7k3Tuv1+LiL9IWsyxJGdRRGyu4nXn7dbVH4BtwFZSq9XNEfFSqQE29jAwXdI20sHBYw3qzwSWSpoJPHSyg7MOdXe9VUEzMEvSYaAN6Lctcl3cR78qaSHp2qx/kk4d1ywG7pZ0gNSqeuDkR925rixTOx8A7pNUa/j5/kkN8ATU+80BWsuLqG+4ZwczMzOzivKpVTMzM7OKciJnZmZmVlFO5MzMzMwqyomcmZmZWUU5kTMzMzOrKCdyZnbKkrRA0qo+mE+zpMjP5DIz6zVO5Mysz0maLmmfpNMLZadL2p+7OSrWHZ2ToAl9H2mvWQ+MAvaUHYiZDSxO5MysDI8Aw4DLC2VXkDrwPjf3mFHTTOpiZ31PZiRpkKTBPYyzV0TEoYh4KU7gwZ2S3qYqPk3bzE4qJ3Jm1uci4hngRWB8oXg8sBJ4gpS8Fcs3RMRBAEnvlvRLSa2SDkhaWey2TtLXJLVJ+lx+ev0hYEzuUu0nebxWSXcCnSZ49U6JSvpwLru0XZ3P5I7T90t6QtIlDabzdUkv5PrLJX1bUhT+PlfS9rw8O0jJ7HBJkyU9mpfhv5L+KGlMnfimSlqd/0ebJV0k6UJJ63Nr6FpJZzVeW2bWnzmRM7OytPDWRI2XQ00AAAOGSURBVG5VfhXLm3PdmsWk1rsvklr09gMPSxpaqDMEmA18C7gAeB64Cbghl40jJXG92Y/sj4Hvkfqi3APc31ELWu47eRHwc2AssAy4tU7Vs4CvAF8mdVx+EBhO6qbvctL/5jVgefE0dXYrcBvwMeBVYClwF3BLHncIML9HS2pm/Yb7WjWzsrQACyS9HRCpQ/hvAv8C5gHkjuJHkU7FImk08AXgqohYk8uuA14gJWWL8rQHA9+JiE21mUn6LnB7RDyYP88EPtuLyzMnIlrytH8IrCX1U7mzTt0ZwJ8i4rb8+RlJl5ESzaLTgesiotgf5nF930qaBuwlJWdrC3/6aUT8Pte5A1gOfKkQ4wJgQbeX0sz6FbfImVlZWkitQuPya3dE7ADWAedIej+pZW4/sDGPMwY4CmyoTSQiXiN1WH5BYdpvAltqHySNJCWExfGOFqbbG7YVhl/M7+/roO75HOvYu6ZeLDvbJXFIOkfSUkk7JO0ldXo+CPhgJ/HUpvFku7LhkoZ1EKOZVYBb5MysFBHxrKTnSacHRTqlSkTsk7QplzcDayPicB6ts4v9izcSvBERR3ohzKN15ntaB3UPF4ZrsXR0sCyOj7cj++qULQd2kU4R7yIlrU+RWu8axdOdGM2sArwBm1mZatfJ1a6Pq1kFTCAlco8Uyp8i7bfG1QokvRNoyn+rK7fa/Zt0+rY2njj+rtl6/pPfRxXKxjYYpyuerjPvRrEg6QxSq+SPImJlRDwNjMAH5WanLCdyZlamFlJydQXHJ3KrgamkU5P/v9EhIv4O/A64R9KnJDUBvyJdI7a0wbzmATdLmiLpPNINA6MajPMP0jV7cyWdK2kS6SaKEzUfmCRpVn5O3jeAa7owXiuwG7hB0kckXQXcTWqVM7NTkBM5MytTC+mU4Cv5+riatcBQUoK2qd0400jXly3L78OAyRFxoMG87gDuI90QsZG0/7u/sxHyKd2pwNnAVtKdoD9ouFQNRMQG0o0NM0jXsl1NusP0YIPxjgLXAhcB20l3vc4hPZrEzE5BOoHnU5qZWS+R9DNgYkQ0lR2LmVWHr6swMyuBpFnACqANmAhMpxda+8zs1OIWOTOzEkj6NelmjpHAc8A9wLwT6cbLzE49TuTMzMzMKso3O5iZmZlVlBM5MzMzs4pyImdmZmZWUU7kzMzMzCrKiZyZmZlZRTmRMzMzM6uo/wHZnC8KxFFW8gAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 720x504 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "labels, values = zip(*english_unigram_freqs[:10])\n",
    "indexes = arange(len(labels))\n",
    "width = 0.8 # width = 1 would give bars that overlap because they are too close.\n",
    "\n",
    "fig = plt.figure(figsize=(10,7))                                                               \n",
    "ax = fig.gca() # Get current axis\n",
    "rects = ax.bar(indexes, values, width)\n",
    "\n",
    "# Add title and axis labels\n",
    "fig.suptitle('Top 10 English word unigrams', fontsize=20)\n",
    "plt.xlabel('Word unigram', fontsize=14)\n",
    "plt.ylabel('Frequency', fontsize=14)\n",
    "\n",
    "# Display value of each bar on bar\n",
    "for rect in rects:\n",
    "        height = rect.get_height()\n",
    "        ax.text(rect.get_x() + rect.get_width() / 2., 50 + height, '%d' % int(height), ha='center', va='bottom') # Can also add color and fontweight arguments.\n",
    "\n",
    "# Remove the default x-axis tick numbers and use tick numbers of your own choosing:\n",
    "ax.set_xticks(indexes)\n",
    "# Replace the tick numbers with strings:\n",
    "ax.set_xticklabels(labels)\n",
    "\n",
    "plt.show()\n",
    "# plt.savefig('top10EnglishWordUnigrams.png')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. Generating unigram and bigram frequencies for English, French and Italian from training files"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_ngram_count_dict(tokens, n):\n",
    "    if n == 1:\n",
    "        n_grams = ngrams(tokens, n)\n",
    "    else:\n",
    "        n_grams = ngrams(tokens, n, pad_left=True, pad_right=True, left_pad_symbol='_', right_pad_symbol='_') # Fun fact: If I remove padding here and later when testing, and also remove the '_' from the unigram dicts, the accuracy rises slightly. However, it's not statistically significant due to the small size of the data.\n",
    "    fdist = FreqDist(n_grams)\n",
    "    ngram_dict = dict()\n",
    "    for k,v in fdist.items():\n",
    "        ngram_dict[' '.join(k)] = v\n",
    "    return ngram_dict\n",
    "\n",
    "# Calls get_ngram_count_dict to get a unigram and bigram dict from file.\n",
    "def get_unigram_bigram_dicts(file):\n",
    "    with open(file, encoding='utf8') as f:\n",
    "        content = f.read()\n",
    "    tokens = ultimate_tokenize(content)\n",
    "    unigram_dict = get_ngram_count_dict(tokens, 1)     \n",
    "    bigram_dict = get_ngram_count_dict(tokens, 2)     \n",
    "    return (unigram_dict, bigram_dict)\n",
    "\n",
    "# Dumps unigram and bigram dictionary of training data of given language to .pickle files.\n",
    "def dump_pickle(language):\n",
    "    file = 'ngram_langid_files/LangId.train.' + language + '.txt'\n",
    "    unigram_dict, bigram_dict = get_unigram_bigram_dicts(file)\n",
    "    with open('ngram_langid_files/' + language + '.unigram.pickle', 'wb') as handle:\n",
    "        pickle.dump(unigram_dict, handle, protocol=pickle.HIGHEST_PROTOCOL) # HIGHEST_PROTOCOL instructs pickle to use the highest protocol version available.\n",
    "    with open('ngram_langid_files/' + language + '.bigram.pickle', 'wb') as handle:\n",
    "        pickle.dump(bigram_dict, handle, protocol=pickle.HIGHEST_PROTOCOL)\n",
    "        \n",
    "dump_pickle('English')\n",
    "dump_pickle('French')\n",
    "dump_pickle('Italian')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Later, it will also be required to know how many sentences there are in the training data for each language. This is because of the method used to calculate probabilities (incorporating the probability of the bigram among other bigrams starting with the same word) and the fact we use padding for our bigrams. \n",
    "\n",
    "In our training data each line is a sentence, which is very convenient for calculating the number of sentences.\n",
    "\n",
    "We go ahead and get the number of sentences (for more efficiency, the following code could be added to `get_unigram_bigram_dicts`):"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "NUMBER OF SENTENCES IN TRAINING DATA\n",
      "English: 2980\n",
      "French: 2980\n",
      "Italian: 2821\n"
     ]
    }
   ],
   "source": [
    "with open('ngram_langid_files/LangId.train.English.txt', encoding='utf8') as f:\n",
    "    for i, l in enumerate(f):\n",
    "        pass\n",
    "number_of_sents_en = i + 1\n",
    "with open('ngram_langid_files/LangId.train.French.txt', encoding='utf8') as f:\n",
    "    for i, l in enumerate(f):\n",
    "        pass\n",
    "number_of_sents_fr = i + 1\n",
    "with open('ngram_langid_files/LangId.train.Italian.txt', encoding='utf8') as f:\n",
    "    for i, l in enumerate(f):\n",
    "        pass\n",
    "number_of_sents_it = i + 1\n",
    "\n",
    "print('NUMBER OF SENTENCES IN TRAINING DATA')\n",
    "print('English:', number_of_sents_en)\n",
    "print('French:', number_of_sents_fr)\n",
    "print('Italian:', number_of_sents_it)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. Identifying language for each line of the test file using bigram probabilities"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "22705"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "with open('ngram_langid_files/English.unigram.pickle', 'rb') as handle:\n",
    "    unigram_english_dict = pickle.load(handle)\n",
    "    \n",
    "with open('ngram_langid_files/English.bigram.pickle', 'rb') as handle:\n",
    "    bigram_english_dict = pickle.load(handle)\n",
    "    \n",
    "with open('ngram_langid_files/French.unigram.pickle', 'rb') as handle:\n",
    "    unigram_french_dict = pickle.load(handle)\n",
    "    \n",
    "with open('ngram_langid_files/French.bigram.pickle', 'rb') as handle:\n",
    "    bigram_french_dict = pickle.load(handle)\n",
    "    \n",
    "with open('ngram_langid_files/Italian.unigram.pickle', 'rb') as handle:\n",
    "    unigram_italian_dict = pickle.load(handle)\n",
    "    \n",
    "with open('ngram_langid_files/Italian.bigram.pickle', 'rb') as handle:\n",
    "    bigram_italian_dict = pickle.load(handle)\n",
    "    \n",
    "vocabulary_size = len(unigram_english_dict) + len(unigram_french_dict) + len(unigram_italian_dict)\n",
    "vocabulary_size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 98.00%\n"
     ]
    }
   ],
   "source": [
    "# Get probability of given bigram belonging to the language which bigram_dict is in\n",
    "def get_bigram_probability(bigram, first_word, bigram_dict, first_word_dict): # first_word is the first word of the word bigram.\n",
    "    bigram_count = bigram_dict.get(bigram)\n",
    "    if bigram_count is None:\n",
    "        bigram_count = 0\n",
    "    \n",
    "    first_word_count = first_word_dict.get(first_word)\n",
    "    if first_word_count is None:\n",
    "        first_word_count = 0\n",
    "    \n",
    "    return (bigram_count + 1) / (first_word_count + vocabulary_size) # To get the logic of this formula, note how the proability is used in the function below. Without the + 1 in the Nr, if you find a bigram which is not in our known bigrams for a language, the probability of it being in that language would become 0. So we would like to assign a small probability of 1 / vocabulary_size in that case. Also note the arbitrariness of this 'probability'. We're saying \"Given a bigram and a language, what is the probability that the bigram is of that language?\" This is arbitrary because to get a meaningful probability we need to know which are the other languages considered and what their bigram frequencies are. That would be another way to do it, but arguable a worse one because it wouldn't be able to give a confidence score for a particular language. The formula just uses common sense to get to a number which works for the purposes. In the denominator, we have both first_word_count and vocabulary_size. Why? We have vocabulary_size for all langs in the denom because the larger this is, the less significant it is that for this particular language the bigram appears so many times. Could we have used a vocab_size of bigrams instead of unigrams? Sure, and the 'probabilities' would end up being much smaller numbers. What about first_word_count? This gives us a way to compare this bigram against other bigrams in this language starting with the same word. In general though, for a given bigram, it's more important to consider how many times it exists than to consider whether it is the usual bigram given a certain first word. The formula achieves that. Take the bigram 'le monseiur' and the English language. Let's say the bigram appears once and 'le' also appears once, while in French 'le monseiur' appears 100 times and le appears 100,000 times. Probability for English = (1 + 1) / (1 + 20,000) = 0.000099995. Probability for French = (100 + 1) / (100,000 + 20,000) = 0.00084166666. Note how the probability for French is still low because 100/100,000 is quite low and maybe it's not French after all if in French le is usually followed by other words. However, it's still significantly higher than the probability for English where both 'le' and 'le monseiur' only appear once.\n",
    "\n",
    "# Get probability that a given bigram list is of a language (specified by its bigram_dict)\n",
    "def get_language_probability(bigram_list, first_words, bigram_dict, first_word_dict):\n",
    "    result = 1.0\n",
    "    index = 0\n",
    "    for bigram in bigram_list:\n",
    "        result *= get_bigram_probability(bigram, first_words[index], bigram_dict, first_word_dict)\n",
    "        index += 1\n",
    "    return result\n",
    "\n",
    "# Load correct solutions\n",
    "solution_dict = dict()\n",
    "with open('ngram_langid_files/LangId.sol.txt') as f:\n",
    "    for line in f:\n",
    "       (key, val) = line.split()\n",
    "       solution_dict[int(key)] = val\n",
    "        \n",
    "line_no = 1\n",
    "result_dict = dict()\n",
    "correct = 0\n",
    "incorrect_line_numbers = []\n",
    "\n",
    "# This needs to be done because I'm using padding for bigrams so the unigram dicts in their raw forms can't be used in get_bigram_probability():\n",
    "unigram_english_dict['_'] = number_of_sents_en\n",
    "unigram_french_dict['_'] = number_of_sents_fr\n",
    "unigram_italian_dict['_'] = number_of_sents_it\n",
    "\n",
    "with open('ngram_langid_files/LangId.test.txt', encoding='utf8') as f:\n",
    "    for line in f:\n",
    "        tokens = ultimate_tokenize(line)\n",
    "        bigrams = ngrams(tokens, 2, pad_left=True, pad_right=True, left_pad_symbol='_', right_pad_symbol='_')\n",
    "        bigram_list = [] # bigram_list will be exactly like bigrams but instead of [('_', 'this'), ...] it will be ['_ this', ...]. It is required because this is how bigrams are represented in the dictionary.\n",
    "        first_words = [] # The first words of each bigram. This is the similar to making a unigram_list. We use it because we don't want something in the form [(this,), ...]. Also because we want this to include '_'. We want it to include '_' because we're not using the unigrams for classification but as part of a formula to judge bigram frequency based on the starting word.\n",
    "        for b in bigrams:\n",
    "            bigram_list.append(' '.join(b))\n",
    "            first_words.append(b[0])\n",
    "        \n",
    "        english_prob = get_language_probability(bigram_list, first_words, bigram_english_dict, unigram_english_dict)\n",
    "        french_prob = get_language_probability(bigram_list, first_words, bigram_french_dict, unigram_french_dict)\n",
    "        italian_prob = get_language_probability(bigram_list, first_words, bigram_italian_dict, unigram_italian_dict)\n",
    "        \n",
    "        max_prob = max(english_prob, french_prob, italian_prob)\n",
    "        if max_prob == english_prob:\n",
    "            result_dict[line_no] = 'English'\n",
    "        elif max_prob == french_prob:\n",
    "            result_dict[line_no] = 'French'\n",
    "        else:\n",
    "            result_dict[line_no] = 'Italian'\n",
    "        \n",
    "        if solution_dict[line_no] == result_dict[line_no]:\n",
    "            correct += 1\n",
    "        else:\n",
    "            incorrect_line_numbers.append(line_no)\n",
    "            \n",
    "        line_no += 1\n",
    "\n",
    "# Storing results from result_dict to file:\n",
    "with open('ngram_langid_files/LangId.result.txt', 'w') as f:\n",
    "    for (key, val) in result_dict.items():\n",
    "        f.write(' '.join([str(key), val]) + '\\n')\n",
    "        \n",
    "print('Accuracy: {:2.2f}%'.format(correct * 100 / len(solution_dict)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Line numbers for incorrectly classified languages: [24, 87, 187, 191, 247, 279]\n"
     ]
    }
   ],
   "source": [
    "print('Line numbers for incorrectly classified languages: {}'.format(str(incorrect_line_numbers)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3. Testing with our own sentence"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Sentence bigrams: ['_ this', 'this is', 'is a', 'a sentence', 'sentence _']\n",
      "Sentence bigrams first words: ['_', 'this', 'is', 'a', 'sentence']\n"
     ]
    }
   ],
   "source": [
    "sent = \"This is a sentence.\"\n",
    "sent_tokens = ultimate_tokenize(sent)\n",
    "sent_bigrams_pre = ngrams(sent_tokens, 2, pad_left=True, pad_right=True, left_pad_symbol='_', right_pad_symbol='_')\n",
    "sent_bigrams = []\n",
    "sent_bigrams_first_words = []\n",
    "for b in sent_bigrams_pre:\n",
    "    sent_bigrams.append(' '.join(b))\n",
    "    sent_bigrams_first_words.append(b[0])\n",
    "print('Sentence bigrams:', sent_bigrams)\n",
    "print('Sentence bigrams first words:', sent_bigrams_first_words)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "RAW 'PROBABILITIES'\n",
      "English: 1.6907852103961136e-18\n",
      "French: 1.4294173998895933e-22\n",
      "Italian: 1.4086603160283195e-22\n"
     ]
    }
   ],
   "source": [
    "sent_english_prob = get_language_probability(sent_bigrams, sent_bigrams_first_words, bigram_english_dict, unigram_english_dict)\n",
    "sent_french_prob = get_language_probability(sent_bigrams, sent_bigrams_first_words, bigram_french_dict, unigram_french_dict)\n",
    "sent_italian_prob = get_language_probability(sent_bigrams, sent_bigrams_first_words, bigram_italian_dict, unigram_italian_dict)\n",
    "print(\"RAW 'PROBABILITIES'\")\n",
    "print('English:', sent_english_prob)\n",
    "print('French:', sent_french_prob)\n",
    "print('Italian:', sent_italian_prob)\n",
    "# As we can see, these 'probabilities' are arbitrary. We can try to convert them to percentages since we are classifying only among these 3 languages:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "RELATIVE PROBABILITIES\n",
      "English: 99.98%\n",
      "French: 0.01%\n",
      "Italian: 0.01%\n"
     ]
    }
   ],
   "source": [
    "def get_normalized_probabilities(list_of_probabilities):\n",
    "    sum_of_probabilities = sum(list_of_probabilities)\n",
    "    result = []\n",
    "    for probability in list_of_probabilities:\n",
    "        result.append(probability / sum_of_probabilities)\n",
    "    return result\n",
    "\n",
    "probabilities = [sent_english_prob, sent_french_prob, sent_italian_prob]\n",
    "normalized_probabilities = get_normalized_probabilities(probabilities)\n",
    "\n",
    "print('RELATIVE PROBABILITIES')\n",
    "print('English: ', round(normalized_probabilities[0] * 100, 2), '%', sep='') # I use sep because I don't want a space before the % sign.\n",
    "print('French: ', round(normalized_probabilities[1] * 100, 2), '%', sep='')\n",
    "print('Italian: ', round(normalized_probabilities[2] * 100, 2), '%', sep='')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**PS:** For a state-of-the-art Greek dialect classifier using n-grams, take a look at [Greek Dialect Classifier](https://github.com/hb20007/greek-dialect-classifier)."
   ]
  }
 ],
 "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": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}