sound-bites.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Sound bites\n",
    "\n",
    "Can we detect sound bites that are evolving?\n",
    "\n",
    "The NLP will be harder than the time series analysis (i.e. what exactly is a sound bite and how would we represent it). We use some open source data and tools to make a start."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "A CSV containing every comment from every politician in the Hansard archive as a .csv\n",
    "\n",
    "1 million statements and counting..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1088983\n"
     ]
    }
   ],
   "source": [
    "df = pd.read_csv('./speeches.csv')\n",
    "df.dropna(inplace=True)\n",
    "print(len(df))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Speaker name, transcript and the original .xml file from the archives"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>date</th>\n",
       "      <th>speaker</th>\n",
       "      <th>text</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>debates2010-01-05a.xml</td>\n",
       "      <td>Philip Hollobone</td>\n",
       "      <td>May I suggest to the Secretary of State that p...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>debates2010-01-05a.xml</td>\n",
       "      <td>Philip Hollobone</td>\n",
       "      <td>\"If you commit a crime you will be deported fr...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>debates2010-01-05a.xml</td>\n",
       "      <td>Jack Straw</td>\n",
       "      <td>I am happy to seek to provide the hon. Gentlem...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>debates2010-01-05a.xml</td>\n",
       "      <td>Jack Straw</td>\n",
       "      <td>What we want to see is prison made effective. ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>debates2010-01-05a.xml</td>\n",
       "      <td>David Miliband</td>\n",
       "      <td>We should certainly have a foreign policy that...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                     date           speaker  \\\n",
       "0  debates2010-01-05a.xml  Philip Hollobone   \n",
       "1  debates2010-01-05a.xml  Philip Hollobone   \n",
       "2  debates2010-01-05a.xml        Jack Straw   \n",
       "3  debates2010-01-05a.xml        Jack Straw   \n",
       "4  debates2010-01-05a.xml    David Miliband   \n",
       "\n",
       "                                                text  \n",
       "0  May I suggest to the Secretary of State that p...  \n",
       "1  \"If you commit a crime you will be deported fr...  \n",
       "2  I am happy to seek to provide the hon. Gentlem...  \n",
       "3  What we want to see is prison made effective. ...  \n",
       "4  We should certainly have a foreign policy that...  "
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Ease computational requirements\n",
    "df = df.loc[500000:]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Compute the day and month from the filename (we can use this for grouping transcripts)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>date</th>\n",
       "      <th>speaker</th>\n",
       "      <th>text</th>\n",
       "      <th>day</th>\n",
       "      <th>month</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>500000</th>\n",
       "      <td>debates2013-03-12b.xml</td>\n",
       "      <td>David Gauke</td>\n",
       "      <td>‘notes that this Coalition Government has cut ...</td>\n",
       "      <td>2013-03-12</td>\n",
       "      <td>2013-03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>500001</th>\n",
       "      <td>debates2013-03-12b.xml</td>\n",
       "      <td>David Gauke</td>\n",
       "      <td>After listening to the speech of</td>\n",
       "      <td>2013-03-12</td>\n",
       "      <td>2013-03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>500002</th>\n",
       "      <td>debates2013-03-12b.xml</td>\n",
       "      <td>David Gauke</td>\n",
       "      <td>We might also have thought that the Labour Gov...</td>\n",
       "      <td>2013-03-12</td>\n",
       "      <td>2013-03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>500003</th>\n",
       "      <td>debates2013-03-12b.xml</td>\n",
       "      <td>David Gauke</td>\n",
       "      <td>Whatever the differences that may exist on the...</td>\n",
       "      <td>2013-03-12</td>\n",
       "      <td>2013-03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>500004</th>\n",
       "      <td>debates2013-03-12b.xml</td>\n",
       "      <td>Alex Cunningham</td>\n",
       "      <td>At Treasury questions, one of the Minister’s c...</td>\n",
       "      <td>2013-03-12</td>\n",
       "      <td>2013-03</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                          date          speaker  \\\n",
       "500000  debates2013-03-12b.xml      David Gauke   \n",
       "500001  debates2013-03-12b.xml      David Gauke   \n",
       "500002  debates2013-03-12b.xml      David Gauke   \n",
       "500003  debates2013-03-12b.xml      David Gauke   \n",
       "500004  debates2013-03-12b.xml  Alex Cunningham   \n",
       "\n",
       "                                                     text         day    month  \n",
       "500000  ‘notes that this Coalition Government has cut ...  2013-03-12  2013-03  \n",
       "500001                  After listening to the speech of   2013-03-12  2013-03  \n",
       "500002  We might also have thought that the Labour Gov...  2013-03-12  2013-03  \n",
       "500003  Whatever the differences that may exist on the...  2013-03-12  2013-03  \n",
       "500004  At Treasury questions, one of the Minister’s c...  2013-03-12  2013-03  "
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df['day'] = df.date.apply(lambda x: x.lstrip('debates').rstrip('abcdefghijklmno.xml'))\n",
    "df['month'] = df['day'].apply(lambda x: x[:7])\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The transcript for each day"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "day\n",
       "2013-03-12    ‘notes that this Coalition Government has cut ...\n",
       "2013-03-13    In 1997 there were no excess deaths in the mor...\n",
       "2013-03-14    What steps he is taking to help households imp...\n",
       "2013-03-18    I want to speak very briefly on the royal char...\n",
       "2013-03-19    What assessment he has made of the value of th...\n",
       "Name: text, dtype: object"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "daytext = df.groupby('day')['text'].agg(lambda x: ' '.join(x))\n",
    "daytext.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Tokenisation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We write some tokenisers for words and sentences. Handling some cases particular to our problem. We want to preserve numbers (so don't split on commas!)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.feature_extraction.text import CountVectorizer, TfidfVectorizer\n",
    "import re"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def sent_split(s):\n",
    "    return re.split('(?<!hon)\\.(?=\\W)', s, re.IGNORECASE)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def tokenizer(s):\n",
    "    return [xx.strip('()[];-\\'`.,\"') for xx in s.split(' ') if xx.strip() != '']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def sentence_tokenizer(s):\n",
    "    return [xx.strip('').lower() for xx in sent_split(s) if xx.strip() != '']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Some words that aren't helpful"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.feature_extraction import text\n",
    "stop_words = text.ENGLISH_STOP_WORDS.union([\n",
    "    'government', 'minister', 'hon', 'friend'\n",
    "])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We'll use a TFIDF representation, which emphasises words that are used very often on a single day relative to their usage in the entire archive."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Warning - takes some time!\n",
    "vec = TfidfVectorizer(ngram_range=(1,2), stop_words=stop_words, tokenizer=tokenizer, min_df=2)\n",
    "vec = vec.fit(daytext)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Function to print the most prominent words for a given sentence and vectorizer."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def top_n_words(text, vec, n=10):\n",
    "    df = pd.DataFrame(data={\n",
    "        'token': vec.get_feature_names(),\n",
    "        'count': vec.transform([text]).toarray()[0],\n",
    "    })\n",
    "    return df[df['count']>0].sort_values('count', ascending=False)[:n]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Find context around the word"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def word_context(text, word):\n",
    "    idx = text.lower().index(word)\n",
    "    return text[idx-100:idx+100]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## See it in action"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's look at the most recent date in our archive. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'2017-04-27'"
      ]
     },
     "execution_count": 88,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "daytext.index[-1]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we look at prominent words used that day. No surprises here!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>count</th>\n",
       "      <th>token</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>50502</th>\n",
       "      <td>0.207737</td>\n",
       "      <td>act 2017</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>525234</th>\n",
       "      <td>0.125958</td>\n",
       "      <td>eu</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>528482</th>\n",
       "      <td>0.123721</td>\n",
       "      <td>european</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1537020</th>\n",
       "      <td>0.119762</td>\n",
       "      <td>uk</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>190413</th>\n",
       "      <td>0.109696</td>\n",
       "      <td>brexit</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>610424</th>\n",
       "      <td>0.107632</td>\n",
       "      <td>friend</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1514736</th>\n",
       "      <td>0.106157</td>\n",
       "      <td>trade</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1514760</th>\n",
       "      <td>0.104083</td>\n",
       "      <td>trade agreement</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>585496</th>\n",
       "      <td>0.101178</td>\n",
       "      <td>financial services</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>398774</th>\n",
       "      <td>0.100129</td>\n",
       "      <td>deal</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            count               token\n",
       "50502    0.207737            act 2017\n",
       "525234   0.125958                  eu\n",
       "528482   0.123721            european\n",
       "1537020  0.119762                  uk\n",
       "190413   0.109696              brexit\n",
       "610424   0.107632              friend\n",
       "1514736  0.106157               trade\n",
       "1514760  0.104083     trade agreement\n",
       "585496   0.101178  financial services\n",
       "398774   0.100129                deal"
      ]
     },
     "execution_count": 100,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "top_n_words(daytext[-1], vec)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's dive deeper. Looks like trade agreements are a hot topic. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'ean businesses do the same in the UK. Financial services is one area for which a bold and ambitious trade agreement will be sought, and we will continue to talk to the industry as we prepare for negot'"
      ]
     },
     "execution_count": 95,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "word_context(daytext[-1], 'trade agreement')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Use of sound bites over time\n",
    "\n",
    "But have they always been?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYoAAAEzCAYAAAAvu0StAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nhTHGmAFZojDGGDMgSxTGGGMG9P8BPFk7oYH6YBkAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1450fe0f0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "daytext.str.lower().str.contains('trade agreement').rolling(100).mean().plot()\n",
    "plt.xticks(rotation=50);"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Mentions of a trade agreement have exploded :)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}