Topic_Modeling_Options.ipynb

{
 "metadata": {
  "name": "",
  "signature": "sha256:bdd461f4faa7434533d105c2ba24e7a6825a5ad1720122099f6e0b0df22e8871"
 },
 "nbformat": 3,
 "nbformat_minor": 0,
 "worksheets": [
  {
   "cells": [
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "### Topic Models in Python\n",
      "\n",
      "There are four main ways to estimate topic models in Python:\n",
      "\n",
      "| Name   | Methodology                                        | Installation                |\n",
      "|--------|----------------------------------------------------|-----------------------------|\n",
      "| [gensim](http://radimrehurek.com/gensim/) | LDA using  online variational Bayes (VB) algorithm | pip install gensim                         |\n",
      "| [lda](https://pypi.python.org/pypi/lda)    | LDA using collapsed Gibbs sampling                 | pip install lda                        |\n",
      "| [Mallet](http://mallet.cs.umass.edu)  | LDA suing collapsed Gibbs sampling                 | [Java download.](http://mallet.cs.umass.edu/download.php) <br>Requires JDK and  [gensim](http://radimrehurek.com/gensim/models/ldamallet.html) wrapper. |\n",
      "| [NMF](http://scikit-learn.org/stable/modules/generated/sklearn.decomposition.NMF.html)    | Non-Negative matrix factorization                  | part of [scikit-learn](http://scikit-learn.org/stable/)        |"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# Note that this will only work if you have stored the data in the current directory. \n",
      "# You can get the data from  \n",
      "# https://github.com/nealcaren/quant-text-fall-2014/tree/master/times_data\n",
      "# Remember that Pandas can read a file or a URL.\n",
      "\n",
      "import pandas as pd\n",
      "\n",
      "\n",
      "\n",
      "df = pd.read_json('times_abortion.json')\n",
      "\n",
      "print 'Article Count:', len(df)\n",
      "print 'Word Count:', df.word_count.sum()\n",
      "df.head()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "Article Count: 14269\n",
        "Word Count: 14626680\n"
       ]
      },
      {
       "html": [
        "<div style=\"max-height:1000px;max-width:1500px;overflow:auto;\">\n",
        "<table border=\"1\" class=\"dataframe\">\n",
        "  <thead>\n",
        "    <tr style=\"text-align: right;\">\n",
        "      <th></th>\n",
        "      <th>byline</th>\n",
        "      <th>dsk</th>\n",
        "      <th>print_page_number</th>\n",
        "      <th>print_section</th>\n",
        "      <th>publication_day_of_month</th>\n",
        "      <th>publication_month</th>\n",
        "      <th>publication_year</th>\n",
        "      <th>text</th>\n",
        "      <th>title</th>\n",
        "      <th>word_count</th>\n",
        "    </tr>\n",
        "  </thead>\n",
        "  <tbody>\n",
        "    <tr>\n",
        "      <th>1000008</th>\n",
        "      <td> Baker, Russell</td>\n",
        "      <td>    Editorial Desk</td>\n",
        "      <td> 23</td>\n",
        "      <td> A</td>\n",
        "      <td> 6</td>\n",
        "      <td> 3</td>\n",
        "      <td> 1998</td>\n",
        "      <td> Despite Monica Lewinsky and Stonewall Clinton,...</td>\n",
        "      <td>               Observer; Comfort for the Afflicted</td>\n",
        "      <td>  749</td>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <th>1000090</th>\n",
        "      <td>    Gross, Jane</td>\n",
        "      <td>     National Desk</td>\n",
        "      <td> 17</td>\n",
        "      <td> A</td>\n",
        "      <td> 6</td>\n",
        "      <td> 3</td>\n",
        "      <td> 1998</td>\n",
        "      <td> Contradicting Sgt. Maj. Gene C. McKinney, Sgt....</td>\n",
        "      <td>      Ex-Top Soldier Is Contradicted By An Accuser</td>\n",
        "      <td>  421</td>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <th>1000104</th>\n",
        "      <td>     Barry, Dan</td>\n",
        "      <td> Metropolitan Desk</td>\n",
        "      <td>  1</td>\n",
        "      <td> A</td>\n",
        "      <td> 6</td>\n",
        "      <td> 3</td>\n",
        "      <td> 1998</td>\n",
        "      <td> Just two months into his second term, Mayor Ru...</td>\n",
        "      <td> Giuliani Fund-Raising Points To Running for Hi...</td>\n",
        "      <td> 1318</td>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <th>1000208</th>\n",
        "      <td>  Schmitt, Eric</td>\n",
        "      <td>      Foreign Desk</td>\n",
        "      <td>  4</td>\n",
        "      <td> A</td>\n",
        "      <td> 7</td>\n",
        "      <td> 3</td>\n",
        "      <td> 1998</td>\n",
        "      <td> The Senate committee that controls spending bi...</td>\n",
        "      <td> Senators Back Extra Military Spending but Faul...</td>\n",
        "      <td>  820</td>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <th>1000225</th>\n",
        "      <td> Purdum, Todd S</td>\n",
        "      <td>     National Desk</td>\n",
        "      <td>  6</td>\n",
        "      <td> A</td>\n",
        "      <td> 7</td>\n",
        "      <td> 3</td>\n",
        "      <td> 1998</td>\n",
        "      <td> Californians are long used to the televised cu...</td>\n",
        "      <td>        Interest Groups Run Own Race in California</td>\n",
        "      <td> 1169</td>\n",
        "    </tr>\n",
        "  </tbody>\n",
        "</table>\n",
        "</div>"
       ],
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 28,
       "text": [
        "                 byline                dsk  print_page_number print_section  \\\n",
        "1000008  Baker, Russell     Editorial Desk                 23             A   \n",
        "1000090     Gross, Jane      National Desk                 17             A   \n",
        "1000104      Barry, Dan  Metropolitan Desk                  1             A   \n",
        "1000208   Schmitt, Eric       Foreign Desk                  4             A   \n",
        "1000225  Purdum, Todd S      National Desk                  6             A   \n",
        "\n",
        "         publication_day_of_month  publication_month  publication_year  \\\n",
        "1000008                         6                  3              1998   \n",
        "1000090                         6                  3              1998   \n",
        "1000104                         6                  3              1998   \n",
        "1000208                         7                  3              1998   \n",
        "1000225                         7                  3              1998   \n",
        "\n",
        "                                                      text  \\\n",
        "1000008  Despite Monica Lewinsky and Stonewall Clinton,...   \n",
        "1000090  Contradicting Sgt. Maj. Gene C. McKinney, Sgt....   \n",
        "1000104  Just two months into his second term, Mayor Ru...   \n",
        "1000208  The Senate committee that controls spending bi...   \n",
        "1000225  Californians are long used to the televised cu...   \n",
        "\n",
        "                                                     title  word_count  \n",
        "1000008                Observer; Comfort for the Afflicted         749  \n",
        "1000090       Ex-Top Soldier Is Contradicted By An Accuser         421  \n",
        "1000104  Giuliani Fund-Raising Points To Running for Hi...        1318  \n",
        "1000208  Senators Back Extra Military Spending but Faul...         820  \n",
        "1000225         Interest Groups Run Own Race in California        1169  "
       ]
      }
     ],
     "prompt_number": 28
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# Trim the dataset to just look at a couple of newsdesks\n",
      "\n",
      "def good_desk(dsk):\n",
      "    if dsk in ['National Desk', 'Editorial Desk',  u'Metropolitan Desk', 'Foreign Desk']:\n",
      "        return True\n",
      "    return False\n",
      "\n",
      "df['viable'] = df['dsk'].apply(good_desk)\n",
      "\n",
      "df = df.loc[df['viable'] == True]\n"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 29
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "df.groupby('dsk').count()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "html": [
        "<div style=\"max-height:1000px;max-width:1500px;overflow:auto;\">\n",
        "<table border=\"1\" class=\"dataframe\">\n",
        "  <thead>\n",
        "    <tr style=\"text-align: right;\">\n",
        "      <th></th>\n",
        "      <th>byline</th>\n",
        "      <th>print_page_number</th>\n",
        "      <th>print_section</th>\n",
        "      <th>publication_day_of_month</th>\n",
        "      <th>publication_month</th>\n",
        "      <th>publication_year</th>\n",
        "      <th>text</th>\n",
        "      <th>title</th>\n",
        "      <th>word_count</th>\n",
        "      <th>viable</th>\n",
        "      <th>...</th>\n",
        "      <th>campaign ms giuliani</th>\n",
        "      <th>house senate president</th>\n",
        "      <th>dole campaign kemp</th>\n",
        "      <th>clinton president mrs</th>\n",
        "      <th>party republican republicans</th>\n",
        "      <th>life right state</th>\n",
        "      <th>judge bork court</th>\n",
        "      <th>dr doctors health</th>\n",
        "      <th>children year health</th>\n",
        "      <th>percent voters poll</th>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <th>dsk</th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "    </tr>\n",
        "  </thead>\n",
        "  <tbody>\n",
        "    <tr>\n",
        "      <th>Editorial Desk</th>\n",
        "      <td> 2446</td>\n",
        "      <td> 2446</td>\n",
        "      <td> 2446</td>\n",
        "      <td> 2446</td>\n",
        "      <td> 2446</td>\n",
        "      <td> 2446</td>\n",
        "      <td> 2446</td>\n",
        "      <td> 2446</td>\n",
        "      <td> 2446</td>\n",
        "      <td> 2446</td>\n",
        "      <td>...</td>\n",
        "      <td> 2446</td>\n",
        "      <td> 2446</td>\n",
        "      <td> 2446</td>\n",
        "      <td> 2446</td>\n",
        "      <td> 2446</td>\n",
        "      <td> 2446</td>\n",
        "      <td> 2446</td>\n",
        "      <td> 2446</td>\n",
        "      <td> 2446</td>\n",
        "      <td> 2446</td>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <th>Foreign Desk</th>\n",
        "      <td>  762</td>\n",
        "      <td>  761</td>\n",
        "      <td>  762</td>\n",
        "      <td>  762</td>\n",
        "      <td>  762</td>\n",
        "      <td>  762</td>\n",
        "      <td>  762</td>\n",
        "      <td>  762</td>\n",
        "      <td>  762</td>\n",
        "      <td>  762</td>\n",
        "      <td>...</td>\n",
        "      <td>  762</td>\n",
        "      <td>  762</td>\n",
        "      <td>  762</td>\n",
        "      <td>  762</td>\n",
        "      <td>  762</td>\n",
        "      <td>  762</td>\n",
        "      <td>  762</td>\n",
        "      <td>  762</td>\n",
        "      <td>  762</td>\n",
        "      <td>  762</td>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <th>Metropolitan Desk</th>\n",
        "      <td> 2109</td>\n",
        "      <td> 2108</td>\n",
        "      <td> 2109</td>\n",
        "      <td> 2109</td>\n",
        "      <td> 2109</td>\n",
        "      <td> 2109</td>\n",
        "      <td> 2109</td>\n",
        "      <td> 2109</td>\n",
        "      <td> 2109</td>\n",
        "      <td> 2109</td>\n",
        "      <td>...</td>\n",
        "      <td> 2109</td>\n",
        "      <td> 2109</td>\n",
        "      <td> 2109</td>\n",
        "      <td> 2109</td>\n",
        "      <td> 2109</td>\n",
        "      <td> 2109</td>\n",
        "      <td> 2109</td>\n",
        "      <td> 2109</td>\n",
        "      <td> 2109</td>\n",
        "      <td> 2109</td>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <th>National Desk</th>\n",
        "      <td> 5406</td>\n",
        "      <td> 5405</td>\n",
        "      <td> 5406</td>\n",
        "      <td> 5406</td>\n",
        "      <td> 5406</td>\n",
        "      <td> 5406</td>\n",
        "      <td> 5406</td>\n",
        "      <td> 5406</td>\n",
        "      <td> 5406</td>\n",
        "      <td> 5406</td>\n",
        "      <td>...</td>\n",
        "      <td> 5406</td>\n",
        "      <td> 5406</td>\n",
        "      <td> 5406</td>\n",
        "      <td> 5406</td>\n",
        "      <td> 5406</td>\n",
        "      <td> 5406</td>\n",
        "      <td> 5406</td>\n",
        "      <td> 5406</td>\n",
        "      <td> 5406</td>\n",
        "      <td> 5406</td>\n",
        "    </tr>\n",
        "  </tbody>\n",
        "</table>\n",
        "<p>4 rows \u00d7 26 columns</p>\n",
        "</div>"
       ],
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 66,
       "text": [
        "                   byline  print_page_number  print_section  \\\n",
        "dsk                                                           \n",
        "Editorial Desk       2446               2446           2446   \n",
        "Foreign Desk          762                761            762   \n",
        "Metropolitan Desk    2109               2108           2109   \n",
        "National Desk        5406               5405           5406   \n",
        "\n",
        "                   publication_day_of_month  publication_month  \\\n",
        "dsk                                                              \n",
        "Editorial Desk                         2446               2446   \n",
        "Foreign Desk                            762                762   \n",
        "Metropolitan Desk                      2109               2109   \n",
        "National Desk                          5406               5406   \n",
        "\n",
        "                   publication_year  text  title  word_count  viable   ...    \\\n",
        "dsk                                                                    ...     \n",
        "Editorial Desk                 2446  2446   2446        2446    2446   ...     \n",
        "Foreign Desk                    762   762    762         762     762   ...     \n",
        "Metropolitan Desk              2109  2109   2109        2109    2109   ...     \n",
        "National Desk                  5406  5406   5406        5406    5406   ...     \n",
        "\n",
        "                   campaign ms giuliani  house senate president  \\\n",
        "dsk                                                               \n",
        "Editorial Desk                     2446                    2446   \n",
        "Foreign Desk                        762                     762   \n",
        "Metropolitan Desk                  2109                    2109   \n",
        "National Desk                      5406                    5406   \n",
        "\n",
        "                   dole campaign kemp  clinton president mrs  \\\n",
        "dsk                                                            \n",
        "Editorial Desk                   2446                   2446   \n",
        "Foreign Desk                      762                    762   \n",
        "Metropolitan Desk                2109                   2109   \n",
        "National Desk                    5406                   5406   \n",
        "\n",
        "                   party republican republicans  life right state  \\\n",
        "dsk                                                                 \n",
        "Editorial Desk                             2446              2446   \n",
        "Foreign Desk                                762               762   \n",
        "Metropolitan Desk                          2109              2109   \n",
        "National Desk                              5406              5406   \n",
        "\n",
        "                   judge bork court  dr doctors health  children year health  \\\n",
        "dsk                                                                            \n",
        "Editorial Desk                 2446               2446                  2446   \n",
        "Foreign Desk                    762                762                   762   \n",
        "Metropolitan Desk              2109               2109                  2109   \n",
        "National Desk                  5406               5406                  5406   \n",
        "\n",
        "                   percent voters poll  \n",
        "dsk                                     \n",
        "Editorial Desk                    2446  \n",
        "Foreign Desk                       762  \n",
        "Metropolitan Desk                 2109  \n",
        "National Desk                     5406  \n",
        "\n",
        "[4 rows x 26 columns]"
       ]
      }
     ],
     "prompt_number": 66
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "print len(df)\n",
      "df.head()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 30,
       "text": [
        "10723"
       ]
      }
     ],
     "prompt_number": 30
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "from datetime import datetime\n",
      "\n",
      "from sklearn.feature_extraction.text import TfidfVectorizer, CountVectorizer\n"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 31
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "tf_vectorizer = CountVectorizer(max_df=0.7, \n",
      "                             min_df=.01,\n",
      "                             stop_words='english')\n",
      "\n",
      "\n",
      "tf_vectorizer.fit(df['text'])\n",
      "corpus_tf = tf_vectorizer.transform(df['text'])"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 33
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "categories = list(set(df.dsk.unique()))\n",
      "\n",
      "number_of_topics = len(categories)  * 4\n",
      "\n",
      "print 'Planning to extract %s topics' % number_of_topics"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "Planning to extract 16 topics\n"
       ]
      }
     ],
     "prompt_number": 63
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "from sklearn.decomposition import NMF\n",
      "\n",
      "\n",
      "start_time = datetime.now()\n",
      "\n",
      "nmf = NMF(n_components = number_of_topics, \n",
      "          max_iter = 5000).fit(corpus_tf)\n",
      "\n",
      "elapsed_time = (datetime.now() - start_time).seconds\n",
      "print elapsed_time, 'seconds'"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "28 seconds\n"
       ]
      }
     ],
     "prompt_number": 35
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "feature_names = tf_vectorizer.get_feature_names()\n",
      "topic_labels = []\n",
      "\n",
      "for topic_idx, topic in enumerate(nmf.components_):\n",
      "    print \"Topic %s: %s\" % (topic_idx,  ' '.join([feature_names[i] for i in topic.argsort()[:-10 - 1:-1]]))\n",
      "    topic_labels.append(\" \".join([feature_names[i]\n",
      "                    for i in topic.argsort()[:-3 - 1:-1]]))\n"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "Topic 0: state governor republican democrat senator democratic race senate won reporting\n",
        "Topic 1: court justice law case supreme federal decision justices opinion state\n",
        "Topic 2: women ms men woman rights family abortions year children say\n",
        "Topic 3: president think people going ve don just know say want\n",
        "Topic 4: bush president campaign gore mccain george vice texas presidential reagan\n",
        "Topic 5: church catholic pope cardinal bishops bishop catholics vatican john religious\n",
        "Topic 6: campaign ms giuliani york city mayor amato political dinkins candidate\n",
        "Topic 7: house senate president republican republicans democrats congress white senator committee\n",
        "Topic 8: dole campaign kemp bob senator buchanan republican presidential forbes tax\n",
        "Topic 9: clinton president mrs gore lazio health perot white campaign democratic\n",
        "Topic 10: party republican republicans conservative democrats political national buchanan like democratic\n",
        "Topic 11: life right state abortions rights law woman roe issue court\n",
        "Topic 12: judge bork court senator thomas supreme nomination committee senate souter\n",
        "Topic 13: dr doctors health medical abortions clinic doctor patients hospital drug\n",
        "Topic 14: children year health government family care people years federal school\n",
        "Topic 15: percent voters poll vote survey points times percentage error results\n"
       ]
      }
     ],
     "prompt_number": 36
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "y_hat = nmf.transform(corpus_tf)\n",
      "y_norm = [y/ y.sum() for y in y_hat] #Normalize\n",
      "\n",
      "df_y_hat = pd.DataFrame(y_norm, \n",
      "                        columns=topic_labels,  # Use our informative topic names\n",
      "                        index=df.index.values) # Makes merge easier\n",
      "\n",
      "df_y_hat = df_y_hat.fillna(0)\n",
      "df_topics = df.join(df_y_hat) # This is a new data set, so you can remerge"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 37
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "df_topics[['dsk'] + topic_labels].groupby('dsk').mean()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "html": [
        "<div style=\"max-height:1000px;max-width:1500px;overflow:auto;\">\n",
        "<table border=\"1\" class=\"dataframe\">\n",
        "  <thead>\n",
        "    <tr style=\"text-align: right;\">\n",
        "      <th></th>\n",
        "      <th>state governor republican</th>\n",
        "      <th>court justice law</th>\n",
        "      <th>women ms men</th>\n",
        "      <th>president think people</th>\n",
        "      <th>bush president campaign</th>\n",
        "      <th>church catholic pope</th>\n",
        "      <th>campaign ms giuliani</th>\n",
        "      <th>house senate president</th>\n",
        "      <th>dole campaign kemp</th>\n",
        "      <th>clinton president mrs</th>\n",
        "      <th>party republican republicans</th>\n",
        "      <th>life right state</th>\n",
        "      <th>judge bork court</th>\n",
        "      <th>dr doctors health</th>\n",
        "      <th>children year health</th>\n",
        "      <th>percent voters poll</th>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <th>dsk</th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "    </tr>\n",
        "  </thead>\n",
        "  <tbody>\n",
        "    <tr>\n",
        "      <th>Editorial Desk</th>\n",
        "      <td> 0.024377</td>\n",
        "      <td> 0.066798</td>\n",
        "      <td> 0.123788</td>\n",
        "      <td> 0.072586</td>\n",
        "      <td> 0.070085</td>\n",
        "      <td> 0.068850</td>\n",
        "      <td> 0.044687</td>\n",
        "      <td> 0.048927</td>\n",
        "      <td> 0.025976</td>\n",
        "      <td> 0.040731</td>\n",
        "      <td> 0.061170</td>\n",
        "      <td> 0.155419</td>\n",
        "      <td> 0.033249</td>\n",
        "      <td> 0.041425</td>\n",
        "      <td> 0.100126</td>\n",
        "      <td> 0.021807</td>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <th>Foreign Desk</th>\n",
        "      <td> 0.012759</td>\n",
        "      <td> 0.030506</td>\n",
        "      <td> 0.142376</td>\n",
        "      <td> 0.058688</td>\n",
        "      <td> 0.020524</td>\n",
        "      <td> 0.195861</td>\n",
        "      <td> 0.028674</td>\n",
        "      <td> 0.066215</td>\n",
        "      <td> 0.005220</td>\n",
        "      <td> 0.032910</td>\n",
        "      <td> 0.058852</td>\n",
        "      <td> 0.074123</td>\n",
        "      <td> 0.010906</td>\n",
        "      <td> 0.041660</td>\n",
        "      <td> 0.174512</td>\n",
        "      <td> 0.046213</td>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <th>Metropolitan Desk</th>\n",
        "      <td> 0.106768</td>\n",
        "      <td> 0.047823</td>\n",
        "      <td> 0.063029</td>\n",
        "      <td> 0.055428</td>\n",
        "      <td> 0.014902</td>\n",
        "      <td> 0.085553</td>\n",
        "      <td> 0.182574</td>\n",
        "      <td> 0.042870</td>\n",
        "      <td> 0.009557</td>\n",
        "      <td> 0.035760</td>\n",
        "      <td> 0.081217</td>\n",
        "      <td> 0.061358</td>\n",
        "      <td> 0.024756</td>\n",
        "      <td> 0.067517</td>\n",
        "      <td> 0.091631</td>\n",
        "      <td> 0.029256</td>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <th>National Desk</th>\n",
        "      <td> 0.056065</td>\n",
        "      <td> 0.078374</td>\n",
        "      <td> 0.062148</td>\n",
        "      <td> 0.062008</td>\n",
        "      <td> 0.075390</td>\n",
        "      <td> 0.050139</td>\n",
        "      <td> 0.055365</td>\n",
        "      <td> 0.100221</td>\n",
        "      <td> 0.043891</td>\n",
        "      <td> 0.046326</td>\n",
        "      <td> 0.081671</td>\n",
        "      <td> 0.070851</td>\n",
        "      <td> 0.050960</td>\n",
        "      <td> 0.054262</td>\n",
        "      <td> 0.074211</td>\n",
        "      <td> 0.038119</td>\n",
        "    </tr>\n",
        "  </tbody>\n",
        "</table>\n",
        "</div>"
       ],
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 38,
       "text": [
        "                   state governor republican  court justice law  women ms men  \\\n",
        "dsk                                                                             \n",
        "Editorial Desk                      0.024377           0.066798      0.123788   \n",
        "Foreign Desk                        0.012759           0.030506      0.142376   \n",
        "Metropolitan Desk                   0.106768           0.047823      0.063029   \n",
        "National Desk                       0.056065           0.078374      0.062148   \n",
        "\n",
        "                   president think people  bush president campaign  \\\n",
        "dsk                                                                  \n",
        "Editorial Desk                   0.072586                 0.070085   \n",
        "Foreign Desk                     0.058688                 0.020524   \n",
        "Metropolitan Desk                0.055428                 0.014902   \n",
        "National Desk                    0.062008                 0.075390   \n",
        "\n",
        "                   church catholic pope  campaign ms giuliani  \\\n",
        "dsk                                                             \n",
        "Editorial Desk                 0.068850              0.044687   \n",
        "Foreign Desk                   0.195861              0.028674   \n",
        "Metropolitan Desk              0.085553              0.182574   \n",
        "National Desk                  0.050139              0.055365   \n",
        "\n",
        "                   house senate president  dole campaign kemp  \\\n",
        "dsk                                                             \n",
        "Editorial Desk                   0.048927            0.025976   \n",
        "Foreign Desk                     0.066215            0.005220   \n",
        "Metropolitan Desk                0.042870            0.009557   \n",
        "National Desk                    0.100221            0.043891   \n",
        "\n",
        "                   clinton president mrs  party republican republicans  \\\n",
        "dsk                                                                      \n",
        "Editorial Desk                  0.040731                      0.061170   \n",
        "Foreign Desk                    0.032910                      0.058852   \n",
        "Metropolitan Desk               0.035760                      0.081217   \n",
        "National Desk                   0.046326                      0.081671   \n",
        "\n",
        "                   life right state  judge bork court  dr doctors health  \\\n",
        "dsk                                                                        \n",
        "Editorial Desk             0.155419          0.033249           0.041425   \n",
        "Foreign Desk               0.074123          0.010906           0.041660   \n",
        "Metropolitan Desk          0.061358          0.024756           0.067517   \n",
        "National Desk              0.070851          0.050960           0.054262   \n",
        "\n",
        "                   children year health  percent voters poll  \n",
        "dsk                                                           \n",
        "Editorial Desk                 0.100126             0.021807  \n",
        "Foreign Desk                   0.174512             0.046213  \n",
        "Metropolitan Desk              0.091631             0.029256  \n",
        "National Desk                  0.074211             0.038119  "
       ]
      }
     ],
     "prompt_number": 38
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "model_stats = {}\n",
      "model_stats['NMF'] = {'time' : elapsed_time ,\n",
      "                      'labels' : topic_labels ,\n",
      "                      'dataframe' : df_topics }"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 39
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "from gensim import corpora, models, similarities, matutils "
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 40
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "id2word = dict((v, k) for k, v in tf_vectorizer.vocabulary_.iteritems())\n",
      "corpus_vect_gensim = matutils.Sparse2Corpus(corpus_tf, documents_columns=False)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 41
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "start_time = datetime.now()\n",
      "\n",
      "lda = models.LdaModel(corpus_vect_gensim, \n",
      "                      id2word = id2word, \n",
      "                      num_topics = number_of_topics,\n",
      "                      passes=10)\n",
      "elapsed_time = (datetime.now() - start_time).seconds\n",
      "print elapsed_time, 'seconds'"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "325 seconds\n"
       ]
      }
     ],
     "prompt_number": 42
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "topic_labels = []\n",
      "\n",
      "\n",
      "for i in range(0, number_of_topics):\n",
      "    temp = lda.show_topic(i, 10)\n",
      "    terms = []\n",
      "    for term in temp:\n",
      "        terms.append(term[1])\n",
      "    print  \"topic %s: %s\" % (i,\" \".join(terms) )\n",
      "    topic_labels.append(\" \".join(terms[:3]))\n"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "topic 0: church catholic pope religious cardinal roman bishops vatican john bishop\n",
        "topic 1: house senate republican republicans congress vote clinton committee president today\n",
        "topic 2: people ms like year old children years family mrs school\n",
        "topic 3: republican state democratic party democrats republicans district governor race election\n",
        "topic 4: president house white clinton political reagan general office washington administration\n",
        "topic 5: clinic dr police anti clinics people federal ms today officials\n",
        "topic 6: women dr abortions woman life medical health birth doctors pregnancy\n",
        "topic 7: york mrs giuliani state city whitman governor party mayor pataki\n",
        "topic 8: family health planning services united women care states administration population\n",
        "topic 9: state tax health year government budget education plan federal school\n",
        "topic 10: percent poll voters black vote survey 10 people percentage points\n",
        "topic 11: clinton president people think american like government going don say\n",
        "topic 12: campaign candidates television money political million candidate 000 debate groups\n",
        "topic 13: bush dole party republican campaign republicans president presidential convention buchanan\n",
        "topic 14: judge court senator nomination bork senate committee supreme kennedy thomas\n",
        "topic 15: court law justice supreme state case decision right federal rights\n"
       ]
      }
     ],
     "prompt_number": 43
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "y_hat = []\n",
      "for item in corpus_vect_gensim:\n",
      "    y_dict = dict(lda[item])\n",
      "    y_list = [y_dict.get(i,0) for i in range(0,number_of_topics)]\n",
      "    y_hat.append(y_list)\n",
      "\n",
      "df_y_hat = pd.DataFrame(y_hat, \n",
      "                        columns=topic_labels,  # Use our informative topic names\n",
      "                        index=df.index.values) # Makes merge easier\n",
      "\n",
      "df_topics = df.join(df_y_hat) # This is a new data set, so you can remerge"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 44
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "df_topics[['dsk'] + topic_labels].groupby('dsk').mean()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "html": [
        "<div style=\"max-height:1000px;max-width:1500px;overflow:auto;\">\n",
        "<table border=\"1\" class=\"dataframe\">\n",
        "  <thead>\n",
        "    <tr style=\"text-align: right;\">\n",
        "      <th></th>\n",
        "      <th>church catholic pope</th>\n",
        "      <th>house senate republican</th>\n",
        "      <th>people ms like</th>\n",
        "      <th>republican state democratic</th>\n",
        "      <th>president house white</th>\n",
        "      <th>clinic dr police</th>\n",
        "      <th>women dr abortions</th>\n",
        "      <th>york mrs giuliani</th>\n",
        "      <th>family health planning</th>\n",
        "      <th>state tax health</th>\n",
        "      <th>percent poll voters</th>\n",
        "      <th>clinton president people</th>\n",
        "      <th>campaign candidates television</th>\n",
        "      <th>bush dole party</th>\n",
        "      <th>judge court senator</th>\n",
        "      <th>court law justice</th>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <th>dsk</th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "    </tr>\n",
        "  </thead>\n",
        "  <tbody>\n",
        "    <tr>\n",
        "      <th>Editorial Desk</th>\n",
        "      <td> 0.073061</td>\n",
        "      <td> 0.034943</td>\n",
        "      <td> 0.070953</td>\n",
        "      <td> 0.025428</td>\n",
        "      <td> 0.037426</td>\n",
        "      <td> 0.034887</td>\n",
        "      <td> 0.221881</td>\n",
        "      <td> 0.023777</td>\n",
        "      <td> 0.056825</td>\n",
        "      <td> 0.033089</td>\n",
        "      <td> 0.013237</td>\n",
        "      <td> 0.138552</td>\n",
        "      <td> 0.023223</td>\n",
        "      <td> 0.078754</td>\n",
        "      <td> 0.036353</td>\n",
        "      <td> 0.089988</td>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <th>Foreign Desk</th>\n",
        "      <td> 0.288316</td>\n",
        "      <td> 0.076504</td>\n",
        "      <td> 0.124704</td>\n",
        "      <td> 0.025453</td>\n",
        "      <td> 0.031320</td>\n",
        "      <td> 0.037744</td>\n",
        "      <td> 0.114384</td>\n",
        "      <td> 0.004876</td>\n",
        "      <td> 0.089751</td>\n",
        "      <td> 0.031175</td>\n",
        "      <td> 0.034584</td>\n",
        "      <td> 0.084523</td>\n",
        "      <td> 0.014810</td>\n",
        "      <td> 0.008589</td>\n",
        "      <td> 0.001980</td>\n",
        "      <td> 0.027179</td>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <th>Metropolitan Desk</th>\n",
        "      <td> 0.073453</td>\n",
        "      <td> 0.029833</td>\n",
        "      <td> 0.110339</td>\n",
        "      <td> 0.101536</td>\n",
        "      <td> 0.037792</td>\n",
        "      <td> 0.123366</td>\n",
        "      <td> 0.042665</td>\n",
        "      <td> 0.188497</td>\n",
        "      <td> 0.030061</td>\n",
        "      <td> 0.065746</td>\n",
        "      <td> 0.021530</td>\n",
        "      <td> 0.036501</td>\n",
        "      <td> 0.062893</td>\n",
        "      <td> 0.020969</td>\n",
        "      <td> 0.008553</td>\n",
        "      <td> 0.041595</td>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <th>National Desk</th>\n",
        "      <td> 0.050247</td>\n",
        "      <td> 0.082805</td>\n",
        "      <td> 0.071415</td>\n",
        "      <td> 0.097086</td>\n",
        "      <td> 0.070709</td>\n",
        "      <td> 0.063910</td>\n",
        "      <td> 0.065029</td>\n",
        "      <td> 0.011789</td>\n",
        "      <td> 0.034246</td>\n",
        "      <td> 0.044175</td>\n",
        "      <td> 0.029137</td>\n",
        "      <td> 0.067538</td>\n",
        "      <td> 0.046692</td>\n",
        "      <td> 0.130682</td>\n",
        "      <td> 0.044354</td>\n",
        "      <td> 0.085937</td>\n",
        "    </tr>\n",
        "  </tbody>\n",
        "</table>\n",
        "</div>"
       ],
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 45,
       "text": [
        "                   church catholic pope  house senate republican  \\\n",
        "dsk                                                                \n",
        "Editorial Desk                 0.073061                 0.034943   \n",
        "Foreign Desk                   0.288316                 0.076504   \n",
        "Metropolitan Desk              0.073453                 0.029833   \n",
        "National Desk                  0.050247                 0.082805   \n",
        "\n",
        "                   people ms like  republican state democratic  \\\n",
        "dsk                                                              \n",
        "Editorial Desk           0.070953                     0.025428   \n",
        "Foreign Desk             0.124704                     0.025453   \n",
        "Metropolitan Desk        0.110339                     0.101536   \n",
        "National Desk            0.071415                     0.097086   \n",
        "\n",
        "                   president house white  clinic dr police  \\\n",
        "dsk                                                          \n",
        "Editorial Desk                  0.037426          0.034887   \n",
        "Foreign Desk                    0.031320          0.037744   \n",
        "Metropolitan Desk               0.037792          0.123366   \n",
        "National Desk                   0.070709          0.063910   \n",
        "\n",
        "                   women dr abortions  york mrs giuliani  \\\n",
        "dsk                                                        \n",
        "Editorial Desk               0.221881           0.023777   \n",
        "Foreign Desk                 0.114384           0.004876   \n",
        "Metropolitan Desk            0.042665           0.188497   \n",
        "National Desk                0.065029           0.011789   \n",
        "\n",
        "                   family health planning  state tax health  \\\n",
        "dsk                                                           \n",
        "Editorial Desk                   0.056825          0.033089   \n",
        "Foreign Desk                     0.089751          0.031175   \n",
        "Metropolitan Desk                0.030061          0.065746   \n",
        "National Desk                    0.034246          0.044175   \n",
        "\n",
        "                   percent poll voters  clinton president people  \\\n",
        "dsk                                                                \n",
        "Editorial Desk                0.013237                  0.138552   \n",
        "Foreign Desk                  0.034584                  0.084523   \n",
        "Metropolitan Desk             0.021530                  0.036501   \n",
        "National Desk                 0.029137                  0.067538   \n",
        "\n",
        "                   campaign candidates television  bush dole party  \\\n",
        "dsk                                                                  \n",
        "Editorial Desk                           0.023223         0.078754   \n",
        "Foreign Desk                             0.014810         0.008589   \n",
        "Metropolitan Desk                        0.062893         0.020969   \n",
        "National Desk                            0.046692         0.130682   \n",
        "\n",
        "                   judge court senator  court law justice  \n",
        "dsk                                                        \n",
        "Editorial Desk                0.036353           0.089988  \n",
        "Foreign Desk                  0.001980           0.027179  \n",
        "Metropolitan Desk             0.008553           0.041595  \n",
        "National Desk                 0.044354           0.085937  "
       ]
      }
     ],
     "prompt_number": 45
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "model_stats['Gensim'] = {'time' : elapsed_time ,\n",
      "                         'labels' : topic_labels ,\n",
      "                         'dataframe' : df_topics }"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 46
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "import lda\n",
      "import numpy as np\n"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 47
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "vocab = [w for w in sorted(tf_vectorizer.vocabulary_)]\n",
      "\n",
      "\n",
      "model = lda.LDA(n_topics= number_of_topics,\n",
      "                n_iter=500)\n",
      "\n"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 48
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "start_time = datetime.now()\n",
      "\n",
      "model.fit(corpus_tf.toarray())\n",
      "\n",
      "elapsed_time = (datetime.now() - start_time).seconds\n",
      "print elapsed_time, 'seconds'"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "445 seconds\n"
       ]
      }
     ],
     "prompt_number": 49
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "topic_word = model.topic_word_\n",
      "topic_labels = []\n",
      "\n",
      "n_top_words = 10\n",
      "for i, topic_dist in enumerate(topic_word):\n",
      "    \n",
      "    topic_words = np.array(vocab)[np.argsort(topic_dist)][:-n_top_words:-1]\n",
      "    print('Topic {}: {}'.format(i, ' '.join(topic_words)))\n",
      "    topic_labels.append(' '.join(topic_words[:3]))"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "Topic 0: people like president time years political say did just\n",
        "Topic 1: women ms children family mrs year woman child old\n",
        "Topic 2: republican state party campaign democratic voters republicans democrats governor\n",
        "Topic 3: house senate president senator committee republican bush clinton republicans\n",
        "Topic 4: abortions state women law life right rights court woman\n",
        "Topic 5: rights law religious civil school public government right american\n",
        "Topic 6: york city giuliani mayor pataki cuomo campaign dinkins yesterday\n",
        "Topic 7: bush dole gore campaign republican party president presidential buchanan\n",
        "Topic 8: health tax care federal government money budget million programs\n",
        "Topic 9: dr drug health medical research doctors patients abortions aids\n",
        "Topic 10: united world states nations china population american government international\n",
        "Topic 11: court judge justice supreme case law federal justices decision\n",
        "Topic 12: federal police case law death attorney officials department charges\n",
        "Topic 13: percent poll survey points percentage error news conducted times\n",
        "Topic 14: church catholic pope cardinal bishops bishop roman catholics john\n",
        "Topic 15: clinic anti clinics group protesters rights operation rescue planned\n"
       ]
      }
     ],
     "prompt_number": 50
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "df_y_hat = pd.DataFrame(model.doc_topic_, \n",
      "                        columns=topic_labels,  # Use our informative topic names\n",
      "                        index=df.index.values) # Makes merge easier\n",
      "\n",
      "df_topics = df.join(df_y_hat) # This is a new data set, so you can remerge"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 51
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "df_topics[['dsk'] + topic_labels].groupby('dsk').mean()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "html": [
        "<div style=\"max-height:1000px;max-width:1500px;overflow:auto;\">\n",
        "<table border=\"1\" class=\"dataframe\">\n",
        "  <thead>\n",
        "    <tr style=\"text-align: right;\">\n",
        "      <th></th>\n",
        "      <th>people like president</th>\n",
        "      <th>women ms children</th>\n",
        "      <th>republican state party</th>\n",
        "      <th>house senate president</th>\n",
        "      <th>abortions state women</th>\n",
        "      <th>rights law religious</th>\n",
        "      <th>york city giuliani</th>\n",
        "      <th>bush dole gore</th>\n",
        "      <th>health tax care</th>\n",
        "      <th>dr drug health</th>\n",
        "      <th>united world states</th>\n",
        "      <th>court judge justice</th>\n",
        "      <th>federal police case</th>\n",
        "      <th>percent poll survey</th>\n",
        "      <th>church catholic pope</th>\n",
        "      <th>clinic anti clinics</th>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <th>dsk</th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "    </tr>\n",
        "  </thead>\n",
        "  <tbody>\n",
        "    <tr>\n",
        "      <th>Editorial Desk</th>\n",
        "      <td> 0.359781</td>\n",
        "      <td> 0.072504</td>\n",
        "      <td> 0.045134</td>\n",
        "      <td> 0.039187</td>\n",
        "      <td> 0.164993</td>\n",
        "      <td> 0.085080</td>\n",
        "      <td> 0.011183</td>\n",
        "      <td> 0.040937</td>\n",
        "      <td> 0.041887</td>\n",
        "      <td> 0.031458</td>\n",
        "      <td> 0.023349</td>\n",
        "      <td> 0.029586</td>\n",
        "      <td> 0.016093</td>\n",
        "      <td> 0.007654</td>\n",
        "      <td> 0.019717</td>\n",
        "      <td> 0.011458</td>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <th>Foreign Desk</th>\n",
        "      <td> 0.372844</td>\n",
        "      <td> 0.111154</td>\n",
        "      <td> 0.032523</td>\n",
        "      <td> 0.052565</td>\n",
        "      <td> 0.065982</td>\n",
        "      <td> 0.036729</td>\n",
        "      <td> 0.002981</td>\n",
        "      <td> 0.003981</td>\n",
        "      <td> 0.052881</td>\n",
        "      <td> 0.027279</td>\n",
        "      <td> 0.156820</td>\n",
        "      <td> 0.004985</td>\n",
        "      <td> 0.013100</td>\n",
        "      <td> 0.007787</td>\n",
        "      <td> 0.051980</td>\n",
        "      <td> 0.006408</td>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <th>Metropolitan Desk</th>\n",
        "      <td> 0.374044</td>\n",
        "      <td> 0.079273</td>\n",
        "      <td> 0.160960</td>\n",
        "      <td> 0.023079</td>\n",
        "      <td> 0.057602</td>\n",
        "      <td> 0.022199</td>\n",
        "      <td> 0.077102</td>\n",
        "      <td> 0.008600</td>\n",
        "      <td> 0.041655</td>\n",
        "      <td> 0.020233</td>\n",
        "      <td> 0.011116</td>\n",
        "      <td> 0.016047</td>\n",
        "      <td> 0.045798</td>\n",
        "      <td> 0.009486</td>\n",
        "      <td> 0.028614</td>\n",
        "      <td> 0.024193</td>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <th>National Desk</th>\n",
        "      <td> 0.383836</td>\n",
        "      <td> 0.049639</td>\n",
        "      <td> 0.118976</td>\n",
        "      <td> 0.085623</td>\n",
        "      <td> 0.074989</td>\n",
        "      <td> 0.036049</td>\n",
        "      <td> 0.004639</td>\n",
        "      <td> 0.055578</td>\n",
        "      <td> 0.040053</td>\n",
        "      <td> 0.028082</td>\n",
        "      <td> 0.008900</td>\n",
        "      <td> 0.041971</td>\n",
        "      <td> 0.026740</td>\n",
        "      <td> 0.012760</td>\n",
        "      <td> 0.015524</td>\n",
        "      <td> 0.016641</td>\n",
        "    </tr>\n",
        "  </tbody>\n",
        "</table>\n",
        "</div>"
       ],
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 52,
       "text": [
        "                   people like president  women ms children  \\\n",
        "dsk                                                           \n",
        "Editorial Desk                  0.359781           0.072504   \n",
        "Foreign Desk                    0.372844           0.111154   \n",
        "Metropolitan Desk               0.374044           0.079273   \n",
        "National Desk                   0.383836           0.049639   \n",
        "\n",
        "                   republican state party  house senate president  \\\n",
        "dsk                                                                 \n",
        "Editorial Desk                   0.045134                0.039187   \n",
        "Foreign Desk                     0.032523                0.052565   \n",
        "Metropolitan Desk                0.160960                0.023079   \n",
        "National Desk                    0.118976                0.085623   \n",
        "\n",
        "                   abortions state women  rights law religious  \\\n",
        "dsk                                                              \n",
        "Editorial Desk                  0.164993              0.085080   \n",
        "Foreign Desk                    0.065982              0.036729   \n",
        "Metropolitan Desk               0.057602              0.022199   \n",
        "National Desk                   0.074989              0.036049   \n",
        "\n",
        "                   york city giuliani  bush dole gore  health tax care  \\\n",
        "dsk                                                                      \n",
        "Editorial Desk               0.011183        0.040937         0.041887   \n",
        "Foreign Desk                 0.002981        0.003981         0.052881   \n",
        "Metropolitan Desk            0.077102        0.008600         0.041655   \n",
        "National Desk                0.004639        0.055578         0.040053   \n",
        "\n",
        "                   dr drug health  united world states  court judge justice  \\\n",
        "dsk                                                                           \n",
        "Editorial Desk           0.031458             0.023349             0.029586   \n",
        "Foreign Desk             0.027279             0.156820             0.004985   \n",
        "Metropolitan Desk        0.020233             0.011116             0.016047   \n",
        "National Desk            0.028082             0.008900             0.041971   \n",
        "\n",
        "                   federal police case  percent poll survey  \\\n",
        "dsk                                                           \n",
        "Editorial Desk                0.016093             0.007654   \n",
        "Foreign Desk                  0.013100             0.007787   \n",
        "Metropolitan Desk             0.045798             0.009486   \n",
        "National Desk                 0.026740             0.012760   \n",
        "\n",
        "                   church catholic pope  clinic anti clinics  \n",
        "dsk                                                           \n",
        "Editorial Desk                 0.019717             0.011458  \n",
        "Foreign Desk                   0.051980             0.006408  \n",
        "Metropolitan Desk              0.028614             0.024193  \n",
        "National Desk                  0.015524             0.016641  "
       ]
      }
     ],
     "prompt_number": 52
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "model_stats['LDA'] = {'time' : elapsed_time ,\n",
      "                      'labels' : topic_labels ,\n",
      "                      'dataframe' : df_topics }\n"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 53
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "from string import punctuation\n",
      "\n",
      "\n",
      "vocab = tf_vectorizer.vocabulary_\n",
      "texts = [[word.strip(punctuation) for word in document.lower().split() if word in vocab ] for document in df['text']]\n",
      "\n",
      "dictionary = corpora.Dictionary(texts)\n",
      "corpus = [dictionary.doc2bow(text) for text in texts]"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 54
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "mallet_path = '/Users/nealcaren/Downloads/mallet-2.0.7/bin/mallet'\n",
      "\n",
      "start_time = datetime.now()\n",
      "\n",
      "model = models.LdaMallet(mallet_path, corpus, num_topics=number_of_topics, id2word=dictionary)\n",
      "\n",
      "elapsed_time = (datetime.now() - start_time).seconds\n",
      "print elapsed_time, 'seconds'"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "284 seconds\n"
       ]
      }
     ],
     "prompt_number": 55
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "topic_labels = []\n",
      "\n",
      "for i in range(0, number_of_topics):\n",
      "    temp = model.show_topic(i, 10)\n",
      "    terms = []\n",
      "    for term in temp:\n",
      "        terms.append(term[1])\n",
      "    print \"topic %s: %s\" % (i,\" \".join(terms))\n",
      "    topic_labels.append(\" \".join(terms[:3]))\n",
      "\n"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "topic 0: women health medical abortions doctors woman drug birth family human\n",
        "topic 1: people make good question time public great issue americans kind\n",
        "topic 2: percent voters poll women 10 black vote support people number\n",
        "topic 3: court law supreme state justice decision federal case states constitutional\n",
        "topic 4: national group groups rights school public political director members executive\n",
        "topic 5: tax federal health money government spending budget care pay state\n",
        "topic 6: house senate vote congress president committee senator representative republicans republican\n",
        "topic 7: people years day man time young told home family children\n",
        "topic 8: catholic church religious john cardinal pope roman moral bishop paul\n",
        "topic 9: united american states government population world foreign international economic family\n",
        "topic 10: police clinic federal death officials clinics operation charges department case\n",
        "topic 11: york state city campaign governor mayor giuliani political jersey support\n",
        "topic 12: judge senator president justice supreme white nomination senate thomas court\n",
        "topic 13: bush president clinton campaign gore vice george white presidential mccain\n",
        "topic 14: democratic republican state democrats race voters won campaign senator election\n",
        "topic 15: republican party dole conservative republicans presidential political convention national bob\n"
       ]
      }
     ],
     "prompt_number": 56
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "y_hat = []\n",
      "predictions = model[corpus]\n",
      "for item in predictions:\n",
      "    y_dict = dict(item)\n",
      "    y_list = [y_dict.get(i,0) for i in range(0,number_of_topics)]\n",
      "    y_hat.append(y_list)\n",
      "\n",
      "df_y_hat = pd.DataFrame(y_hat, \n",
      "                        columns=topic_labels,  # Use our informative topic names\n",
      "                        index=df.index.values) # Makes merge easier\n",
      "\n",
      "df_topics = df.join(df_y_hat) # This is a new data set, so you can remerge"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 57
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "df_topics[['dsk'] + topic_labels].groupby('dsk').mean()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "html": [
        "<div style=\"max-height:1000px;max-width:1500px;overflow:auto;\">\n",
        "<table border=\"1\" class=\"dataframe\">\n",
        "  <thead>\n",
        "    <tr style=\"text-align: right;\">\n",
        "      <th></th>\n",
        "      <th>women health medical</th>\n",
        "      <th>people make good</th>\n",
        "      <th>percent voters poll</th>\n",
        "      <th>court law supreme</th>\n",
        "      <th>national group groups</th>\n",
        "      <th>tax federal health</th>\n",
        "      <th>house senate vote</th>\n",
        "      <th>people years day</th>\n",
        "      <th>catholic church religious</th>\n",
        "      <th>united american states</th>\n",
        "      <th>police clinic federal</th>\n",
        "      <th>york state city</th>\n",
        "      <th>judge senator president</th>\n",
        "      <th>bush president clinton</th>\n",
        "      <th>democratic republican state</th>\n",
        "      <th>republican party dole</th>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <th>dsk</th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "      <th></th>\n",
        "    </tr>\n",
        "  </thead>\n",
        "  <tbody>\n",
        "    <tr>\n",
        "      <th>Editorial Desk</th>\n",
        "      <td> 0.092654</td>\n",
        "      <td> 0.094697</td>\n",
        "      <td> 0.047464</td>\n",
        "      <td> 0.077416</td>\n",
        "      <td> 0.053034</td>\n",
        "      <td> 0.062629</td>\n",
        "      <td> 0.053476</td>\n",
        "      <td> 0.063879</td>\n",
        "      <td> 0.058888</td>\n",
        "      <td> 0.064754</td>\n",
        "      <td> 0.053171</td>\n",
        "      <td> 0.046907</td>\n",
        "      <td> 0.058816</td>\n",
        "      <td> 0.062368</td>\n",
        "      <td> 0.048687</td>\n",
        "      <td> 0.061160</td>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <th>Foreign Desk</th>\n",
        "      <td> 0.086260</td>\n",
        "      <td> 0.054748</td>\n",
        "      <td> 0.055477</td>\n",
        "      <td> 0.043552</td>\n",
        "      <td> 0.051797</td>\n",
        "      <td> 0.053247</td>\n",
        "      <td> 0.055588</td>\n",
        "      <td> 0.086536</td>\n",
        "      <td> 0.111544</td>\n",
        "      <td> 0.183808</td>\n",
        "      <td> 0.051817</td>\n",
        "      <td> 0.030563</td>\n",
        "      <td> 0.030022</td>\n",
        "      <td> 0.036952</td>\n",
        "      <td> 0.034226</td>\n",
        "      <td> 0.033862</td>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <th>Metropolitan Desk</th>\n",
        "      <td> 0.057032</td>\n",
        "      <td> 0.053483</td>\n",
        "      <td> 0.049501</td>\n",
        "      <td> 0.049470</td>\n",
        "      <td> 0.060170</td>\n",
        "      <td> 0.065847</td>\n",
        "      <td> 0.050115</td>\n",
        "      <td> 0.083564</td>\n",
        "      <td> 0.060285</td>\n",
        "      <td> 0.041193</td>\n",
        "      <td> 0.083917</td>\n",
        "      <td> 0.139276</td>\n",
        "      <td> 0.040069</td>\n",
        "      <td> 0.040835</td>\n",
        "      <td> 0.076978</td>\n",
        "      <td> 0.048264</td>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <th>National Desk</th>\n",
        "      <td> 0.062436</td>\n",
        "      <td> 0.060166</td>\n",
        "      <td> 0.055373</td>\n",
        "      <td> 0.075767</td>\n",
        "      <td> 0.060930</td>\n",
        "      <td> 0.059593</td>\n",
        "      <td> 0.076177</td>\n",
        "      <td> 0.064007</td>\n",
        "      <td> 0.047899</td>\n",
        "      <td> 0.043885</td>\n",
        "      <td> 0.058512</td>\n",
        "      <td> 0.036781</td>\n",
        "      <td> 0.065893</td>\n",
        "      <td> 0.076576</td>\n",
        "      <td> 0.081945</td>\n",
        "      <td> 0.074061</td>\n",
        "    </tr>\n",
        "  </tbody>\n",
        "</table>\n",
        "</div>"
       ],
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 58,
       "text": [
        "                   women health medical  people make good  \\\n",
        "dsk                                                         \n",
        "Editorial Desk                 0.092654          0.094697   \n",
        "Foreign Desk                   0.086260          0.054748   \n",
        "Metropolitan Desk              0.057032          0.053483   \n",
        "National Desk                  0.062436          0.060166   \n",
        "\n",
        "                   percent voters poll  court law supreme  \\\n",
        "dsk                                                         \n",
        "Editorial Desk                0.047464           0.077416   \n",
        "Foreign Desk                  0.055477           0.043552   \n",
        "Metropolitan Desk             0.049501           0.049470   \n",
        "National Desk                 0.055373           0.075767   \n",
        "\n",
        "                   national group groups  tax federal health  \\\n",
        "dsk                                                            \n",
        "Editorial Desk                  0.053034            0.062629   \n",
        "Foreign Desk                    0.051797            0.053247   \n",
        "Metropolitan Desk               0.060170            0.065847   \n",
        "National Desk                   0.060930            0.059593   \n",
        "\n",
        "                   house senate vote  people years day  \\\n",
        "dsk                                                      \n",
        "Editorial Desk              0.053476          0.063879   \n",
        "Foreign Desk                0.055588          0.086536   \n",
        "Metropolitan Desk           0.050115          0.083564   \n",
        "National Desk               0.076177          0.064007   \n",
        "\n",
        "                   catholic church religious  united american states  \\\n",
        "dsk                                                                    \n",
        "Editorial Desk                      0.058888                0.064754   \n",
        "Foreign Desk                        0.111544                0.183808   \n",
        "Metropolitan Desk                   0.060285                0.041193   \n",
        "National Desk                       0.047899                0.043885   \n",
        "\n",
        "                   police clinic federal  york state city  \\\n",
        "dsk                                                         \n",
        "Editorial Desk                  0.053171         0.046907   \n",
        "Foreign Desk                    0.051817         0.030563   \n",
        "Metropolitan Desk               0.083917         0.139276   \n",
        "National Desk                   0.058512         0.036781   \n",
        "\n",
        "                   judge senator president  bush president clinton  \\\n",
        "dsk                                                                  \n",
        "Editorial Desk                    0.058816                0.062368   \n",
        "Foreign Desk                      0.030022                0.036952   \n",
        "Metropolitan Desk                 0.040069                0.040835   \n",
        "National Desk                     0.065893                0.076576   \n",
        "\n",
        "                   democratic republican state  republican party dole  \n",
        "dsk                                                                    \n",
        "Editorial Desk                        0.048687               0.061160  \n",
        "Foreign Desk                          0.034226               0.033862  \n",
        "Metropolitan Desk                     0.076978               0.048264  \n",
        "National Desk                         0.081945               0.074061  "
       ]
      }
     ],
     "prompt_number": 58
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "model_stats['MALLET'] = {'time' : elapsed_time ,\n",
      "                      'labels' : topic_labels ,\n",
      "                      'dataframe' : df_topics }\n",
      "\n"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 59
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "from sklearn.linear_model import LogisticRegression\n",
      "from sklearn.metrics import accuracy_score, confusion_matrix, classification_report \n",
      "\n",
      "%matplotlib inline\n",
      "import matplotlib.pyplot as plt\n",
      "\n",
      "\n",
      "lr = LogisticRegression()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 60
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "def cmplot(y_true, y_pred, labels):\n",
      "    print classification_report(y_true, y_pred, labels)\n",
      "    cm = confusion_matrix(y_true, y_pred, labels)\n",
      "    correct = accuracy_score(y_true,y_pred )*100\n",
      "    fig = plt.figure()\n",
      "    ax = fig.add_subplot(111)\n",
      "    cax = ax.matshow(cm)\n",
      "    plt.title('Confusion matrix of the classifier\\n %.1f%% correctly classified' %  correct)\n",
      "    fig.colorbar(cax)\n",
      "    ax.set_xticklabels([''] )\n",
      "    ax.set_yticklabels([''] + categories)\n",
      "    plt.xlabel('Predicted')\n",
      "    plt.ylabel('True')\n",
      "    plt.show()\n",
      "\n"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 61
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "for method in sorted(model_stats):\n",
      "    print '%s took %s seconds.' % (method, model_stats[method]['time'])\n",
      "    predictors = model_stats[method]['labels'][:-1]\n",
      "    df = model_stats[method]['dataframe']\n",
      "    \n",
      "    lr.fit(df[predictors], df['dsk'] )\n",
      "    \n",
      "    prediction = lr.predict(df[predictors])\n",
      "    cmplot(df['dsk'], prediction, categories)\n",
      "    print ''\n",
      "    \n",
      "    "
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "Gensim took 325 seconds.\n",
        "             precision    recall  f1-score   support\n",
        "\n",
        "Editorial Desk       0.60      0.46      0.52      2446\n",
        "Metropolitan Desk       0.75      0.58      0.66      2109\n",
        "National Desk       0.68      0.85      0.75      5406\n",
        "Foreign Desk       0.36      0.20      0.25       762\n",
        "\n",
        "avg / total       0.65      0.66      0.65     10723\n"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n"
       ]
      },
      {
       "metadata": {},
       "output_type": "display_data",
       "png": "iVBORw0KGgoAAAANSUhEUgAAAZMAAAElCAYAAAAhjw8JAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRU5ErkJggg==\n",
       "text": [
        "<matplotlib.figure.Figure at 0x15463b590>"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "LDA took 445 seconds.\n",
        "             precision    recall  f1-score   support\n",
        "\n",
        "Editorial Desk       0.60      0.38      0.47      2446\n",
        "Metropolitan Desk       0.81      0.50      0.62      2109\n",
        "National Desk       0.67      0.90      0.77      5406\n",
        "Foreign Desk       0.65      0.56      0.60       762\n",
        "\n",
        "avg / total       0.68      0.68      0.66     10723\n"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n"
       ]
      },
      {
       "metadata": {},
       "output_type": "display_data",
       "png": "iVBORw0KGgoAAAANSUhEUgAAAZMAAAElCAYAAAAhjw8JAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nXctA4UgAAAAASUVORK5CYII=\n",
       "text": [
        "<matplotlib.figure.Figure at 0x1543bf9d0>"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "MALLET took 284 seconds.\n",
        "             precision    recall  f1-score   support\n",
        "\n",
        "Editorial Desk       0.58      0.27      0.36      2446\n",
        "Metropolitan Desk       0.80      0.61      0.69      2109\n",
        "National Desk       0.66      0.90      0.76      5406\n",
        "Foreign Desk       0.64      0.52      0.58       762\n",
        "\n",
        "avg / total       0.67      0.67      0.64     10723\n"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n"
       ]
      },
      {
       "metadata": {},
       "output_type": "display_data",
       "png": "iVBORw0KGgoAAAANSUhEUgAAAZMAAAElCAYAAAAhjw8JAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nwxxMzMysYQ4mZmbWMAcTMzNr2P8Dnt6duy4DsQsAAAAASUVORK5CYII=\n",
       "text": [
        "<matplotlib.figure.Figure at 0x155438250>"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "NMF took 28 seconds.\n",
        "             precision    recall  f1-score   support\n",
        "\n",
        "Editorial Desk       0.54      0.36      0.43      2446\n",
        "Metropolitan Desk       0.65      0.48      0.55      2109\n",
        "National Desk       0.63      0.84      0.72      5406\n",
        "Foreign Desk       0.35      0.14      0.20       762\n",
        "\n",
        "avg / total       0.59      0.61      0.59     10723\n"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n"
       ]
      },
      {
       "metadata": {},
       "output_type": "display_data",
       "png": "iVBORw0KGgoAAAANSUhEUgAAAZMAAAElCAYAAAAhjw8JAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJzt3XmcXEW9///XmywQSQggyBo2ScQgQoiXyKbhqvkhsl4X\nUNk0V/HGC/7YvKBeCaDgiiIKLgiELRIFESQgiOQKKkQgYCBsAQIkkIDssmb5fP+oaubQ6e6Z6Z5J\nL3k/H4/zmNN16pxT3TNzPl1V51QpIjAzM2vEKs0ugJmZtT8HEzMza5iDiZmZNczBxMzMGuZgYmZm\nDXMwMTOzhjmYWL+TNETSlZKek3RJA8f5tKQ/9GXZmkXSrpLu7Yfj9vqzljRD0sS+LkvZOQ6VdGM/\nHn+6pIMKr78h6SlJj0saIelFSeqv8xsMbHYBrHVI+hRwFPAO4EXgDuCbEfGXBg/9MeBtwNoRsaze\ng0TERcBFDZal30laBmwZEQ9VyxMRNwJb9cPpa37WkiYDb4+IgwrJkZe2FRF7lNYlbUL6Ox4REU/n\n5GFNKdhKxDUTA0DSUcAPgG+QLkYjgJ8Ae/fB4TcF7m8kkLShqt+CJfXnl7iV8bMutwnwdCGQ1E3S\ngD4oz8ohIrys5AswnFQT+WiNPKsCPwQW5OUHwOC8bTwwn/RtcBHwOHBo3nYi8Brwej7HZ4HJwAWF\nY28GLANWya8PBR4EXgAeAj5VSL+xsN9OwN+B54CZwI6FbTOAk4Cb8nH+ALy1ynsrlf9Y4Mlc/n2B\nPYD7gaeB4wr5dwD+Bjyb854BDMrb/pzfy7/y+/144fhfBp4ApuS0x/I+b8/nGJNfbwg8BbyvSnnf\nmd/fs8BdwF5VPuvPlO23e9n2WTn9hlqfFfBe4K/5fHcA76/xdzICuCx/jv8EzqjyuzsdeBR4HrgV\n2KXs8701b1sIfD+nrwZcmI/7bP6dr1v4fU8EPgC8DCzN7/Eclv/7Gg78Mv/u5gMn8+a/vb8Ap+Xz\nnNTs/892WZpeAC/NX/JFZnHpH6pKnpPyBWWdvPyl9I+WL4yLSUFiAPBh4CVgeN5+AnB+4VgnUCWY\nAKvni8jIvG09YHRef+OCBKydLyifzvsdADwDrJW3zwAeALbMF6EbgFOrvLdS+b+Wy/+f+UJyUS7P\n6HyB2jTn3z5f8FYh1QTmAF8qHG8ZsEWF458KDMrlGU8OJjnPfwJ3A0NIF/PvVCnrIGAucBypmXo3\nUgAYVemzrrD/ctvzZzW30mcFbJQ/i93z6w/m1+tUOPYA4E7g+/l9rArsVP67y68/DayVP8OjSEG2\n9OXkb8Cn8/pbgB3y+mHAFbmMAsYAw/K2G4DP5vX3l322m/HmYPJb4KxcxnWBW4DPF8q5GPhiLttq\nzf7/bJfFzVwG8Fbgn1G7aeRTpODxz4j4J+lbcLHdfXHevjQiriZ9M39H3ibe3OzTXUfoMmAbSUMi\nYlFEzKmQ5yPAfRFxUUQsi4hfAffS1SwXwLkRMTciXgWmAdvVOOdiUv/QUuASUrD6YUS8lM8/p7R/\nRNweETPzeR8Bfk66gHX3nk6IiMW5PG8SEWeTLugzSQH0q1WO815g9Yj4VkQsiYgbgN8Dn8zbyz/r\ncpW2B3BOlc/qQGB6RFyTy/lHUq1hD5a3A7ABcGxEvBIRr0XEXysVIv/ens2f4WmkwFP6e3kdGClp\nnYh4OSJmFtLfSvqiERExKyJerPIeK795aT3Sl50jcxmfItW4DyhkezwifpLLttzvyipzMDFITSzr\nSKr197Ah8Ejh9aM57Y1jlAWjl4GhvS1IRLwE7A98AXhc0u8lvaNC1g1zGYoeKSvTwsL6K92U5+mI\niEJeSE12xf1XB5A0KpfrCUnPA98kXeRqeSoiXu8mz9nA1qSmocVV8mwIPFaW9gipBtGIap/VpsDH\nJT1bWoCdgfUrHGME8Eg3X0oAkHSMpDn5rrNnSU1P6+TNE4FRwD2SZkr6SE6/gFRr+5WkBZK+XUf/\n06ak2t0ThffzU1INpaT887UecDAxSM0KrwH71cjzOKm5oGSTnFaPf5GaL0redGGKiGsjYkJOvxf4\nRYVjLCBdGIo2zen97SxSTWXLiBhOqkV0979U824pSUNJ35DPBk6UtFaVrI8DI8puc92U1PbfE73t\nmH+U1CS5VmEZFhHfqZD3MWCT7jqtJe1K6p/6eESsGRFrkZo2BZBrSJ+KiHWBbwO/ybXUJRFxUkRs\nTeov2xM4uJfv5zHS3/pbC+9neERsU8jT1ne2NYuDiRERzwNfB34iaR9Jb5E0SNKHJX07Z5sKfE3S\nOpLWyfkvqPOUdwDvy/f/DweOL22Q9LZchtVJTU8vkTpTy10NjJL0SUkDJe1PutX294U8/fVcwVBS\n5+7LkrYC/qts+yJSp3pvnA7MjIjPA1eRvi1XcjOp1vfl/DsaT7qo/qqH51kEbFbhmYtqn9WFwF6S\nJkgaIGk1SeMlVaoJ3ULq+/hW/htaTdJOFfINA5YA/5Q0WNLXgTXeKIh0oKRSTeF50sV9maTdJG2T\ng9WLpL+PSn8bVUXEE8C1wGmShklaRdLbJb2vN8ex5TmYGAC53fooUif0k6RvpJNInZWQbhm+FfhH\nXm7NaW8cotbhi9tzu/sl+Th/B64sbF8FOJJUw3ga2JWui/Ubx4l02+eewNGkDuFjgD0j4pkqZeru\nWYrybbXyHkPqQ3qB1F/yq7L8k4EpuRnlYzXOHQCS9gEm0PU+jwK2l/TJ5XZIzV97kdr9nwJ+DBwU\nEfcXjlmr7L/OP5+WdGt5WcqPERHzgX2Ar9D1d3E0Fa4duXlrL1JH/qOkWsAnKpTrmrzcD8wjNasV\nmyz/P+AuSS+S7ho8ICJeI/Ul/ZoUYOaQbhyo9oWm1u/zYGBwPsYz+ZjrF/K5ZlIHdTUTm5mZ1cc1\nEzMza5iDiZmZNczBxMzMGuZgYmZmDXMwMWtD+fbcPn24rl2HiZe0maRl3Tx0a/3MH76tUJK+JOkh\nSf/KT0CPzOnrS7oiP9m8LA8jXu0Y60qamvM+J+kmSTsUtm8r6e58oTqykD5I0s1VnpFoafkz2aLZ\n5WhEROwRERfAm4aJ3yoiNoyIx/LDkL69tE05mNgKI+k/SaMG7xERQ0nja/0zb14GTAc+2oNDDSU9\nILc9abDAKcBVkkpP1Z9KulBtC3xV0tty+lHAbyKi35+Sr/QUeB98c+6kyZ36bJh4axErclRJLyvv\nQvri8hiwWzf5BpICyya9PP7zdA3hPoeuIeH/BryHNOTILcCAHhxrF7qGXH8UOCSnDwfOJz28N480\njErpWa1DefPQ5ScD55KGXplOGkLm30lja12aj/EQcHjZZ/QV0oCPL5Ae6NyY6sPal4awP5YUJIvv\n4UekgSorvb92HyZ+FeB7pIc2HySN8PvGvl6a9D/e7AJ4WTkW0jfRZcAR+QL1EOlJcZXl63UwIY1w\n+wpdw5FPIz0dvzFpeI+1gcuBXXtwrE3zhXx/0pDqawPb5m3nk0YEWD3nu4+uYc8PpWzocuA80lwr\nO+Y8Q4DbSKMMDAQ2zxfDCXn7saRRAUrD77+bNGMiVB7WvhRMNiAFmuGFz3ARObiWvb9OGCb+C8A9\npMEt18rHXYqDSXP/x5tdAC8rx0IamG8ZaeiUNQoX4/8sy9erYJKPNRv4n0LaJqTxrW4jBYW9SU1h\nI4Dfkb4hf6zK8Y4HLq2QPoA0QOBWhbTPAzfk9UNJI+YW9zkXOK/welyFPMeThn8nfx57VSlX1WCS\nX19d+ixJgfSuKsfZkVQjWe7CWx5MKmx/Btgmr/8f6cvAOmV5PkOqoW1TYf9iMCkv/xvBhDRsyqsU\n5hIhDbH/p7z+p1Jgya8/hGsmTV88B7ytKKVh3b8TES8AL0j6GWlejLPrOaCkIaTg9NeIKA1ISUQ8\nSuqPIfej/JV0wfkJacDK6aSxn66PiGfLDrsxqdZUbh3S0OXlw/AXO/Mr3V1VHM13U2DDPOx5yQBS\nM1bp3A9WOEZPTCF9Yz+bNAdJtTGrejVMPKmPa0PSeFVr8OZh4k8iDRP/MHBiRFyVzzuCNEz8mqQm\nr69GxJJevJfiMPGltFXoGr9rA978WZdPRWBN4A54W1HuI01uVK6uu3ckrUpquno0Ig6rkfXrwM8j\nTYL0LuDWHMzmU3lk38eqpP+T1Iy1WSFtE94cLKoO5pg9Cjwcbx7OfY2I2LNw7i1rvJdafge8W9K7\nSIH0oir5OmGY+CdIn31J1Tv/bMVxMLEVIiJeJo0U/GVJQyVtDHyOwpDxklYjtbUDrJZfL0fSIOA3\npE7cQ6udU9JoUtv8WTnpYeADSrPtjaTyN9qLgA9K+nge2v6tkraNNAPjNOCbufybkkY3vrDG2y6/\n+2om8KKkL0sakod0f5ek9+TtZwMnS9pSybslrZ231RzWPiJeIXXsXwzcEmm030o6YZj4acARkjZS\nmvfluN4c3/qHg4mtSP9N6ih+nNT0dFFEnFvY/jKp8ztIk2K9VNog6SxJpaCwE+nb94eA5/LDbi9K\n2rnsfD8GjoiIUu3geNINAHeRpuh9sryAEfEYqentaNIQ+LNIHeEAh+cyPQTcSAo8pfJXGrq8fOj9\nZaRv6tvlYzxFGsK+dJE+jXShvJZ0gf4FXcF1Mt0Paz+FVPuqOs9MdMYw8b8gzbh4J+mOsksrHMtW\nMA9Bb9YhJI0gBeH1IuJfzS6PrVxcMzHrAPmByKOBqQ4k1gy+m8uszSlNcbyI1Ce0e5OLYyspN3OZ\nmVnD3MxlZmYNczAxM7OGOZiYmVnDHEzMzKxhDiZmZtYwBxMzM2uYg4mZmTXMwcTMzBrmYGJmZg1z\nMDEzs4Y5mJiZdZg8V84sSVfm15Mlzc9psyR9uJD3eEkPSLpX0oRC+lhJs/O207s7p4OJmVnn+RJp\nLpjS4IsBnBYRY/JyNbwxgdz+wGjSIKFnqmuu5LOAiRExEhgpqeYgoh41uANJ8uidZk0SEeUzbPZK\nb/9/y8+XZzHdA/gmcFQpmeVn/gTYhzRtwWJgnqS5wDhJjwDDImJmznc+sC9pwrSKHEw61S4rMJ48\nMhk2nbziznfTeSvuXECaan7fFXe6Cw9dcecCuHQyfHTyijvfgd9ccefieuADK/B8X+uTo3yjsbP9\nADiWwjTLpJrJ4ZIOJs1OeXREPAdsCNxcyDcf2Ig03XJx6ucFOb0qN3OZmbWYQT1cyknaE3gyImbx\n5prIWcDmpCmjnwC+39dlds3EzKzFVLswzwUerL3rTsDekvYAVgPWkHR+RBxcyiDpbODK/HIBMKKw\n/8akGsmCvF5MX1DrxK6ZWOOGj292CfrZVs0uQP965/hml6Afbd7sAtRlSJVlG1KDa2kpFxFfiYgR\nEbE5cADwp4g4WNIGhWz7AbPz+hXAAZIGS9ocGAnMjIiFwAuSxuUO+YNI7b1VuWZijVtzfLNL0M86\nPJiMHt/sEvSjLZpdgLpUasKqg+i6m+s7krbNrx8GDgOIiDmSppHu/FoCTIqu6XcnAeeR4tj0iKja\n+Q4OJmZmLacvLswRMQOYkdcPqpHvFOCUCum3kSpDPeJgYmbWYvqoZrJCOZiYmbWYdrwwt2OZzcw6\nmmsmZmbWMAcTMzNr2JBmF6AODiZmZi2mHS/M7VhmM7OO5mYuMzNrWDtemNuxzGZmHc01EzMza1g7\nXpjbscxmZh3NNRMzM2uYbw02M7OGuWZiZmYNa8cLczuW2cysow3q6ZV5Sb8Wo1ccTMzMWszANgwm\nbTltr6SlkmYVli9XyDNe0pV5fS9J/5PX95X0zjrOeZikqhPM5DyTJR1dJX1+Luv9ki6tpwy1zmFm\nnWPQgJ4t1UgakK83pWvg2pKuy9efayWtWch7vKQHJN0raUIhfayk2Xnb6d2VuS2DCfByRIwpLN+p\nlTkiroyIb+eX+wKje3MySQMi4mcRcUE3WaNG+mm5rKOAS4A/SVqnN+Xo5hxm1iEGDuzZUsOXSFPx\nlq4XxwHX5evP9fk1kkYD+5OuibsDZ+Y53wHOAiZGxEhgpKTda52wXYNJRZJ2l3SPpNuA/Qrph0o6\nQ9KOwF7Ad3PU3kLSdpJulnSnpMtKEVvSDEk/kPR34EuSTijVCCR9TtJMSXdI+o2kntzJV/oFERHT\ngGuBT+Xjjc3nu1XSNZLWz+lHSLo7l+3iwrGiUI7pklZr5HMzs9YyaNWeLZVI2hjYAzibruvO3sCU\nvD6F9KUaYB9gakQsjoh5wFxgnKQNgGERMTPnO7+wT0XtGkyGlDVzfTxfUH8O7BkRY4H1KfsWHxF/\nA64Ajsm1hIdIH9KxEbEtMBs4oZQdGBQR/xYRp5Wd/9KI2CEitgPuASbW8R5uB7aSNBA4A/hoRLwH\nOBf4Zs7zP8B2uWxfKOwrSf9N+oPZJyJereP8ZtaqBvZwqewHwLHAskLaehGxKK8vAtbL6xsC8wv5\n5gMbVUhfkNNrFrkdvRIRY4oJkrYDHo6IB3PShcDnq+yvvM9wYHhE3JjTpwC/LuS7pMr+20j6BjAc\nGApc0/u38EYgfwewNfDHXLscADyet/0DuFjS5cDlhbIfDDxGCiRL6zi3mbWyKlfmGa+mpRpJewJP\nRsQsSeMr5YmIkNTnzeXtGkwqKf9wVDFX5bzV9nmpyn7nAXtHxGxJhwDje1LAMmOAmfmcd0fEThXy\nfAR4H6lp7quStsllmA1sC4wA5lU8+iOTu9aHj4c16ymimdX2EPBw3x+2ypV5/NC0lJz4wnJZdgL2\nlrQHsBqwhqQLgEWS1o+IhbkJ68mcfwHpOlKyMalGsiCvF9MX1CpyuzZzVXIfsJmkLfLrT1bJ9yKw\nBkBEPA88K2mXvO0gYEaNc5SCzVBgoaRBwIF0BZlaAazrINJHgQ8BU4H7gXUlvTdvGyRpdO4E2yQi\nZpA6y0q1IIBZpGavK/IfxvI2ndy1OJCY9ZMtgA8Ulj4yoIdLmYj4SkSMiIjNgQOAP0XEQaTm/UNy\ntkPoaum4AjhA0mBJmwMjgZkRsRB4QdK4fC06qLBPRe1aMxkiaVbh9dUR8RVJnweukvQycCOwet4e\ndF3wfwX8QtLhwMdJH+xPJb0FeBD4TI3zlo7xv8AtwFP559DC9mq1niMlHZjLNBvYLSKeBpD0MeBH\nudltIKnN837ggpwm4PSIeD43hUVE/EXSMfn9fjAinqlRbjNrJ313ZS5dj74FTJM0kdSa8QmAiJgj\naRrpzq8lwKSIKO0zidQKMwSYHhE1m/PVtZ91CknBLh38e73pvGaXoH9deGizS9C/Dvxm93na1teI\niB61UFQjKd3a05O8d9Dw+fpKu9ZMzMw6VxtemduwyGZmHa4Nr8xtWGQzsw5XY6iUVuVgYmbWatrw\nytyGRTYz63BteGVuwyKbmXW4Nrwyt2GRzcw6XJVBHFuZg4mZWatpwytzGxbZzKzD+W4uMzNrWBte\nmduwyGZmHa4Nr8xtWGQzsw7nZi4zM2tYG16Z27DIZmYdbrVmF6D3HEzMzFqNm7nMzKxhbXhl7qRp\ne83MOsPAHi5lJK0m6RZJd0iaI+nUnD5Z0nxJs/Ly4cI+x0t6QNK9kiYU0sdKmp23nd6TIpuZWSup\ns5krIl6VtFtEvCxpIHCTpF1I0/eeFhGnFfNLGg3sD4wGNgL+KGlknrr3LGBiRMyUNF3S7rWm7nXN\nxMys1dRZMwGIiJfz6mBSWHo2v640ve8+wNSIWBwR84C5wDhJGwDDImJmznc+sG+tIjuYmJm1mgaC\niaRVJN0BLAJuiIi786bDJd0p6ZeS1sxpGwLzC7vPJ9VQytMX5PSaRTYzs1ZSZdTgGQ/CjIdq7xoR\ny4DtJA0H/iBpPKnJ6qSc5WTg+8DEvils4mBiZtZqqlyZx78jLSUn/rH6ISLieUlXAe+JiBmldEln\nA1fmlwuAEYXdNibVSBbk9WL6glpFdjOXmVmrqf9urnVKTViShgAfAmZJWr+QbT9gdl6/AjhA0mBJ\nmwMjgZkRsRB4QdI4SQIOAi7vrsjWiW46q9kl6D8H/lezS9C/zmt2Afrb4mYXoPXV/9DiBsAUSauQ\nKgsXRMT1ks6XtB3prq6HgcMAImKOpGnAHGAJMCnfyQUwifTXOASYXutOLnAwMTNrPXVemSNiNrB9\nhfSDa+xzCnBKhfTbgG16em4HEzOzVtOGV+Y2LLKZWYfz2FxmZtYwjxpsZmYNa8MrcxsW2cysw7mZ\ny8zMGtaGV+Y2LLKZWYdrwytzGxbZzKzDuZnLzMwa5ru5zMysYa6ZmJlZw9rwytyGRTYz63BteGVu\nwyKbmXW4Nrwyt2GRzcw6nPtMzMysYW14ZW7DIpuZdbgqc8C3Mk/ba2bWauqftnc1SbdIukPSHEmn\n5vS1JV0n6X5J15am9s3bjpf0gKR7JU0opI+VNDtvO727IjuYmJm1mjqDSUS8CuwWEdsB7wZ2k7QL\ncBxwXUSMAq7Pr5E0GtgfGA3sDpyZ53wHOAuYGBEjgZGSdq9VZAcTM7NWU2cwAYiIl/PqYFJX/rPA\n3sCUnD4F2Dev7wNMjYjFETEPmAuMk7QBMCwiZuZ85xf2qcjBxMysxcSAni2VSFpF0h3AIuCGiLgb\nWC8iFuUsi4D18vqGwPzC7vOBjSqkL8jpVbkD3sysxSytcmX+vxvh/26qvW9ELAO2kzQc+IOk3cq2\nh6Tom5J2cTAxM2sx1YLJLrulpeTkb1U/RkQ8L+kqYCywSNL6EbEwN2E9mbMtAEYUdtuYVCNZkNeL\n6Qtqlblfm7kkLZN0QeH1QElPSbqym/22lfTh/ixbPs/4Ulkk7SXpf/L6vpLe2YfnWSpplqS78l0W\nRxU6uXp7rH/1VbnMrDW9turgHi3lJK1TulNL0hDgQ8As4ArgkJztEODyvH4FcICkwZI2B0YCMyNi\nIfCCpHH5WnVQYZ+K+rtm8hKwtaTV8l0GHyJFve6qWGNI0fTq8g2SBkbEkr4uaERcCZSC3L55/Z4+\nOvzLETEGQNK6wMXAGsDkOo7V59VTM2stSwfU/Qj8BsAUSauQKgsXRMT1kmYB0yRNBOYBnwCIiDmS\npgFzgCXApIgoXWMmAecBQ4DpEXFNrROviGau6cBHgEuBTwJTgV0BJK0OnAFsDQwiXVyvBk4CVsu3\ntJ1Kum3t7cDmwCOSvgKcC7wVeAr4TEQ8Juk84FVSIFoDOCoirpK0Guk2t7GkD+yoiJhRLKSkQ/P2\ni4G9gPdJ+irwMeADwOdId0fMBQ6KiFfy+Z4H3gOsD3w5Ii6t9WFExFOSPg/8HZgsaQDwLeD9pEeV\nfhIRP89V0UuAYaTf0xci4i+F8q5D+lZxckQsF3TNrH0trXM8lYiYDWxfIf0Z4INV9jkFOKVC+m3A\nNj0994q4m+sSUjVqVVLBbils+ypwfUSMA/4d+C4pqPwv8KuIGBMR03LerYAPRMSngR8D50bEtsBF\nwI8Kx9wkIv6NFMB+ms/7RWBpRLybFNCm5PTlRMTfSBfpYyJi+4h4CLg0InbI927fA0ws7LJ+ROwM\n7EkKCt2KiIeBAZLelo/1XETsAOwAfE7SZrmc1+QazbbAnaX9836/B/7XgcSs8yxhQI+WVtLvNZOI\nmF24OF5VtnkCsJekY/LrVYFNAOXljcMAV0TEa/n1e+m65/lC4DuFfNPyeedKeogUhHYmB5yIuE/S\nI8CobopePP82kr4BDAeGAqXqXpDbESPiHknr0XsT8vE/ll+vAWxJqrmcI2kQcHlElILJYNJDR5Mi\n4sbqh/19YX0U3b9dM+u9eXnpW0vb8N6oFVXiK4DvkZpy1i3b9h8R8UAxQdK4Csd4uex1TzuwS+1/\n5fm763sobj8P2DsHxkOA8YVtr/e2TJK2INWUnsz98P8dEddVyLcrqcZznqTTIuICYDFwK+lp1RrB\nZM+eFMXMGrJZXkr+r0+OWm8zVzOtqIcWzwEm54dniv4AHFF6IWlMXn2R1FdQzV+BA/L6p4E/lw4B\nfFzJ24EtgHtJF91P53OMItV+7qtx/BdJNYSSocDCXEs4kAY6wXMH/E9JfUWQPoNJkgaWyifpLZI2\nAZ6KiLOBX5JuSiCf+7PAVpK+XG85zKx1LWVAj5ZW0t81kwCIiAWkfo5SWulifDLwQ0n/IAW2h0iP\n/d8AHJfvQDi1eKzscOBcSceS7pf+TCHPo8BMUjA4LCJel3QmcFY+zxLgkIhYnB/cicK+pfVfAb+Q\ndDjwcVIfzi2kzv5bSMHlTe+xwnrRkPxeBuXznw/8IG87m/TV5vZ8C96TwH6k2s+xkhaTgtvBpXPk\nh44+CVwh6YWI+GmV85pZG3qN5W/7bXXqugus/Uk6F7gyIi5rdlmaKQXJM5tdjP5z4H81uwT9a2Gz\nC9DP/ji52SXoRycSEXU9Q1YiKe6Kt/co77v0YMPn6yvt18tjZtbhWq0Jqyc6KphExGe6z2Vm1toc\nTMzMrGGt9gxJTziYmJm1GD9nYmZmDXMzl5mZNez1Nrw12MHEzKzFuM/EzMwa5j4TMzNrWDv2mayo\nsbnMzKyH6h2bS9IISTdIujvP7HpETp8saX6e8XVWcSZbScdLekDSvZImFNLHSpqdt53eXZldMzEz\nazEN9JksBo6MiDskDQVuk3QdadzA0yLitGJmSaOB/UkTEG4E/FHSyDzb4lnAxIiYKWm6pN1rzbbo\nYGJm1mJep+Lcfd3Kc7cvzOv/knQPKUhA5Sky9gGmRsRiYJ6kucC4POfTsIiYmfOdT5pDqmowcTOX\nmVmL6Ysh6POkhGOAm3PS4ZLulPRLSWvmtA2B+YXd5pOCT3n6ArqCUkWumZiZtZhqzVz3zHiSe2c8\n2e3+uYnrN8CXcg3lLOCkvPlk4Pu8efrxhnUbTCStQppYavOIOClP2rR+ofpjZmZ9qNqtwaPGb8io\n8Ru+8fp3J85ZLk+exO9S4MKIKE0r/mRh+9nAlfnlAmBEYfeNSTWSBXm9mL6gVpl70sx1JrAj8Kn8\n+l909GQZZmbN1cDdXCLNzDonIn5YSN+gkG0/YHZevwI4QNJgSZsDI4GZue/lBUnj8jEPAi6vVeae\nNHONi4gxeaZAIuKZHPnMzKwfNPCcyc6kqcX/UbpmA18BPilpO9JdXQ8DhwFExBxJ04A5pFlgJ0XX\njImTgPPEgeo/AAASlElEQVSAIcD0WndyQc+CyeuS3nhneQ7zZT18Y2Zm1kv1BpOIuInKLU5X19jn\nFOCUCum3Adv09Nw9CSZnAL8F3ibpFOBjwNd6egIzM+ud1+q8NbiZug0mEXGhpNuAD+SkfSLinv4t\nlpnZyqsdh1Ppyd1cmwAv0dX7H5I2iYhH+7VkZmYrqY4MJsB0UqcNwGrA5sB9wNb9VSgzs5VZRw5B\nHxHvKr6WtD3wxX4rkZnZSm6lGII+Im6XNK4/CmN96YPNLkD/ufC2ZpegX53Ae5pdhH51It9vdhFa\nXkc2c0k6uvByFWB7unkS0szM6teRwQQYWlhfAvye9Ki+mZn1g9c6bQ74/LDiGhFxdK18ZmbWdzqq\nz0TSwIhYImlnSSo8Ym9mZv2o05q5ZpL6R+4Afifp18DLeVtExGX9XTgzs5VRpwWT0qxcqwFPA/9e\ntt3BxMysH3TacybrSjqKrqGKzcxsBeioPhNgADBsRRXEzMySTmvmWhgRJ66wkpiZGQCvd9qtwWZm\ntuK1Y59JrWl7O3g8DjOz1rWUgT1aykkaIekGSXdLukvSETl9bUnXSbpf0rWS1izsc7ykByTdK2lC\nIX2spNl52+ndlblqMImIp3v9CZiZWcPqnQMeWAwcGRFbA+8FvijpncBxwHURMQq4Pr9G0mhgf2A0\nsDtwZp7zHeAsYGJEjARGStq9Vplr1UzMzKwJ6g0mEbEwIu7I6/8C7gE2AvYGpuRsU4B98/o+wNSI\nWBwR84C5wDhJGwDDImJmznd+YZ+K3GdiZtZi+qLPRNJmwBjgFmC9iFiUNy0C1svrGwI3F3abTwo+\ni/N6yYKcXpWDiZlZi6n2nMkzM2bzzIy7ut1f0lDSgLxfiogXu1qu0vAlkvp8eCwHEzOzFlPt1uCh\n48cydPzYN14/dOKvlssjaRApkFwQEZfn5EWS1o+IhbkJ68mcvgAYUdh9Y1KNZEFeL6bXnHrEfSZm\nZi1mCQN6tJTLnee/BOZExA8Lm64ADsnrhwCXF9IPkDRY0ubASGBmRCwEXpA0Lh/zoMI+FblmYmbW\nYhoYTmVn4EDgH5Jm5bTjgW8B0yRNBOYBnwCIiDmSpgFzSPNVTSqMED8JOA8YAkyPiGtqndjBxMys\nxdQ7nEpE3ET1FqeKzw5GxCnAKRXSbwO26em5HUzMzFpMp43NZWZmTeBgYmZmDXuNVZtdhF5zMDEz\nazHtWDNpu1uDJS2T9L3C62MkndDNPu+XtGPh9WGSDurjcm0mabmJxHL6K5JulzRH0i2SDql0jHrP\nYWadpYGxuZqmHWsmrwP7STo1D0bZkyc5dwNeBP4GEBE/68fyVTI3IrYHyPdyXyZJEXHeCi6HmbWB\nThuCvlUtBn4OHFm+QdJekm7OtYDrJL0tj09zGHCkpFmSdpE0WdLReZ/t8j53SrqsNDSzpBmSvpVr\nEvdJ2iWnbybpz5Juy8uO5eWoJSIeBo4CSkNDry7pnHye2yXtndO3zmmzctneXvZet8j5xy5/FjNr\nZ/UOQd9M7RhMAM4EPi1pjbL0GyPivbkWcAnw5TwS5k+B0yJiTL4PO+iq0ZwPHBsR25Lmuy81mQUw\nICLGAf9/IX0R8KGIGAscAPyojvLPArbK618Frs/n+Xfgu5LeQgqAp0fEGGAshaEMJL0D+A1wSL4X\n3Mw6iJu5VpA8cNn5pG/3rxQ2jchPc64PDAYeKmwTZXIwGh4RN+akKcCvC1kuyz9vBzbL64OBH0va\nFlgKjKrjLRTLMgHYS9Ix+fWqwCakJrmvStoYuCwi5ubB2t5GGtZgv4i4t/opijFuXF7MrG/NBR7s\n86O2WqDoibYMJtkPSRf5cwtpZwDfi4jfS3o/MLmXxywPOK/ln0vp+qyOBJ6IiIMkDQBe7eU5IA0L\nPafw+j8i4oGyPPdKuhnYE5gu6TBScHwOeATYFagRTI6oo1hm1jtb5qXk2j456muvt98c8O3azEVE\nPAtMAybS1WS1BvB4Xj+0kP1FYFjZIRQRLwDPlvpDSIOZzejm1GsAC/P6wdC7rxC5D+e7pMAH8AcK\nV35JY/LPzSPi4Yg4A/gdXcMavA78B3CwpE/25txm1h6WLhnYo6WVtGMwKd699X1gncLrycCvJd0K\nPFXIeyXpDrDbC4GjtO0QUj/FncC7gZO6Oe+ZwCGS7gDeAfyrStmK3l66NZjUl3N6RJRmPTsZGCTp\nH5LuAk7M6Z/IczjPArYm9e2INB3By6Qay5GS9qxyTjNrU0uXDOjR0krUNUCkdYo08c39zS5GP3qh\n2QXoVyfwnmYXoV+dyPebXYR+dDQRsVz/bG9IisFPP9+jvK+/dXjD5+srrVVPMjMzlixurVpHTziY\nmJm1mGVL2+/S3H4lNjPrdC3WH9IT7dgBb2bW2V4d2LOlTB5NY1FxDL884sf8PJrGLEkfLmw7XtID\nku6VNKGQPlbS7Lzt9J4U2cHEzKzVLOnhsrxzgd3L0oKuEUDGRMTVAJJGA/sDo/M+Z+b53gHOAiZG\nxEhgpKTyYy7HwcTMrNXUGUzyaB7PVjhipTu+9gGmRsTiPOzUXGCcpA2AYRExM+c7H9i3uyI7mJiZ\ntZr6aybVHJ4HjP1laTBbYENgfiHPfGCjCukLcnpN7oA3M2s1i6uk3zYDbp/R26OdRdfD2CeTHvae\nWFe5anAwMTNrNUurpG83Pi0lvzyxSsYuEfFkaV3S2aQRQSDVOEYUsm5MqpEsyOvF9AV0w81cZmat\npg+buXIfSMl+pKk2AK4ADpA0OE/aNxKYGRELgRckjcsd8geRRiqvyTUTM7NWU89Y5ICkqcD7gXUk\nPUaah2m8pO1Id3U9TJoriYiYk6fsmEMKTZOia3ytScB5wBBgekRc0925HUzMzFpN7zrX3xARlUYS\nP6dG/lOAUyqk30bXSOU94mBiZtZq6gwmzeRgYmbWahxMzMysYdVuDW5hDiZmZq2m2q3BLczBxMys\n1biZy8zMGlbnrcHN5GBiZtZqXDMxM7OGOZiYmVnDHEzMzKxhvjXYzMwa5luDrXXc0uwC9KO5zS5A\nvzqR+5tdhH52ZfdZVna+m8vMzBrmPhMzM2uY+0zMzKxhbdhn4pkWzcxaTZ0zLUo6R9IiSbMLaWtL\nuk7S/ZKulbRmYdvxkh6QdK+kCYX0sZJm522n96TIDiZmZq2m/ml7zwV2L0s7DrguIkYB1+fXSBoN\n7A+MzvucmafpBTgLmBgRI4GRksqPuRwHEzOzVrO4h0uZiLgReLYseW9gSl6fAuyb1/cBpkbE4oiY\nR7pNclyeM35YRMzM+c4v7FOV+0zMzFrNa316tPUiYlFeXwSsl9c3BG4u5JsPbEQKU/ML6Qtyek0O\nJmZmrabarcGLZsCTM+o+bESEpKj7ADU4mJiZtZpqtwavPT4tJXed2JOjLZK0fkQszE1YT+b0BcCI\nQr6NSTWSBXm9mL6gu5O4z8TMrNUs7eHSM1cAh+T1Q4DLC+kHSBosaXNgJDAzIhYCL0galzvkDyrs\nU5VrJmZmrabOJ+AlTQXeD6wj6THg68C3gGmSJgLzgE8ARMQcSdOAOfmMkyKi1AQ2CTgPGAJMj4hr\nuju3g4mZWaupM5hExCerbPpglfynAKdUSL8N2KY353YwMTNrNR5OxczMGta3twavEA4mZmatxqMG\nm5lZw9zMZWZmDWvDUYMdTMzMWo2buczMrGEOJmZm1jD3mZiZWcN8a7CZmTXMzVxmZtawNmzm6ohR\ngyUtlTSrsGzSB8c8UdIH+qh88yT9Iy93SzpZ0qoNHGvtviiXmbWovh01eIXolJrJyxExprc7SRoQ\nERV/JRFxQuPF6jocMD4inpG0OvBz4GfAoXUey8w6WRs2c3VEzaQSSdtJulnSnZIuk7RmTp8h6QeS\n/g4cIWlsTrtV0jWS1s/5zpP00by+h6R7cp4fSboyp0+WdI6kGyQ9KOnw7soVES8BXwD2LZTpWEkz\nc1kn57TVJV0l6Q5JsyV9vOz9DZF0dR5W2sw6yZIeLi2kU2omQyTNyusPRcRHgfOBL0bEjZJOBE4A\njiR9sx8UEf8maSDwZ2CviHha0v7AN4GJOV9IWg34KbBrRDwi6WLeXDsYBewGrAHcJ+nMarWdkoh4\nUdLDwKgcULaMiB0krQL8TtKuwLrAgoj4CICkYYVDDAMuAaZExIX1fWRm1rLasM+kU4LJK8VmLknD\ngeERcWNOmgL8upD/kvxzK2Br4I9pQjEGAI8X8inneSgiHslpU4HP5/UAroqIxcDTkp4E1is7RjXK\nPycAEwrBcHVgS+Am4PuSvgX8PiJuKuz3O+DbETG1+uEvK6y/My9m1rfmAg/2/WFbrNbRE50STLqj\nstcvFdLvjoidauxb3kdRfqzXC+tL6cFnmmsZmwH356RTI+LnFfKNAT4CfEPS9RFxci7PTcCHSYGt\niv/orhhm1rAt81JybbMK8gZJ84AXSNejxbnVY23Sl+hNybMtRsRzOf/xwGdz/iMioq430ZF9JhHx\nPPCspF1y0kHAjEKWUkC4D1hX0nsBJA2SNLp4qJxnC0mb5rT96Qow5YGlFuVzDAXOBH6bf5l/AD6b\nO+aRtJGkdSVtALwaERcB3wOKNxh8Pb+/n/Ti/Ga2cijd8DMmInbIaccB10XEKOD6/Jp8vdsfGA3s\nDpyZm9t7rVOCSaU7nA4BvivpTuDdwEnl+SPideBjwLcl3QHMAnZ804EjXiXNh3yNpFtJEf/5wnF6\nenfVDZJmA7eQvhkclo9/HXAx8DdJ/wCmkfpEtgFuyc1fXwe+UVauL5H6ir7dw/Ob2cqj/Ivu3qTm\nfvLPffP6PsDUiFgcEfNI7XY7UAd1zR9v1UhaPd+FRa4N3B8Rpze5WFVJCrig2cXoR3ObXYB+9ulm\nF6CfXdnsAvSjo4mI3rRYLCf9/77efUYABi93PkkPkb7wLgV+FhG/kPRsRKyVtwt4JiLWknQGcHNu\nAUHS2cDVEXFpb8u9svSZNOpzkg4BBgO3k54RMTPrJ9V64P+cl5p2jognJK0LXCfp3uLGiIgUsKqq\nq4bhYNIDEfFD4IfNLoeZrSyq3Ru8I29uiT9luRwR8UT++ZSk35KarRZJWj8iFub+2Cdz9gXAiMLu\nG+e0XuuUPhMzsw7ySg+XN5P0ltIzafmmngnAbOAKUj8y+eflef0K4ABJgyVtDowEZtZTYtdMzMxa\nTt1PLa4H/DY/NzcQuCgirs03D03LI2bMAz4BEBFzJE0D5pDa1iZFnR3pDiZmZi2nvqcWI+JhYLsK\n6c8AH6yyzylUai/rJQcTM7OW037jqTiYmJm1nPYbT8XBxMys5bhmYmZmDVv+Tq1W52BiZtZy3Mxl\nZmYNczOXmZk1zDUTMzNrmGsmZmbWMNdMzMysYa6ZmJlZw3xrsJmZNcw1EzMza5j7TMzMrGHtVzPx\n5FjWB+5pdgH62bxmF6Cf3dLsAvSjuc0uQJ2W9HBpHQ4m1gccTNpbJweTB5tdgDot7uHSOtzMZWbW\nclqr1tETDiZmZi2n/W4NVp3T/VoLk+RfqlmTRIQa2b+3/7+Nnq+vOJiYmVnD3AFvZmYNczAxM7OG\nOZiYmVnDHEzMzKxhDiZmdZC0VNIsSbMlTZM0pIFjnSfpo3n9F5LeWSPv+yXtWMc55klau94ymnXH\nwcSsPi9HxJiI2AZ4HfhCcaOk3jzDFXkhIj4XEbWGFNgN2Km3hS0d36y/OJiYNe5GYMtca7hR0u+A\nuyStIum7kmZKulPS5wGU/FjSvZKuA95WOpCkGZLG5vXdJd0m6Q5J10naFDgMODLXinaWtK6k3+Rz\nzJS0U973rZKulXSXpF8ALfEsgnUuPwFv1oBcA9kDmJ6TxgBbR8QjOXg8FxE7SFoVuEnStcD2wCjg\nncD6wBzgl3n/AELSusDPgV3zsdaMiOck/RR4MSJOy+e/GPhBRPxF0ibANcBo4ATgzxHxDUl7ABP7\n/cOwlZqDiVl9hkialdf/DJwD7AzMjIhHcvoEYBtJH8uv1wBGArsCF0d6YvgJSX8qO7aA95KCwSMA\nEfFc2faSDwLvlN5IGiZp9XyO/fK+0yU929C7NeuGg4lZfV6JiDHFhHxBf6ks339HxHVl+fag+2an\nnvZxCBgXEa9XKIubtmyFcZ+JWf/5AzCp1BkvaZSkt5BqMvvnPpUNSJ3qRQHcDLxP0mZ539KdWC8C\nwwp5rwWOKL2QtG1e/TPwqZz2YWCtvntbZstzMDGrT6WaQ5Sln03qD7ld0mzgLGBARPwWeCBvmwL8\ndbkDRfwT+DxwmaQ7gKl505XAfqUOeFIgeU/u4L+b1EEPcCIpGN1Fau56BLN+5IEezcysYa6ZmJlZ\nwxxMzMysYQ4mZmbWMAcTMzNrmIOJmZk1zMHEzMwa5mBiZmYN+397ERogb/BnQgAAAABJRU5ErkJg\ngg==\n",
       "text": [
        "<matplotlib.figure.Figure at 0x1564ecb50>"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n"
       ]
      }
     ],
     "prompt_number": 64
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [],
     "language": "python",
     "metadata": {},
     "outputs": []
    }
   ],
   "metadata": {}
  }
 ]
}