2 - Support Vector Machine (Separating Hyperplane).ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from __future__ import division\n",
    "\n",
    "import numpy as np\n",
    "\n",
    "from math import sqrt, log\n",
    "\n",
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "from matplotlib.gridspec import GridSpec\n",
    "\n",
    "import seaborn as sns\n",
    "import colorcet as cc\n",
    "sns.set(style=\"white\", palette=\"deep\", color_codes=True)\n",
    "sns.set_context(\"notebook\", font_scale=1.2)\n",
    "plt.rcParams[\"axes.labelsize\"] = 16\n",
    "\n",
    "def force_aspect(ax, aspect=1):\n",
    "    \"\"\"force a matplotlib axis to have the specified aspect ratio on the screen\"\"\"\n",
    "    xlim = ax.get_xlim()\n",
    "    ylim = ax.get_ylim()\n",
    "    \n",
    "    a = (np.max(xlim) - np.min(xlim)) / (np.max(ylim) - np.min(ylim))\n",
    "    ax.set_aspect(a / aspect)\n",
    "    \n",
    "golden_ratio = 1.61803398875"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "stopwords = [\"a\", \"about\", \"above\", \"according\", \"across\", \"affect\", \"after\", \\\n",
    "\"again\", \"against\", \"all\", \"almost\", \"alone\", \"along\", \"already\", \\\n",
    "\"also\", \"although\", \"always\", \"among\", \"an\", \"and\", \"another\", \"any\", \\\n",
    "\"anyone\", \"anything\", \"anywhere\", \"apt\", \"are\", \"argue\", \"around\", \\\n",
    "\"as\", \"asserted\", \"at\", \"back\", \"be\", \"became\", \"because\", \"become\", \\\n",
    "\"becomes\", \"been\", \"before\", \"behind\", \"being\", \"better\", \"between\", \\\n",
    "\"both\", \"but\", \"by\", \"can\", \"cannot\", \"common\", \"commonly\", \\\n",
    "\"comparable\", \"consequently\", \"considerable\", \"could\", \"course\", \\\n",
    "\"degree\", \"did\", \"do\", \"done\", \"down\", \"dramatically\", \"during\", \\\n",
    "\"each\", \"either\", \"enough\", \"even\", \"ever\", \"every\", \"everyone\", \\\n",
    "\"everything\", \"everywhere\", \"extent\", \"few\", \"find\", \"first\", \\\n",
    "\"follow\", \"for\", \"four\", \"from\", \"full\", \"further\", \"get\", \"give\", \\\n",
    "\"go\", \"had\", \"has\", \"have\", \"he\", \"hence\", \"her\", \"here\", \"herself\", \\\n",
    "\"him\", \"himself\", \"his\", \"how\", \"however\", \"if\", \"imply\", \\\n",
    "\"importance\", \"in\", \"interest\", \"into\", \"is\", \"it\", \"its\", \"itself\", \\\n",
    "\"join\", \"keep\", \"kind\", \"largely\", \"last\", \"least\", \"less\", \"made\", \\\n",
    "\"many\", \"may\", \"me\", \"might\", \"more\", \"moreover\", \"most\", \"mostly\", \\\n",
    "\"much\", \"must\", \"my\", \"myself\", \"necessary\", \"never\", \"next\", \"no\", \\\n",
    "\"nobody\", \"noone\", \"nor\", \"not\", \"nothing\", \"now\", \"nowhere\", \"of\", \\\n",
    "\"off\", \"often\", \"on\", \"once\", \"one\", \"only\", \"or\", \"other\", \"others\", \\\n",
    "\"our\", \"out\", \"over\", \"part\", \"particularly\", \"per\", \"perhaps\", \\\n",
    "\"possible\", \"principle\", \"put\", \"rather\", \"requisite\", \"same\", \\\n",
    "\"second\", \"see\", \"seem\", \"seemed\", \"seeming\", \"seems\", \"several\", \\\n",
    "\"shall\", \"she\", \"should\", \"show\", \"side\", \"since\", \"so\", \"some\", \\\n",
    "\"someone\", \"something\", \"somewhere\", \"still\", \"substance\", \"such\", \\\n",
    "\"suggest\", \"take\", \"than\", \"that\", \"the\", \"their\", \"them\", \"then\", \\\n",
    "\"there\", \"therefore\", \"these\", \"they\", \"thing\", \"this\", \"those\", \\\n",
    "\"though\", \"three\", \"through\", \"throughout\", \"thus\", \"to\", \"together\", \\\n",
    "\"too\", \"toward\", \"two\", \"typical\", \"typically\", \"under\", \\\n",
    "\"unexpected\", \"until\", \"up\", \"upon\", \"us\", \"use\", \"using\", \"very\", \\\n",
    "\"was\", \"we\", \"well\", \"were\", \"what\", \"when\", \"where\", \"whether\", \\\n",
    "\"which\", \"while\", \"whilst\", \"who\", \"whole\", \"whose\", \"why\", \"will\", \\\n",
    "\"with\", \"within\", \"without\", \"would\", \"yet\", \"you\", \"your\", \"yours\"]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Exploratory Analysis: Finding the words which best distinguish Madison from Hamilton"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We'll first try this the standard way, which is easy and informative, but unfortunately also inadequate for the present task.  The basic idea is to determine which words most distinguish Hamilton from Madison, build some linear function which discriminates them, and then apply this model to the disputed papers to determine their authorship.\n",
    "\n",
    "This approach is easy to do, but it suffers from two problems: since Hamilton and Madison are *so* similar in their writing style, most of what we fit is likely to be noise.  And more importantly, there's no way for a linear discriminant to attach an uncertainty to its determination... recall that my entire purpose here is to find an honest and quantitative assessment of the probability that each man wrote the disputed papers."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Tally up word counts"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We want to tally up the occurrences of each of these stop words in our training sets.  We'll split the sets into chunks of 1000 words, so that we can also measure the variance:\n",
    "\n",
    "(My choice of 1000 words here is arbitrary, but I think it's reasonable for writing.  If word rates are Poisonian, then the mean rate and variance should both scale linearly with this fragment size.  Later, we'll see evidence that word rates are \"over-dispersed,\" with a larger variance than Poisson can accomodate.  Fortunately, this non-Poisson-ness seems to be intrisic to each word, and not to depend strongly on fragment length.  In either case, this arbitrary choice does not much influence my results.)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import nltk\n",
    "from collections import Counter"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def sample_usage(filename):\n",
    "    \"\"\"Split filename into chunks of 1000 words, then tally the instances of\n",
    "    each of the stopwords in each chunk.  Return the mean and variance for\n",
    "    each word.\"\"\"\n",
    "    with open(filename, 'r') as f:\n",
    "        raw = f.read()\n",
    "\n",
    "    all_words = nltk.word_tokenize(raw.lower())\n",
    "\n",
    "    # split into chunks of 1000 words\n",
    "    n=1000\n",
    "    samples = [ all_words[i:i + n] \n",
    "                for i in xrange(0, len(all_words), n) ]\n",
    "\n",
    "    # tally each of the stopwords, for each chunk\n",
    "    c = [ Counter(s) for s in samples ]\n",
    "    x = [ [ cc[w] for w in stopwords ] for cc in c ]\n",
    "    x = np.asarray(x)\n",
    "\n",
    "    # return lists of the mean and variance for each stop word\n",
    "    mu = np.mean(x, axis=0)\n",
    "    va = np.var(x,  axis=0)\n",
    "    \n",
    "    return (mu, va)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "mu_m, v_m = sample_usage('../authors/madison-corpus-small.txt')\n",
    "mu_h, v_h = sample_usage('../authors/hamilton-corpus-small.txt')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Find words for which the difference in Madison's and Hamilton's usage exceeds the variance:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# recall that i floor the occurrence rates at one per million...\n",
    "# for Poissonian statistics, this also floors the variance at one per million\n",
    "delta = (mu_m - mu_h) / np.sqrt(v_m + v_h + 1.0e-6)\n",
    "delta = np.abs(delta)\n",
    "ind = np.flipud(np.argsort(delta))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can get a sense of the available information by making a quick plot:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAk0AAAGMCAYAAADHppbUAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngIQmIYQQQogG+P/9guhejF1s3gAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a17939fd0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(figsize=(12,6))\n",
    "\n",
    "num = 10\n",
    "ax.plot(delta[ind][:num], '-o')\n",
    "ax.set_ylabel('discriminating power $\\Delta$')\n",
    "\n",
    "# label x-ticks with words and force it to show every one of them\n",
    "ax.set_xticks(range(num))\n",
    "labels = [ stopwords[i] for i in ind[:num] ]\n",
    "ax.set_xticklabels(labels, rotation=60);\n",
    "ax.axhline(1.0, color='k')\n",
    "\n",
    "ax.set_title('Most Important Words');\n",
    "force_aspect(ax, aspect=golden_ratio)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Clearly, the first 3 points stand out from the rest, with $\\Delta > 1$ (if barely).  The rest of the detections are not technically significant.  So we'll have a look at these first three:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['on', 'there', 'upon']\n"
     ]
    }
   ],
   "source": [
    "keep = np.where(delta>1)[0]\n",
    "set1 = [ stopwords[i] for i in keep ]\n",
    "print set1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "And let's also keep everything with delta>0.5, just in case it helps us:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['and', 'at', 'by', 'kind', 'on', 'there', 'to', 'upon', 'would']\n"
     ]
    }
   ],
   "source": [
    "keep = np.where(delta>0.5)[0]\n",
    "set2 = [ stopwords[i] for i in keep ]\n",
    "print set2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Investigate key words by eye"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can start by plotting our training set according to the occurrences of each of our key words.  If Madison's and Hamilton's clearly seperate, then this might suggest a way to tell them apart.  Once can always hope!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import glob"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def measure(filename, wordlist=stopwords):\n",
    "    \"\"\"count frequency for each word in wordlist for the file filename\"\"\"\n",
    "    with open(filename, 'r') as f:\n",
    "        raw = f.read()\n",
    "\n",
    "    all_words = nltk.word_tokenize(raw.lower())\n",
    "    c = Counter(all_words)\n",
    "    \n",
    "    tally = [ c[word] for word in wordlist ]    \n",
    "    \n",
    "    return np.asarray(tally) / len(all_words)\n",
    "\n",
    "xm = [ measure(f, stopwords) for f in glob.glob('../authors/madison/*.txt')]\n",
    "xh = [ measure(f, stopwords) for f in glob.glob('../authors/hamilton/*.txt')]\n",
    "\n",
    "# transpose data so that each row corresponds to a word\n",
    "# (and each column corresponds to a different federalist paper)\n",
    "xm = np.asarray(xm).T\n",
    "xh = np.asarray(xh).T"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAiYAAAIzCAYAAAA01dJuAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a17967b10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig=plt.figure(figsize=(8, 8))\n",
    "gs=GridSpec(2,2) # one row, two columns\n",
    "\n",
    "ax1=fig.add_subplot(gs[0,0]) # first row, first column\n",
    "ax2=fig.add_subplot(gs[0,1]) # first row, second column\n",
    "ax3=fig.add_subplot(gs[1,0]) # second row, first column\n",
    "ax4=fig.add_subplot(gs[1,1]) # second row, second column\n",
    "\n",
    "def compare_usage(x, word1, word2):\n",
    "    return (x[stopwords.index(word1)] * 1000, \n",
    "            x[stopwords.index(word2)] * 1000)\n",
    "\n",
    "def plot_comparison(w1, w2, ax):\n",
    "    x, y = compare_usage(xm, w1, w2)\n",
    "    ax.plot(x, y, 'o', color='r', label = 'madison')\n",
    "    x, y = compare_usage(xh, w1, w2)\n",
    "    ax.plot(x, y, '^', color='g', label = 'hamilton')\n",
    "    ax.set_xlabel(w1)\n",
    "    ax.set_ylabel(w2)\n",
    "\n",
    "plot_comparison('on',     'upon',  ax1)\n",
    "plot_comparison('to',     'there', ax2)\n",
    "plot_comparison('by',     'kind',  ax3)\n",
    "plot_comparison('whilst', 'while', ax4)\n",
    "\n",
    "ax1.legend(loc='upper right')\n",
    "    \n",
    "for ax in (ax1, ax2, ax3, ax4):\n",
    "    force_aspect(ax, 1)\n",
    "\n",
    "fig.suptitle('Word Frequencies (per 1000 words)')\n",
    "fig.tight_layout(rect=[0.0, 0.0, 1.0, 0.95])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Clearly, Hamilton uses the word *upon* much more than Madison, while Madison uses the word *on* much more than Hamilton.  Hamilton uses the words *there* and *to* more frequently on average, but the scatter is so large that this doesn't seem very predictive at all.  The words *while* and *whilst* seem promising, but they're so rare that their non-occurrence in a paper hardly tells us anything at all.\n",
    "\n",
    "While the *on*-*upon* pair comes close, unfortunately none of these conclusively disambiguate Madison from Hamilton.  We can check whether some *linear combination* of these words disambiguates the authors, however: maybe in those few cases where Hamilton rarely uses *upon*, he also uses the words *there* and *to* a lot.  (This actually is a standard method, even if it seems like a stretch!)\n",
    "\n",
    "Equivalently, if we plot these points in some high-dimenional space, maybe there exists some hyperplane which does separate the points, even if it's not aligned with any of these axes.  We can search for such a hyperplane using the \"support vector machine\" implemented in scikit-learn:"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Support Vector Machine"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## First, let's take a look at set1 (with $\\Delta > 1$):"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Tally up the occurrences of each word in our set for Madison's papers, for Hamilton's papers, and for the disputed papers."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "mad_data_1 = [ measure(f, set1) for f in glob.glob('../authors/madison/*.txt') ]\n",
    "ham_data_1 = [ measure(f, set1) for f in glob.glob('../authors/hamilton/*.txt') ]\n",
    "\n",
    "test_files = glob.glob('../authors/unknown-federalist/*.txt')\n",
    "u_data_1 = [ measure(f, set1) for f in test_files ]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For convenience, store the names of the papers as well:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import re\n",
    "names = [ re.sub(r'.*?([0-9]+).*', r'\\1', f) for f in test_files ]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn import svm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Federalist # 18 : madison\n",
      "Federalist # 19 : madison\n",
      "Federalist # 20 : madison\n",
      "Federalist # 49 : madison\n",
      "Federalist # 50 : madison\n",
      "Federalist # 51 : madison\n",
      "Federalist # 52 : madison\n",
      "Federalist # 53 : madison\n",
      "Federalist # 54 : madison\n",
      "Federalist # 55 : madison\n",
      "Federalist # 56 : madison\n",
      "Federalist # 57 : madison\n",
      "Federalist # 62 : madison\n",
      "Federalist # 63 : madison\n"
     ]
    }
   ],
   "source": [
    "# x is a list of points in 3-dimensional cartesian space\n",
    "# - one point for each file in the training set\n",
    "# - one dimenension for each word in set1\n",
    "#\n",
    "x = np.vstack((mad_data_1, ham_data_1))\n",
    "\n",
    "# y is a list of labels for each of the points.  let's let\n",
    "# 0 = 'Madison' and 1 = 'Hamilton'\n",
    "#\n",
    "y = [0] * len(mad_data_1) + [1] * len(ham_data_1)\n",
    "\n",
    "# we have a cartesian space of words, which honestly doesn't\n",
    "# make a whole lot of sense.  we'll scale each axis by its\n",
    "# standard deviation so that distances are measured in units\n",
    "# of the variance, rather than in units of 'upons'...\n",
    "#\n",
    "scale = np.std(x, axis=0)\n",
    "x = x / scale\n",
    "\n",
    "# use a linear support vector machine, to look for a simple\n",
    "# separating hyperplane\n",
    "#\n",
    "clf = svm.SVC(kernel='linear')\n",
    "clf.fit(x, y)\n",
    "\n",
    "# now let's see what the model predicts!\n",
    "result = clf.predict(u_data_1/scale)\n",
    "result = [ 'hamilton' if r else 'madison' for r in result ]\n",
    "\n",
    "for name, author in zip(names, result):\n",
    "    print \"Federalist #\", name, \":\", author"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "So we can see that the model did in fact find a separating hyperplane, and that it implies that Madison wrote every single one of the disputed Federalist papers, including numbers 18, 19, and 20, which are traditionally thought to be co-written.\n",
    "\n",
    "This is great, but it's hard to know how certain we can be in this conclusion.  This has been the subject of more than 200 years of historical debate... are the historians *really* going to believe this analysis?\n",
    "\n",
    "As a first check, we can make sure the model correctly classifies all of the *known* federalist papers.  We trained the model on this data, so it doesn't seem like a whole lot to ask:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Should all be Madison:\n",
      "madison\n",
      "madison\n",
      "madison\n",
      "madison\n",
      "madison\n",
      "madison\n",
      "madison\n",
      "madison\n",
      "madison\n",
      "-> hamilton <-\n",
      "madison\n",
      "madison\n",
      "madison\n",
      "madison\n",
      "madison\n"
     ]
    }
   ],
   "source": [
    "result = clf.predict(mad_data_1/scale)\n",
    "result = [ '-> hamilton <-' if r else 'madison' for r in result ]\n",
    "\n",
    "print \"Should all be Madison:\"\n",
    "for author in result:\n",
    "    print author"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Should all be Hamilton:\n",
      "hamilton\n",
      "hamilton\n",
      "hamilton\n",
      "hamilton\n",
      "hamilton\n",
      "hamilton\n",
      "hamilton\n",
      "-> madison <-\n",
      "hamilton\n",
      "hamilton\n",
      "hamilton\n",
      "hamilton\n",
      "hamilton\n",
      "hamilton\n",
      "hamilton\n"
     ]
    }
   ],
   "source": [
    "result = clf.predict(ham_data_1/scale)\n",
    "result = [ 'hamilton' if r else '-> madison <-' for r in result ]\n",
    "\n",
    "print \"Should all be Hamilton:\"\n",
    "for author in result[:15]:\n",
    "    print author"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Doh! We can inspect the model more visually by plotting each paper's distance \"above\" or \"below\" the separating hyperplane:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "b = clf.intercept_[0]\n",
    "v = clf.coef_[0] / scale"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "m = [ np.dot(v, el) + b for el in mad_data_1 ]\n",
    "h = [ np.dot(v, el) + b for el in ham_data_1 ]\n",
    "u = [ np.dot(v, el) + b for el in   u_data_1 ]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAekAAAHBCAYAAABNMUPMAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a17954c90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(figsize=(8, 8))\n",
    "\n",
    "ax.plot(np.random.normal(0.0, 1.0, len(m)), m, 'o', color='r', label='madison')\n",
    "ax.plot(np.random.normal(0.0, 1.0, len(h)), h, '^', color='g', label='hamilton')\n",
    "ax.plot(np.random.normal(0.0, 1.0, len(u)), u, 's', color='b', label='disputed')\n",
    "\n",
    "ax.axhline(y=0, color='k')\n",
    "\n",
    "ax.set_xticks([])\n",
    "ax.set_ylabel('\"Hamilton-ness\" (<0 if Madison, >0 if Hamilton)')\n",
    "ax.legend(loc=1);\n",
    "\n",
    "force_aspect(ax, 1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "You can see that, although all of the disputed papers happen to fall below this plane, the separation really isn't all that strong.  You can also see (as we found before) that one each of Madison's and Hamilton's papers from the training set are mis-attributed.\n",
    "\n",
    "So unfortunately this method doesn't seem terribly reliable... certainly not enough to settle a 200+ year-old academic and political feud!\n",
    "\n",
    "We can redo it with the larger sample of words, just to see if we do any better:"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Next, let's take a look at set2 (with $\\Delta > 0.5$):"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This data is noisier, but there is more of it, so maybe it will help us in our determinations.  For convenience, I'll roll our analysis into a function:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def analyze_svc(wordlist):\n",
    "    # tally word occurrences in the the known and disputed federalist papers\n",
    "    mad_data = [ measure(f, wordlist) for f in glob.glob('../authors/madison/*.txt') ]\n",
    "    ham_data = [ measure(f, wordlist) for f in glob.glob('../authors/hamilton/*.txt') ]\n",
    "\n",
    "    test_files = glob.glob('../authors/unknown-federalist/*.txt')\n",
    "    u_data = [ measure(f, wordlist) for f in test_files ]\n",
    "    \n",
    "    # fit a linear classifier\n",
    "    x = np.vstack((mad_data, ham_data))\n",
    "    y = [0] * len(mad_data) + [1] * len(ham_data)\n",
    "\n",
    "    scale = np.std(x, axis=0) + 1.0e-6\n",
    "    x = x / scale\n",
    "\n",
    "    clf = svm.SVC(kernel='linear')\n",
    "    clf.fit(x, y)\n",
    "\n",
    "    # print predictions for the disputed papers\n",
    "    result = clf.predict(u_data/scale)\n",
    "    result = [ 'hamilton' if r else 'madison' for r in result ]\n",
    "\n",
    "    for name, author in zip(names, result):\n",
    "        print \"Federalist #\", name, \":\", author\n",
    "        \n",
    "    # plot the results visually\n",
    "    b = clf.intercept_[0]\n",
    "    v = clf.coef_[0] / scale\n",
    "    \n",
    "    m = [ np.dot(v, el) + b for el in mad_data ]\n",
    "    h = [ np.dot(v, el) + b for el in ham_data ]\n",
    "    u = [ np.dot(v, el) + b for el in   u_data ]\n",
    "    \n",
    "    fig, ax = plt.subplots(figsize=(8, 8))\n",
    "\n",
    "    ax.plot(np.random.normal(0.0, 1.0, len(m)), m, 'o', color='r', label='madison')\n",
    "    ax.plot(np.random.normal(0.0, 1.0, len(h)), h, '^', color='g', label='hamilton')\n",
    "    ax.plot(np.random.normal(0.0, 1.0, len(u)), u, 's', color='b', label='disputed')\n",
    "\n",
    "    ax.axhline(y=0, color='k')\n",
    "\n",
    "    ax.set_xticks([])\n",
    "    ax.set_ylabel('\"Hamilton-ness\" (<0 if Madison, >0 if Hamilton)')\n",
    "    ax.legend(loc='best');\n",
    "\n",
    "    force_aspect(ax, 1)\n",
    "    \n",
    "    return (b, v)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Federalist # 18 : madison\n",
      "Federalist # 19 : madison\n",
      "Federalist # 20 : madison\n",
      "Federalist # 49 : madison\n",
      "Federalist # 50 : madison\n",
      "Federalist # 51 : madison\n",
      "Federalist # 52 : madison\n",
      "Federalist # 53 : madison\n",
      "Federalist # 54 : madison\n",
      "Federalist # 55 : hamilton\n",
      "Federalist # 56 : madison\n",
      "Federalist # 57 : madison\n",
      "Federalist # 62 : madison\n",
      "Federalist # 63 : madison\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAekAAAHBCAYAAABNMUPMAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAADYFCENAIBN/X/WdW/ttf101AAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a17e95e50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "b, v = analyze_svc(set2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Oh, no -- one of our determinations has changed!  Now it looks like Hamilton may have written Federalist #55.\n",
    "\n",
    "Again, we can plot distances from the hyperplane to gauge how well the model performs:"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we've separated our training data much more effectively, which looks promising.  But note that Federalist #55 has moved significantly, and is now *firmly* in Hamilton's territory.  It's not comforting to see such a marked change in our predictions.\n",
    "\n",
    "Moreover, I just don't think this model is very well-motivated for our present calculation.  Even if Hamilton tends to us \"on\" while Madison prefers \"upon,\" and Hamilton uses \"there\" more than Madison, who says that in the few instances where Madison uses \"on\" more often, he also uses \"there\" less often?  I see no reason why such a correlation should exist.\n",
    "\n",
    "The computer will always be able to find such correlations, if you give it enough words to check, but I think it's essentially just picking up on noise in the training data.\n",
    "\n",
    "To demonstrate the fungibility of our model, I can add Federalist #55 (which is very clearly on the Hamilton side) to the training data.  If I tell the model that Madison wrote it, the SVM will happily find a new separating hyperplane, and now reports that Madison wrote everything:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Federalist # 18 : madison\n",
      "Federalist # 19 : madison\n",
      "Federalist # 20 : madison\n",
      "Federalist # 49 : madison\n",
      "Federalist # 50 : madison\n",
      "Federalist # 51 : madison\n",
      "Federalist # 52 : madison\n",
      "Federalist # 53 : madison\n",
      "Federalist # 54 : madison\n",
      "Federalist # 55 : madison\n",
      "Federalist # 56 : madison\n",
      "Federalist # 57 : madison\n",
      "Federalist # 62 : madison\n",
      "Federalist # 63 : madison\n"
     ]
    }
   ],
   "source": [
    "ind = names.index('55')\n",
    "\n",
    "# tally word occurrences in the the known and disputed federalist papers\n",
    "mad_data = [ measure(f, set2) for f in glob.glob('../authors/madison/*.txt') ]\n",
    "ham_data = [ measure(f, set2) for f in glob.glob('../authors/hamilton/*.txt') ]\n",
    "\n",
    "test_files = glob.glob('../authors/unknown-federalist/*.txt')\n",
    "u_data = [ measure(f, set2) for f in test_files ]\n",
    "\n",
    "# fit a linear classifier, adding federalist #55 to the training data:\n",
    "x = np.vstack((mad_data, ham_data, [u_data[ind]]))\n",
    "y = [0] * len(mad_data) + [1] * len(ham_data) + [0]\n",
    "\n",
    "scale = np.std(x, axis=0) + 1.0e-6\n",
    "x = x / scale\n",
    "\n",
    "clf = svm.SVC(kernel='linear')\n",
    "clf.fit(x, y)\n",
    "\n",
    "# print predictions for the disputed papers\n",
    "result = clf.predict(u_data/scale)\n",
    "result = [ 'hamilton' if r else 'madison' for r in result ]\n",
    "\n",
    "for name, author in zip(names, result):\n",
    "    print \"Federalist #\", name, \":\", author"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "So our model is sensitive to noise, has possibly been trained on noise, and is subject to huge swings in its predictions based on small changes in the training set.\n",
    "\n",
    "Nonetheless, this model is the best we have so far, and it does seem suggestive that Madison wrote many, most, or perhaps all of the disputed papers.  This model is in fact better than at least a few which have been published in peer-reviewed journals, so let's save it for posterity:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'H = -14.3 + (6.4 kind) + (4.7 upon) + (4.2 there) + (3.9 at) - (1.5 on) - (1.0 by) + (0.8 would) + (0.6 to) - (0.5 and)'"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "fact = 10 * v / np.linalg.norm(v)\n",
    "\n",
    "terms = sorted(zip(fact, set2), \n",
    "               key = lambda(a, b) : abs(a), \n",
    "               reverse=True)\n",
    "\n",
    "def format_term(term):\n",
    "    weight, word = term\n",
    "    \n",
    "    s = '-' if np.sign(weight) < 0 else '+'\n",
    "    weight = '{0:.1f}'.format(abs(weight))\n",
    "    return s + ' (' + weight + ' ' + word + ')'\n",
    "    \n",
    "\n",
    "'H = ' '{0:.1f}'.format(10*b) + ' ' + \\\n",
    "' '.join(map(format_term, terms))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We might call this the \"Hamilton index:\" we predict Hamilton wrote the paper if $H>0$, otherwise Madison."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## What happens if we use *even more* data?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "I can try re-running the model with the *full* list of stopwords, hoping that I can surmount the problems with this method by overwhelming it with data:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Federalist # 18 : madison\n",
      "Federalist # 19 : madison\n",
      "Federalist # 20 : hamilton\n",
      "Federalist # 49 : madison\n",
      "Federalist # 50 : madison\n",
      "Federalist # 51 : madison\n",
      "Federalist # 52 : madison\n",
      "Federalist # 53 : madison\n",
      "Federalist # 54 : madison\n",
      "Federalist # 55 : madison\n",
      "Federalist # 56 : madison\n",
      "Federalist # 57 : madison\n",
      "Federalist # 62 : madison\n",
      "Federalist # 63 : madison\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfMAAAHBCAYAAAB9vVLtAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x10e08ea50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "b, v = analyze_svc(stopwords)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "again, the predictions have changed, which is not encouraging.  Moreover, while the model now does an excellent job at separating the training data, it performs much less well on the disputed papers.  This is a clear sign of regression or selection effects; i.e., that we've trained the model on noise in the training data and that it isn't terribly predictive."
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.14"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}