3-3-2_versification_vs_lexical.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div style='color:white;font-size:30px;background-color:#196363;text-align:center;padding-top:20px; padding-bottom:20px;margin-bottom:40px'>VERSIFICATION & AUTHORSHIP ATTRIBUTION</div>\n",
    "<h1>3.3.2 Versification based vs. lexical based attribution</h1>\n",
    "<h2 style='margin-bottom: 50px'>Replication code</h2>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "import matplotlib.pyplot as plt \n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "import sticho\n",
    "from collections import defaultdict\n",
    "from scipy import stats"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<h3>Read data from JSON</h3>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "#----------------------------------------------------------------------------------------------------------#\n",
    "# Files contain a list of poems with following structure:                                                  #   \n",
    "# [{                                                                                                       #\n",
    "#    'author': string (name of the author),                                                                #\n",
    "#    'title': string (title of the poem),                                                                  #\n",
    "#    'subcorpus': int (id of the subcorpus to which the poem belongs)                                      #\n",
    "#    'body': [{                                                                                            #\n",
    "#        'text': string (text of the line)                                                                 #\n",
    "#        'metre': string (metre of the line)                                                               #\n",
    "#        'stress_pattern': string (bit string representing stressed/unstressed syllables)                  #\n",
    "#        'rhyme': list (indices of lines with which this one rhymes)                                       #\n",
    "#        'words': [{                                                                                       #\n",
    "#            'token': string,                                                                              #\n",
    "#            'lemma': string,                                                                              #\n",
    "#            'tag': string (POS-tag)                                                                       #\n",
    "#            'xsampa': string (phonetic transcription of word in XSampa)                                   #\n",
    "#         }, ...]                                                                                          #\n",
    "#     }, ... ]                                                                                             #\n",
    "# }, ...]                                                                                                  #\n",
    "#                                                                                                          #\n",
    "# CZECH CORPUS                                                                                             #\n",
    "# Dataset comes from The Corpus of Czech Verse (http://versologie.cz/v2/web_content/corpus.php?lang=en).   #\n",
    "#                                                                                                          #\n",
    "# SPANISH CORPUS                                                                                           #\n",
    "# Dataset comes from Corpus de Sonetos del Siglo de Oro                                                    #\n",
    "#    (https://github.com/bncolorado/CorpusSonetosSigloDeOro).                                              #\n",
    "# Tokenization, lemmatization, pos-tagging provided by TreeTagger                                          #\n",
    "#    (https://www.cis.uni-muenchen.de/~schmid/tools/TreeTagger/).                                          #\n",
    "# Phonetic transcription provided by ESpeak (http://espeak.sourceforge.net/)                               #\n",
    "#                                                                                                          #\n",
    "# GERMAN CORPUS                                                                                            #\n",
    "# Dataset not published here as it comes from proprietary corpus Metricalizer (http://metricalizer.de).    #\n",
    "#----------------------------------------------------------------------------------------------------------#\n",
    "\n",
    "data = dict()\n",
    "with open('data/cs.json', 'r') as f:\n",
    "    data['cs'] = json.load(f)\n",
    "with open('data/es.json', 'r') as f:\n",
    "    data['es'] = json.load(f)\n",
    "    \n",
    "# Subcorpora available for each corpus\n",
    "subcorpora = {'cs': (1,2,3), 'es': (1,2)}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<h3>Cross-validation</h3>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iteration #30 | mft = 150 | lang = cs | subc = 1              \n",
      "\tp(Wilcoxon) = 1.1442144430510537e-05\n",
      "iteration #30 | mft = 250 | lang = cs | subc = 1              \n",
      "\tp(Wilcoxon) = 0.0002981699421895982\n",
      "iteration #30 | mft = 500 | lang = cs | subc = 1              \n",
      "\tp(Wilcoxon) = 2.1577047385332823e-05\n",
      "iteration #30 | mft = 150 | lang = cs | subc = 2              \n",
      "\tp(Wilcoxon) = 8.084788301662718e-06\n",
      "iteration #30 | mft = 250 | lang = cs | subc = 2              \n",
      "\tp(Wilcoxon) = 7.61946721684823e-06\n",
      "iteration #30 | mft = 500 | lang = cs | subc = 2              \n",
      "\tp(Wilcoxon) = 7.737777422379613e-05\n",
      "iteration #30 | mft = 150 | lang = cs | subc = 3              \n",
      "\tp(Wilcoxon) = 1.5983101993404364e-06\n",
      "iteration #30 | mft = 250 | lang = cs | subc = 3              \n",
      "\tp(Wilcoxon) = 1.6363965084210647e-06\n",
      "iteration #30 | mft = 500 | lang = cs | subc = 3              \n",
      "\tp(Wilcoxon) = 1.5942375258103098e-06\n",
      "iteration #30 | mft = 150 | lang = es | subc = 1              \n",
      "\tp(Wilcoxon) = 0.8724228556121767\n",
      "iteration #30 | mft = 250 | lang = es | subc = 1              \n",
      "\tp(Wilcoxon) = 0.23623560111787156\n",
      "iteration #30 | mft = 500 | lang = es | subc = 1              \r"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/scipy/stats/morestats.py:2879: UserWarning: Sample size too small for normal approximation.\n",
      "  warnings.warn(\"Sample size too small for normal approximation.\")\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\tp(Wilcoxon) = 1.0\n",
      "iteration #30 | mft = 150 | lang = es | subc = 2              \n",
      "\tp(Wilcoxon) = 0.15935193824976623\n",
      "iteration #30 | mft = 250 | lang = es | subc = 2              \n",
      "\tp(Wilcoxon) = 0.2093193697720107\n",
      "iteration #30 | mft = 500 | lang = es | subc = 2              \n",
      "\tp(Wilcoxon) = 0.7684977098758217\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x288 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x288 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x288 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x288 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x288 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Container to store predictions\n",
    "decisions = defaultdict(lambda: defaultdict(lambda: defaultdict(lambda: defaultdict(list))))\n",
    "acc = defaultdict(lambda: defaultdict(lambda: defaultdict(list)))\n",
    "\n",
    "# Iterate over subcorpora\n",
    "for lang in sorted(data):\n",
    "    for subc in subcorpora[lang]:\n",
    "        \n",
    "        # Iterate over different mft settings\n",
    "        for mft in (150, 250, 500):\n",
    "           \n",
    "            # Perform 30 iterations\n",
    "            for iteration in range(30):            \n",
    "\n",
    "                print('iteration #{0} | mft = {1} | lang = {2} | subc = {3}              '.format(\n",
    "                    iteration+1, mft, lang, subc\n",
    "                ), end='\\r')\n",
    "\n",
    "                # SAMPLING\n",
    "                s = sticho.Sampler(data[lang], n=100, subcorpus=subc, min_rhymes=40)\n",
    "                # Select 5 authors by random\n",
    "                s.level_authors(n=5)\n",
    "                # Randomly select 10 samples by each author\n",
    "                s.level_samples(max_=10)\n",
    "\n",
    "                # CLASSIFICATION WITH LEMMATA\n",
    "                f1 = sticho.Features(s)\n",
    "                f1.bow(domain='lemma', mft=mft)\n",
    "                f1.build_dataframe()\n",
    "                # Cross-validation\n",
    "                c = sticho.Classification(f1, clf='svm', kernel='linear', C=1)\n",
    "                acc[mft][lang+str(subc)]['L'].append(c.cross_validation())\n",
    "                for d in c.results['decisions']:\n",
    "                    decisions[mft][lang+str(subc)][iteration]['L'].append((d[0], c.results['decisions'][d]))                    \n",
    "                    \n",
    "                # CLASSIFICATION WITH VERSIFICATION\n",
    "                f2 = sticho.Features(s)    \n",
    "                f2.rhythmic_ngrams(n=2, mft=0)\n",
    "                f2.rhythmic_ngrams(n=3, mft=0)\n",
    "                f2.rhythmic_ngrams(n=4, mft=0)\n",
    "                f2.rhyme_profile(method='word_length', mft=0, ending='mf')\n",
    "                f2.rhyme_profile(method='stress', mft=0, ending='mf')\n",
    "                f2.rhyme_profile(method='pos', mft=0, ending='mf')\n",
    "                f2.rhyme_profile(method='sounds', mft=0, ending='mf', snd_position=1)\n",
    "                f2.rhyme_profile(method='sounds', mft=0, ending='mf', snd_position=2)\n",
    "                f2.rhyme_profile(method='sounds', mft=0, ending='f', snd_position=3)\n",
    "                f2.rhyme_profile(method='sounds', mft=0, ending='f', snd_position=4)\n",
    "                f2.sound_frequencies(mft=30)\n",
    "                f2.build_dataframe()    \n",
    "                # Cross-validation\n",
    "                c = sticho.Classification(f2, clf='svm', kernel='linear', C=1)\n",
    "                acc[mft][lang+str(subc)]['V'].append(c.cross_validation())\n",
    "                for d in c.results['decisions']:\n",
    "                    decisions[mft][lang+str(subc)][iteration]['V'].append((d[0], c.results['decisions'][d]))\n",
    "                \n",
    "                # CLASSIFICATION WITH VERSIFICATION + LEMMATA \n",
    "                df3 = pd.concat([f1.df_, f2.df_], axis=1, sort=False)\n",
    "                # Cross-validation\n",
    "                c = sticho.Classification(df3, clf='svm', kernel='linear', C=1)\n",
    "                acc[mft][lang+str(subc)]['LV'].append(c.cross_validation())                \n",
    "                for d in c.results['decisions']:\n",
    "                    decisions[mft][lang+str(subc)][iteration]['LV'].append((d[0], c.results['decisions'][d]))\n",
    "                                \n",
    "            # Wilcoxon signed rank test\n",
    "            w = stats.wilcoxon(acc[mft][lang+str(subc)]['L'], acc[mft][lang+str(subc)]['LV'])\n",
    "            print('\\n\\tp(Wilcoxon) = {0}'.format(w.pvalue))\n",
    "                    \n",
    "        # Plot the charts\n",
    "        fig, axs = plt.subplots(1, 3, figsize=(15,4))\n",
    "        plot_i = 0\n",
    "        for mft in sorted(acc):\n",
    "            axs[plot_i].boxplot([ \n",
    "                acc[mft][lang+str(subc)]['L'], \n",
    "                acc[mft][lang+str(subc)]['V'], \n",
    "                acc[mft][lang+str(subc)]['LV'] \n",
    "            ], whis=[0,100])\n",
    "            axs[plot_i].set_ylim(0.5,1.1)\n",
    "            axs[plot_i].set_xticklabels(['L', 'V', 'LV'])\n",
    "            axs[plot_i].set_title('{0}{1} ({2})'.format(lang, subc, mft))\n",
    "            axs[plot_i].set_yticks(axs[plot_i].get_yticks()[:-1])\n",
    "            axs[plot_i].set_xlabel('features')\n",
    "            axs[plot_i].set_ylabel('accuracy')\n",
    "            plot_i += 1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<h3>Voting</h3>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x288 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x288 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Container for plotting data\n",
    "plotdata = defaultdict(lambda: defaultdict(lambda: defaultdict(list)))\n",
    "\n",
    "# Iterate over predictions from particular iterations\n",
    "for mft in decisions:\n",
    "    for subcorpus in decisions[mft]:\n",
    "        for iteration in decisions[mft][subcorpus]:\n",
    "            ambiguous = 0\n",
    "            unambiguous_correct = 0\n",
    "            for i, x in enumerate(decisions[mft][subcorpus][iteration]['L']):\n",
    "                decision_l = decisions[mft][subcorpus][iteration]['L'][i]\n",
    "                decision_v = decisions[mft][subcorpus][iteration]['V'][i]\n",
    "                # Prediction by lemmata != prediction by versification\n",
    "                if decision_l[1] != decision_v[1]:\n",
    "                    ambiguous += 1\n",
    "                # Both predict the same author and the prediction is correct\n",
    "                elif decision_l[1] == decision_l[0]:\n",
    "                    unambiguous_correct += 1\n",
    "            # Store data to the main container\n",
    "            plotdata['unambiguous'][mft][subcorpus].append(unambiguous_correct/(len(decisions[mft][subcorpus][iteration]['L']) - ambiguous))\n",
    "            plotdata['ambiguous'][mft][subcorpus].append(ambiguous/len(decisions[mft][subcorpus][iteration]['L']))\n",
    "            \n",
    "# Iterate over plotting data (ambiguous/unambiguous)\n",
    "for t in sorted(plotdata):\n",
    "    # Plot the charts\n",
    "    fig, axs = plt.subplots(1, 3, figsize=(15,4))\n",
    "    plot_i = 0\n",
    "    for mft in sorted(plotdata[t]):\n",
    "        axs[plot_i].boxplot([ \n",
    "            plotdata[t][mft]['cs1'], \n",
    "            plotdata[t][mft]['cs2'], \n",
    "            plotdata[t][mft]['cs3'],\n",
    "            plotdata[t][mft]['es1'],\n",
    "            plotdata[t][mft]['es2']\n",
    "        ], whis=[0,100])\n",
    "        axs[plot_i].set_ylim(0,1.1)\n",
    "        axs[plot_i].set_xticklabels(['cs1', 'cs2', 'cs3', 'es1', 'es2'])\n",
    "        axs[plot_i].set_title('{0}({1})'.format(t, mft))\n",
    "        axs[plot_i].set_yticks(axs[plot_i].get_yticks()[:-1])\n",
    "        axs[plot_i].set_xlabel('corpus')\n",
    "        plot_i += 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "250 es2 L 1.8957631424603906e-05\n",
      "250 es2 V 1.6678313975036149e-06\n",
      "250 es2 LV 1.6897314186860455e-05\n",
      "250 es1 L 0.009454332030834373\n",
      "250 es1 V 2.3434959542201058e-06\n",
      "250 es1 LV 0.0511998362745388\n",
      "250 cs2 L 3.716291950908611e-06\n",
      "250 cs2 V 2.584007955542239e-06\n",
      "250 cs2 LV 1.2296882968809649e-05\n",
      "250 cs1 L 3.2931495080422715e-06\n",
      "250 cs1 V 2.3418941242887347e-06\n",
      "250 cs1 LV 0.0001320456763959241\n",
      "250 cs3 L 1.580051428024809e-06\n",
      "250 cs3 V 2.875572273617061e-06\n",
      "250 cs3 LV 4.124102095924912e-05\n",
      "500 es2 L 0.0004698669578488691\n",
      "500 es2 V 1.659399844033728e-06\n",
      "500 es2 LV 0.00030607170265197994\n",
      "500 es1 V 1.5312284726953685e-06\n",
      "500 es1 LV 0.31731050786291415\n",
      "500 cs2 L 7.338295563576314e-06\n",
      "500 cs2 V 2.2974108281097315e-06\n",
      "500 cs2 LV 0.00028352415776931704\n",
      "500 cs1 L 5.503938136582391e-05\n",
      "500 cs1 V 2.291113790811059e-06\n",
      "500 cs1 LV 0.04108722452782128\n",
      "500 cs3 L 1.6229293364132389e-06\n",
      "500 cs3 V 1.0771899135604078e-05\n",
      "500 cs3 LV 0.00025812880774004254\n",
      "150 es2 L 2.3161206293822223e-06\n",
      "150 es2 V 1.7213435893906393e-06\n",
      "150 es2 LV 1.7029908934686465e-06\n",
      "150 es1 L 0.00019283302182767056\n",
      "150 es1 V 2.357953433878093e-06\n",
      "150 es1 LV 0.0001496806311640834\n",
      "150 cs2 L 1.6890670049981201e-06\n",
      "150 cs2 V 5.374892921113508e-06\n",
      "150 cs2 LV 2.1168374326957036e-05\n",
      "150 cs1 L 2.357953433878093e-06\n",
      "150 cs1 V 7.638952448849794e-06\n",
      "150 cs1 LV 0.00015101762028621782\n",
      "150 cs3 L 1.678421145178411e-06\n",
      "150 cs3 V 2.1014359560124373e-06\n",
      "150 cs3 LV 3.254780192663229e-05\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/scipy/stats/morestats.py:2879: UserWarning: Sample size too small for normal approximation.\n",
      "  warnings.warn(\"Sample size too small for normal approximation.\")\n"
     ]
    }
   ],
   "source": [
    "# Print p-values\n",
    "for mft in plotdata['unambiguous']:\n",
    "    for subc in plotdata['unambiguous'][mft]:\n",
    "        if sorted(plotdata['unambiguous'][mft][subc]) != sorted(acc[mft][subc]['L']):            \n",
    "            w = stats.wilcoxon(plotdata['unambiguous'][mft][subc], acc[mft][subc]['L'])\n",
    "            print(mft, subc, 'L', w.pvalue)\n",
    "        if sorted(plotdata['unambiguous'][mft][subc]) != sorted(acc[mft][subc]['V']):            \n",
    "            w = stats.wilcoxon(plotdata['unambiguous'][mft][subc], acc[mft][subc]['V'])\n",
    "            print(mft, subc, 'V', w.pvalue)\n",
    "        if sorted(plotdata['unambiguous'][mft][subc]) != sorted(acc[mft][subc]['LV']):            \n",
    "            w = stats.wilcoxon(plotdata['unambiguous'][mft][subc], acc[mft][subc]['LV'])\n",
    "            print(mft, subc, 'LV', w.pvalue)  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}