stimmenfryslan/notebooks/Pronunciation map.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Geographical pronunciation statistics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas\n",
    "import MySQLdb\n",
    "import numpy\n",
    "import json\n",
    "\n",
    "db = MySQLdb.connect(user='root', passwd='Nmmxhjgt1@', db='stimmen', charset='utf8')\n",
    "\n",
    "%matplotlib notebook\n",
    "from matplotlib import pyplot\n",
    "import folium\n",
    "from IPython.display import display\n",
    "from shapely.geometry import Polygon, MultiPolygon, shape, Point\n",
    "from jupyter_progressbar import ProgressBar\n",
    "from collections import defaultdict\n",
    "from ipy_table import make_table\n",
    "from html import escape\n",
    "\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from matplotlib.colors import LogNorm\n",
    "from sklearn import mixture\n",
    "from skimage.measure import find_contours\n",
    "from collections import Counter\n",
    "from random import shuffle"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Borders of Frysian municipalities\n",
    "\n",
    "with open('Friesland_AL8.GeoJson') as f:\n",
    "    gemeentes = json.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "shapes = [shape(feature['geometry']) for feature in gemeentes['features']]\n",
    "gemeente_names = [feature['properties']['name'] for feature in gemeentes['features']]\n",
    "\n",
    "def get_gemeente(point):\n",
    "    for i, shape in enumerate(shapes):\n",
    "        if shape.contains(point):\n",
    "            return i\n",
    "    return -1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Answers to how participants state a word should be pronounces.\n",
    "\n",
    "answers = pandas.read_sql('''\n",
    "SELECT prediction_quiz_id, user_lat, user_lng, question_text, answer_text\n",
    "FROM       core_surveyresult as survey\n",
    "INNER JOIN core_predictionquizresult as result ON survey.id = result.survey_result_id\n",
    "INNER JOIN core_predictionquizresultquestionanswer as answer\n",
    "    ON result.id = answer.prediction_quiz_id\n",
    "''', db)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Takes approximately 2 minutes\n",
    "\n",
    "gemeente_map = {\n",
    "    (lng, lat): get_gemeente(Point(lng, lat))\n",
    "    for lng, lat in set(zip(answers['user_lng'], answers['user_lat']))\n",
    "}\n",
    "\n",
    "answers['gemeente'] = [\n",
    "    gemeente_map[(lng, lat)]\n",
    "    for lat, lng in zip(answers['user_lat'], answers['user_lng'])\n",
    "]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Mapping pronunciations\n",
    "\n",
    "The idea is to plot each pronunciation as a point of a different color, now only seems to show participation density.\n",
    "\n",
    "Slow, so started with the first question."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "# cmap = pyplot.get_cmap('gist_rainbow')\n",
    "\n",
    "# std = (1.89, 1.35)\n",
    "\n",
    "# for _, (question, rows) in zip(range(3), answers.groupby('question_text')):\n",
    "#     plt.figure()\n",
    "#     n_answers = len(rows.groupby('answer_text').count())\n",
    "#     colors = cmap(range(256))[::256 // n_answers]\n",
    "#     for (answer, rows_), color in zip(rows.groupby('answer_text'), colors):\n",
    "#         if len(rows_) < 100:\n",
    "#             continue\n",
    "#         color = '#%02x%02x%02x' % tuple(int(c*255) for c in color[:3])\n",
    "#         X = rows_[['user_lat', 'user_lng']].as_matrix()\n",
    "\n",
    "#         clf = mixture.GaussianMixture(n_components=5, covariance_type='full')\n",
    "#         clf.fit(X)\n",
    "#         xlim = numpy.percentile(X[:, 0], [1, 99.5])\n",
    "#         ylim = numpy.percentile(X[:, 1], [1, 99.5])\n",
    "#         xlim = [2*xlim[0] - xlim[1], 2*xlim[1] - xlim[0]]\n",
    "#         ylim = [2*ylim[0] - ylim[1], 2*ylim[1] - ylim[0]]\n",
    "        \n",
    "#         x = np.linspace(*xlim, 1000)\n",
    "#         y = np.linspace(*ylim, 1000)\n",
    "#         xx, yy = np.meshgrid(x, y)\n",
    "#         xxyy = np.array([xx.ravel(), yy.ravel()]).T\n",
    "#         z = np.exp(clf.score_samples(xxyy))\n",
    "#         z = z.reshape(xx.shape)\n",
    "        \n",
    "#         z_sorted = sorted(z.ravel(), reverse=True)\n",
    "#         z_sorted_cumsum = np.cumsum(z_sorted)\n",
    "#         split = np.where(z_sorted_cumsum > (z_sorted_cumsum[-1] * 0.5))[0][0]\n",
    "#         threshold = z_sorted[split]\n",
    "#         threshold\n",
    "\n",
    "# #         p = list(range(0, 100, 5))\n",
    "\n",
    "#         p = [80]\n",
    "#         plt.contour(xx, yy, z, levels=[threshold], colors=[color])\n",
    "#         plt.plot(X[:, 0], X[:, 1], '.', c=color)\n",
    "#         plt.xlim(*xlim)\n",
    "#         plt.ylim(*ylim)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "zero_latlng_questions = {\n",
    "    q\n",
    "    for q, row in answers.groupby('question_text').agg('std').iterrows()\n",
    "    if row['user_lat'] == 0 and row['user_lng'] == 0\n",
    "}\n",
    "answers_filtered = answers[answers['question_text'].map(lambda x: x not in zero_latlng_questions)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "answers_filtered['question_text_url'] = answers_filtered['question_text'].map(\n",
    "    lambda x: x.replace('\"', '').replace('*', ''))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_palette(n, no_black=True, no_white=True):\n",
    "    with open('glasbey/{}_colors.txt'.format(n + no_black + no_white)) as f:\n",
    "        return [\n",
    "            '#%02x%02x%02x' % tuple(int(c) for c in line.replace('\\n', '').split(','))\n",
    "            for line in f\n",
    "            if not no_black or line != '0,0,0\\n'\n",
    "            if not no_white or line != '255,255,255\\n'\n",
    "        ]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "options = [x[1] for x in sorted([\n",
    "    (row['user_lng'], answer_text)\n",
    "    for answer_text, row in rows.groupby('answer_text').agg({'user_lng': 'count'}).iterrows()\n",
    "], reverse=True)]\n",
    "\n",
    "groups = [options[:len(options) // 2], options[len(options) // 2:]]\n",
    "groups"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "80000 / 350"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import glob\n",
    "with open('index.html', 'w') as f:\n",
    "    f.write('<html><head></head><body>' + \n",
    "        '<br/>\\n'.join(\n",
    "            '\\t<a href=\"http://herbertkruitbosch.com/pronunciation_maps/{}\">{}<a>'.format(fn, fn[:-4].replace('_', ' '))\n",
    "            for fn in sorted(\n",
    "                glob.glob('*_all.html') +\n",
    "                glob.glob('*_larger.html') +\n",
    "                glob.glob('*_smaller.html')\n",
    "            )\n",
    "    ) + \"</body></html>\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "# cmap = pyplot.get_cmap('gist_rainbow')\n",
    "# colors = pyplot.get_cmap('tab20')\n",
    "# colors = ['#e6194b', '#3cb44b', '#ffe119', '#0082c8', '#f58231', '#911eb4', '#46f0f0', '#f032e6', '#d2f53c', '#fabebe', '#008080', '#e6beff', '#aa6e28', '#fffac8', '#800000', '#aaffc3', '#808000', '#ffd8b1', '#000080', '#808080']\n",
    "\n",
    "std = (1.89, 1.35)\n",
    "\n",
    "for question, rows in answers_filtered.groupby('question_text_url'):\n",
    "#     question = rows['question_text_url'][0]\n",
    "    n_answers = len(rows.groupby('answer_text').count())\n",
    "    \n",
    "    \n",
    "    options = [x[1] for x in sorted([\n",
    "        (row['user_lng'], answer_text)\n",
    "        for answer_text, row in rows.groupby('answer_text').agg({'user_lng': 'count'}).iterrows()\n",
    "    ], reverse=True)]\n",
    "    groups = [options]\n",
    "    if n_answers > 6:\n",
    "        groups.extend([options[:6], options[6:]])\n",
    "    \n",
    "    for group, group_name in zip(groups, ['all', 'larger', 'smaller']):\n",
    "        m = folium.Map((rows['user_lat'].median(), rows['user_lng'].median()), tiles='stamentoner', zoom_start=9)\n",
    "    #     colors = cmap(range(256))[::256 // n_answers]\n",
    "        colors = get_palette(len(group))\n",
    "        for answer, color in zip(group, colors):\n",
    "            rows_ = rows[rows['answer_text'] == answer]\n",
    "    #         color = '#%02x%02x%02x' % tuple(int(c*255) for c in color[:3])\n",
    "            name = '<span style=\\\\\"color:{}; \\\\\">{} ({})'.format(color, escape(answer), len(rows_))\n",
    "\n",
    "            group = folium.FeatureGroup(name=name)\n",
    "            colormap[name] = color\n",
    "\n",
    "            for point in zip(rows_['user_lat'], rows_['user_lng']):\n",
    "                point = tuple(p + 0.01 * s * numpy.random.randn() for p, s in zip(point, std))\n",
    "                folium.Circle(\n",
    "                    point, color=None, fill_color=color,\n",
    "                    radius=400*min(1, 100 / len(rows_)), fill_opacity=1 #1 - 0.5 * len(rows_) / len(rows)\n",
    "                ).add_to(group)\n",
    "            group.add_to(m)\n",
    "        folium.map.LayerControl('topright', collapsed=False).add_to(m)\n",
    "    \n",
    "        print(group_name, question)\n",
    "        if group_name == 'larger':\n",
    "            display(m)\n",
    "        m.save('{}_{}.html'.format(question, group_name))"
   ]
  }
 ],
 "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}