cdp_topic_models.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Copy of Copy of topic_models.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yOwTyRgrKmT0"
      },
      "source": [
        "## Google Authentications & Libraries\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sTZdPRlfKSsZ"
      },
      "source": [
        "# Code to read csv file into Colaboratory:\n",
        "# reference: https://towardsdatascience.com/3-ways-to-load-csv-files-into-colab-7c14fcbdcb92\n",
        "!pip install -U -q PyDrive\n",
        "from pydrive.auth import GoogleAuth\n",
        "from pydrive.drive import GoogleDrive\n",
        "from google.colab import auth\n",
        "from oauth2client.client import GoogleCredentials\n",
        "\n",
        "# Authenticate and create the PyDrive client.\n",
        "auth.authenticate_user()\n",
        "gauth = GoogleAuth()\n",
        "gauth.credentials = GoogleCredentials.get_application_default()\n",
        "my_drive = GoogleDrive(gauth)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hNkxSxAGLC2s",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 71
        },
        "outputId": "692debad-24ec-4101-d965-c594641fb872"
      },
      "source": [
        "%matplotlib inline\n",
        "import pandas as pd\n",
        "import numpy as np\n",
        "import nltk\n",
        "from nltk.corpus import stopwords\n",
        "import gensim\n",
        "from gensim.models import LdaModel\n",
        "from gensim import models, corpora, similarities\n",
        "import re\n",
        "from nltk.stem.porter import PorterStemmer\n",
        "import time\n",
        "from nltk import FreqDist\n",
        "from scipy.stats import entropy\n",
        "import matplotlib.pyplot as plt\n",
        "import seaborn as sns\n",
        "sns.set_style(\"darkgrid\")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/statsmodels/tools/_testing.py:19: FutureWarning: pandas.util.testing is deprecated. Use the functions in the public API at pandas.testing instead.\n",
            "  import pandas.util.testing as tm\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jbhhYIjiLOcr"
      },
      "source": [
        "## Import Data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2Y-0r7gvLJZK",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 357
        },
        "outputId": "987c8980-2050-43d7-b9e5-7318c6be0eb4"
      },
      "source": [
        "# Cities Questionnaire\n",
        "cities_q_id = '1MdA6rqOjPAugJR4qFOgoNGBiLtpqcLmV'\n",
        "cities_q_file = my_drive.CreateFile({'id':cities_q_id}) \n",
        "cities_q_file.GetContentFile('cities_q.csv')\n",
        "cities_q_df = pd.read_csv('cities_q.csv', encoding='latin1')\n",
        "print(cities_q_df.shape)\n",
        "cities_q_df.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(135511, 12)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/IPython/core/interactiveshell.py:2718: DtypeWarning: Columns (7) have mixed types.Specify dtype option on import or set low_memory=False.\n",
            "  interactivity=interactivity, compiler=compiler, result=result)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>Project Year</th>\n",
              "      <th>Account Number</th>\n",
              "      <th>Account Name</th>\n",
              "      <th>Question Number</th>\n",
              "      <th>Question Name</th>\n",
              "      <th>Row Number</th>\n",
              "      <th>Row Name</th>\n",
              "      <th>Column Number</th>\n",
              "      <th>Column Name</th>\n",
              "      <th>Response Answer</th>\n",
              "      <th>Comments</th>\n",
              "      <th>File Name</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>2017</td>\n",
              "      <td>63999</td>\n",
              "      <td>City of Miami Beach, FL</td>\n",
              "      <td>0.1</td>\n",
              "      <td>Please give a general description and introdu...</td>\n",
              "      <td>1</td>\n",
              "      <td>NaN</td>\n",
              "      <td>C1</td>\n",
              "      <td>Administrative boundary</td>\n",
              "      <td>City/Municipality</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>2017</td>\n",
              "      <td>62864</td>\n",
              "      <td>City of Lancaster, PA</td>\n",
              "      <td>0.1</td>\n",
              "      <td>Please give a general description and introdu...</td>\n",
              "      <td>1</td>\n",
              "      <td>NaN</td>\n",
              "      <td>C1</td>\n",
              "      <td>Administrative boundary</td>\n",
              "      <td>City/Municipality</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>2017</td>\n",
              "      <td>61790</td>\n",
              "      <td>City of Emeryville, CA</td>\n",
              "      <td>0.1</td>\n",
              "      <td>Please give a general description and introdu...</td>\n",
              "      <td>1</td>\n",
              "      <td>NaN</td>\n",
              "      <td>C1</td>\n",
              "      <td>Administrative boundary</td>\n",
              "      <td>City/Municipality</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>2017</td>\n",
              "      <td>58485</td>\n",
              "      <td>Abington Township</td>\n",
              "      <td>0.1</td>\n",
              "      <td>Please give a general description and introdu...</td>\n",
              "      <td>1</td>\n",
              "      <td>NaN</td>\n",
              "      <td>C1</td>\n",
              "      <td>Administrative boundary</td>\n",
              "      <td>City/Municipality</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>2017</td>\n",
              "      <td>54102</td>\n",
              "      <td>City of Albany</td>\n",
              "      <td>0.1</td>\n",
              "      <td>Please give a general description and introdu...</td>\n",
              "      <td>1</td>\n",
              "      <td>NaN</td>\n",
              "      <td>C1</td>\n",
              "      <td>Administrative boundary</td>\n",
              "      <td>City/Municipality</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "   Project Year  Account Number  ... Comments File Name\n",
              "0          2017           63999  ...      NaN       NaN\n",
              "1          2017           62864  ...      NaN       NaN\n",
              "2          2017           61790  ...      NaN       NaN\n",
              "3          2017           58485  ...      NaN       NaN\n",
              "4          2017           54102  ...      NaN       NaN\n",
              "\n",
              "[5 rows x 12 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 4
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jlEMNpNz-Rj1",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 323
        },
        "outputId": "7fd87c28-7c61-409f-c722-63fec0ca656b"
      },
      "source": [
        "# Corporations Questionnaire\n",
        "corp_q_id = '11Ax7FfLRBKqAyGc_ctieDp5BjlI5pPqa'\n",
        "corp_q_file = my_drive.CreateFile({'id':corp_q_id}) \n",
        "corp_q_file.GetContentFile('corp_q.csv')\n",
        "corp_q_df = pd.read_csv('corp_q.csv', encoding='latin1')\n",
        "print(corp_q_df.shape)\n",
        "corp_q_df.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(747904, 10)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>Project Year</th>\n",
              "      <th>Account Number</th>\n",
              "      <th>Account Name</th>\n",
              "      <th>Question Number</th>\n",
              "      <th>Question Name</th>\n",
              "      <th>Row Number</th>\n",
              "      <th>Row Name</th>\n",
              "      <th>Column Number</th>\n",
              "      <th>Column Name</th>\n",
              "      <th>Response Answer</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>2018</td>\n",
              "      <td>285</td>\n",
              "      <td>3M Company</td>\n",
              "      <td>CC0.1</td>\n",
              "      <td>Give a general description and introduction to...</td>\n",
              "      <td>0.0</td>\n",
              "      <td>NaN</td>\n",
              "      <td>C0</td>\n",
              "      <td>NaN</td>\n",
              "      <td>3M is a global science company committed to in...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>2018</td>\n",
              "      <td>64</td>\n",
              "      <td>Abbott Laboratories</td>\n",
              "      <td>CC0.1</td>\n",
              "      <td>Give a general description and introduction to...</td>\n",
              "      <td>0.0</td>\n",
              "      <td>NaN</td>\n",
              "      <td>C0</td>\n",
              "      <td>NaN</td>\n",
              "      <td>Abbott is a global healthcare company that hel...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>2018</td>\n",
              "      <td>45103</td>\n",
              "      <td>AbbVie Inc</td>\n",
              "      <td>CC0.1</td>\n",
              "      <td>Give a general description and introduction to...</td>\n",
              "      <td>0.0</td>\n",
              "      <td>NaN</td>\n",
              "      <td>C0</td>\n",
              "      <td>NaN</td>\n",
              "      <td>AbbVie is a global, research-driven biopharmac...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>2018</td>\n",
              "      <td>291</td>\n",
              "      <td>Abercrombie &amp; Fitch Co.</td>\n",
              "      <td>CC0.1</td>\n",
              "      <td>Give a general description and introduction to...</td>\n",
              "      <td>0.0</td>\n",
              "      <td>NaN</td>\n",
              "      <td>C0</td>\n",
              "      <td>NaN</td>\n",
              "      <td>Abercrombie &amp; Fitch Co. (''A&amp;F ), a company i...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>2018</td>\n",
              "      <td>58</td>\n",
              "      <td>ABM INDUSTRIES INC</td>\n",
              "      <td>CC0.1</td>\n",
              "      <td>Give a general description and introduction to...</td>\n",
              "      <td>0.0</td>\n",
              "      <td>NaN</td>\n",
              "      <td>C0</td>\n",
              "      <td>NaN</td>\n",
              "      <td>Building Value With Integrated Facilities Solu...</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "   Project Year  ...                                    Response Answer\n",
              "0          2018  ...  3M is a global science company committed to in...\n",
              "1          2018  ...  Abbott is a global healthcare company that hel...\n",
              "2          2018  ...  AbbVie is a global, research-driven biopharmac...\n",
              "3          2018  ...  Abercrombie & Fitch Co. (''A&F ), a company i...\n",
              "4          2018  ...  Building Value With Integrated Facilities Solu...\n",
              "\n",
              "[5 rows x 10 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 5
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HCIXHHxFMlJu",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 425
        },
        "outputId": "c29ca41e-3b06-4790-b997-35cc2df8af71"
      },
      "source": [
        "# Data Catalog (contains metadata & links from all datasets)\n",
        "data_catalog_id = '1llkhxscCNYR1CE9LgRPBngVwuC2SP0ab'\n",
        "data_catalog_file = my_drive.CreateFile({'id':data_catalog_id}) \n",
        "data_catalog_file.GetContentFile('data_catalog.csv')\n",
        "data_catalog_df = pd.read_csv('data_catalog.csv', encoding='latin1')\n",
        "print(data_catalog_df.shape)\n",
        "data_catalog_df.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(314, 14)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>ï»¿CDP Dataset Source File Name</th>\n",
              "      <th>Project Name</th>\n",
              "      <th>Datapoint Theme</th>\n",
              "      <th>2019 Question Number</th>\n",
              "      <th>2019 Question Name</th>\n",
              "      <th>Use</th>\n",
              "      <th>2019 Row Number</th>\n",
              "      <th>2019 Row Name</th>\n",
              "      <th>2019 Column Number</th>\n",
              "      <th>2019 Column Name</th>\n",
              "      <th>Comparable 2018 datapoint provided</th>\n",
              "      <th>Comparable 2017 datapoint provided</th>\n",
              "      <th>Comparable 2016 datapoint provided</th>\n",
              "      <th>Comparable 2015 datapoint provided</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>Cities_Data_2017-2019_mb2.cv</td>\n",
              "      <td>Cities</td>\n",
              "      <td>City details</td>\n",
              "      <td>0.1</td>\n",
              "      <td>Please give a general description and introduc...</td>\n",
              "      <td>0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>City boundary</td>\n",
              "      <td>1</td>\n",
              "      <td>Administrative boundary</td>\n",
              "      <td>0.1</td>\n",
              "      <td>0.1</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>Cities_Data_2017-2019_mb2.cv</td>\n",
              "      <td>Cities</td>\n",
              "      <td>City details</td>\n",
              "      <td>0.1</td>\n",
              "      <td>Please give a general description and introduc...</td>\n",
              "      <td>1</td>\n",
              "      <td>1.0</td>\n",
              "      <td>City boundary</td>\n",
              "      <td>2</td>\n",
              "      <td>Description of city</td>\n",
              "      <td>0.1</td>\n",
              "      <td>0.1</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>Cities_Data_2017-2019_mb2.cv</td>\n",
              "      <td>Cities</td>\n",
              "      <td>City details</td>\n",
              "      <td>0.2</td>\n",
              "      <td>If you have not previously submitted your Lett...</td>\n",
              "      <td>0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "      <td>0.2</td>\n",
              "      <td>0.2</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>Cities_Data_2017-2019_mb2.cv</td>\n",
              "      <td>Cities</td>\n",
              "      <td>City details</td>\n",
              "      <td>0.3</td>\n",
              "      <td>Please provide information about your cityâs...</td>\n",
              "      <td>0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>Please complete</td>\n",
              "      <td>1</td>\n",
              "      <td>Leader title</td>\n",
              "      <td>0.3</td>\n",
              "      <td>0.3</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>Cities_Data_2017-2019_mb2.cv</td>\n",
              "      <td>Cities</td>\n",
              "      <td>City details</td>\n",
              "      <td>0.3</td>\n",
              "      <td>Please provide information about your cityâs...</td>\n",
              "      <td>0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>Please complete</td>\n",
              "      <td>2</td>\n",
              "      <td>Leader name</td>\n",
              "      <td>0.3</td>\n",
              "      <td>0.3</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "  ï»¿CDP Dataset Source File Name  ... Comparable 2015 datapoint provided\n",
              "0    Cities_Data_2017-2019_mb2.cv  ...                                NaN\n",
              "1    Cities_Data_2017-2019_mb2.cv  ...                                NaN\n",
              "2    Cities_Data_2017-2019_mb2.cv  ...                                NaN\n",
              "3    Cities_Data_2017-2019_mb2.cv  ...                                NaN\n",
              "4    Cities_Data_2017-2019_mb2.cv  ...                                NaN\n",
              "\n",
              "[5 rows x 14 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yK2VwRfDNSIe"
      },
      "source": [
        "## Clean Data\n",
        "\n",
        "Filters:\n",
        "\n",
        "\n",
        "*   2019 Questions\n",
        "*   Recommended Challenge 1 Questions w/ Themes City/Company details & Climate-related targets and goals\n",
        "*   Free text questions (manually identified)\n",
        "\n",
        "\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hrhR1RVHNRL2",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 751
        },
        "outputId": "07f21538-a9ce-40e8-8fd1-17e642f0aff9"
      },
      "source": [
        "\n",
        "catalog_cols = ['Project Name', 'Datapoint Theme', '2019 Question Number', '2019 Question Name', '2019 Row Number', '2019 Row Name', '2019 Column Number', '2019 Column Name']\n",
        "data_catalog_rec_df = data_catalog_df.loc[(data_catalog_df['Use'] == 1) & \n",
        "                                          (data_catalog_df['Datapoint Theme'].isin(['City details', 'Climate-related targets and goals', 'Company details', 'Climate targets and goals'])), \n",
        "                                          catalog_cols]\n",
        "\n",
        "cities_q_rec_2019_df = cities_q_df[cities_q_df['Project Year'] == 2019].merge(data_catalog_rec_df, \n",
        "                                         left_on = ['Question Number', 'Column Number'],\n",
        "                                         right_on = ['2019 Question Number', '2019 Column Number'])\n",
        "cities_q_rec_2019_df['Account Type'] = \"City\"\n",
        "\n",
        "corp_q_rec_2019_df = corp_q_df[corp_q_df['Project Year'] == 2019].merge(data_catalog_rec_df, \n",
        "                                     left_on = ['Question Number', 'Column Number'],\n",
        "                                     right_on = ['2019 Question Number', '2019 Column Number'])\n",
        "corp_q_rec_2019_df['Account Type'] = \"Corporation\"\n",
        "\n",
        "\n",
        "#cities_q_rec_2019_df.head()\n",
        "print(\"\\nNumber of Relevant Responses: \" + str(cities_q_rec_2019_df.shape[0] + corp_q_rec_2019_df.shape[0]))\n",
        "print(\"\\nRelevant City Questions: \" + \"\\n\\t- \" + \"\\n\\t- \".join(cities_q_rec_2019_df['Question Name'].unique()))\n",
        "print(\"\\nRelevant Corp Questions: \" + \"\\n\\t- \" + \"\\n\\t- \".join(corp_q_rec_2019_df['Question Name'].unique()))\n",
        "\n",
        "#distinct_entries = cities_q_rec_2019_df[['Question Number','Question Name', 'Row Number', 'Row Name', 'Column Number', 'Column Name']].drop_duplicates()\n",
        "#print(\"\\nNumber of Distinct Questions: \", distinct_entries.shape[0])\n",
        "#distinct_entries.head(31)\n",
        "\n",
        "data_cols = ['Account Type', 'Account Number', 'Account Name', 'Question Number', \n",
        "             'Question Name', 'Row Number', 'Row Name', 'Column Number', \n",
        "             'Column Name', 'Response Answer']\n",
        "\n",
        "# union dataframes\n",
        "q_rec_2019_df = pd.concat([cities_q_rec_2019_df[data_cols],corp_q_rec_2019_df[data_cols]])\n",
        "\n",
        "num_cities_with_text_responses = q_rec_2019_df.loc[q_rec_2019_df['Account Type'] == \"City\", 'Account Number'].drop_duplicates().shape[0]\n",
        "num_corps_with_text_responses = q_rec_2019_df.loc[q_rec_2019_df['Account Type'] == \"Corporation\", 'Account Number'].drop_duplicates().shape[0]\n",
        "\n",
        "print(\"\\nNumber of Cities w/ Text Responses: \" + str(num_cities_with_text_responses))\n",
        "print(\"\\nNumber of Corporations w/ Text Responses: \" + str(num_corps_with_text_responses))\n",
        "\n",
        "print(\"\\nCombined Data Frame:\")\n",
        "q_rec_2019_df.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\n",
            "Number of Relevant Responses: 19433\n",
            "\n",
            "Relevant City Questions: \n",
            "\t- Please give a general description and introduction to your city including your citys reporting boundary in the table below.\n",
            "\t- Please provide details of your total city-wide base year intensity target. An intensity target is usually measured per capita or per unit GDP. If you have an absolute emissions reduction target, please select Base year emissions (absolute) target in question 5.0.\n",
            "\t- Please provide details of your total city-wide base year emissions reduction (absolute) target. In addition, you may add rows to provide details of your sector-specific targets, by providing the base year emissions specific to that target.\n",
            "\t- Please provide details of your total fixed level target.\n",
            "\t- Please provide details of your total city-wide baseline scenario target, including projected business as usual emissions.\n",
            "\n",
            "Relevant Corp Questions: \n",
            "\t- Give a general description and introduction to your organization.\n",
            "\t- Why did you not have any emissions reduction initiatives active during the reporting year?\n",
            "\t- Provide details of your absolute emissions target(s) and progress made against those targets. \n",
            "\t- Provide details of your emissions intensity target(s) and progress made against those target(s). \n",
            "\t- Explain why you do not have emissions target and forecast how your emissions will change over the next five years. \n",
            "\t- Provide details of other key climate-related targets not already reported in question C4.1/a/b.\n",
            "\t- Provide details on the initiatives implemented in the reporting year in the table below.\n",
            "\t- What methods do you use to drive investment in emissions reduction activities? \n",
            "\n",
            "Number of Cities w/ Text Responses: 25\n",
            "\n",
            "Number of Corporations w/ Text Responses: 963\n",
            "\n",
            "Combined Data Frame:\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>Account Type</th>\n",
              "      <th>Account Number</th>\n",
              "      <th>Account Name</th>\n",
              "      <th>Question Number</th>\n",
              "      <th>Question Name</th>\n",
              "      <th>Row Number</th>\n",
              "      <th>Row Name</th>\n",
              "      <th>Column Number</th>\n",
              "      <th>Column Name</th>\n",
              "      <th>Response Answer</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>City</td>\n",
              "      <td>58485</td>\n",
              "      <td>Abington Township</td>\n",
              "      <td>0.1</td>\n",
              "      <td>Please give a general description and introduc...</td>\n",
              "      <td>1.0</td>\n",
              "      <td>City boundary</td>\n",
              "      <td>2</td>\n",
              "      <td>Description of city</td>\n",
              "      <td>Abington Township is the 52nd certified STAR C...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>City</td>\n",
              "      <td>55799</td>\n",
              "      <td>Arlington, VA</td>\n",
              "      <td>0.1</td>\n",
              "      <td>Please give a general description and introduc...</td>\n",
              "      <td>1.0</td>\n",
              "      <td>City boundary</td>\n",
              "      <td>2</td>\n",
              "      <td>Description of city</td>\n",
              "      <td>Arlington Va., is a world-class residential, b...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>City</td>\n",
              "      <td>74414</td>\n",
              "      <td>Boulder County</td>\n",
              "      <td>0.1</td>\n",
              "      <td>Please give a general description and introduc...</td>\n",
              "      <td>1.0</td>\n",
              "      <td>City boundary</td>\n",
              "      <td>2</td>\n",
              "      <td>Description of city</td>\n",
              "      <td>Located in north-central Colorado, northwest o...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>City</td>\n",
              "      <td>63941</td>\n",
              "      <td>Broward County, FL</td>\n",
              "      <td>0.1</td>\n",
              "      <td>Please give a general description and introduc...</td>\n",
              "      <td>1.0</td>\n",
              "      <td>City boundary</td>\n",
              "      <td>2</td>\n",
              "      <td>Description of city</td>\n",
              "      <td>Broward County, Florida</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>City</td>\n",
              "      <td>49339</td>\n",
              "      <td>City and County of Honolulu</td>\n",
              "      <td>0.1</td>\n",
              "      <td>Please give a general description and introduc...</td>\n",
              "      <td>1.0</td>\n",
              "      <td>City boundary</td>\n",
              "      <td>2</td>\n",
              "      <td>Description of city</td>\n",
              "      <td>Established by city charter in 1907, the City ...</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "  Account Type  ...                                    Response Answer\n",
              "0         City  ...  Abington Township is the 52nd certified STAR C...\n",
              "1         City  ...  Arlington Va., is a world-class residential, b...\n",
              "2         City  ...  Located in north-central Colorado, northwest o...\n",
              "3         City  ...                            Broward County, Florida\n",
              "4         City  ...  Established by city charter in 1907, the City ...\n",
              "\n",
              "[5 rows x 10 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 7
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "w6R6ICuJBmQA"
      },
      "source": [
        "Concatenate all relevant text responses by Account to get one long text response.\n",
        "\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "34sqFg5uCO1j",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        },
        "outputId": "68d04344-4539-45e9-8196-75c5701f7d38"
      },
      "source": [
        "q_rec_2019_df['Response Answer'] = q_rec_2019_df['Response Answer'].astype('str')\n",
        "#q_rec_2019_df['Response Answer'] = q_rec_2019_df['Response Answer'].str.lower()\n",
        "\n",
        "groupby_cols = ['Account Type', 'Account Number', 'Account Name']\n",
        "df2 = q_rec_2019_df.sort_values(by='Row Number').groupby(groupby_cols)['Response Answer'].apply('. '.join).reset_index()\n",
        "\n",
        "df2.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>Account Type</th>\n",
              "      <th>Account Number</th>\n",
              "      <th>Account Name</th>\n",
              "      <th>Response Answer</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>City</td>\n",
              "      <td>1093</td>\n",
              "      <td>City of Atlanta</td>\n",
              "      <td>Benchmarking and retrofits for buildings large...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>City</td>\n",
              "      <td>1184</td>\n",
              "      <td>City of Austin</td>\n",
              "      <td>Our target is net-zero emissions by 2050 and i...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>City</td>\n",
              "      <td>35268</td>\n",
              "      <td>City of Boston</td>\n",
              "      <td>Public facility. Commercial buildings. Residen...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>City</td>\n",
              "      <td>35853</td>\n",
              "      <td>City of Baltimore</td>\n",
              "      <td>Key findings by sector provide guidance for wh...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>City</td>\n",
              "      <td>37241</td>\n",
              "      <td>City of Berkeley</td>\n",
              "      <td>Public facility. Industrial facilities. Reside...</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "  Account Type  ...                                    Response Answer\n",
              "0         City  ...  Benchmarking and retrofits for buildings large...\n",
              "1         City  ...  Our target is net-zero emissions by 2050 and i...\n",
              "2         City  ...  Public facility. Commercial buildings. Residen...\n",
              "3         City  ...  Key findings by sector provide guidance for wh...\n",
              "4         City  ...  Public facility. Industrial facilities. Reside...\n",
              "\n",
              "[5 rows x 4 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "iRFLoTocPfEp"
      },
      "source": [
        "Use NLTK library to extract tokens from cleaned text"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7G_h3SJMMhV7",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 85
        },
        "outputId": "832a76ff-aebe-4f86-cf45-4d61ce705c61"
      },
      "source": [
        "from nltk.util import ngrams\n",
        "nltk.download('stopwords')\n",
        "nltk.download('punkt')\n",
        "\n",
        "def initial_clean(text):\n",
        "    \"\"\"\n",
        "    Cleans text using string replace & tokenizes:\n",
        "    - removes websites, emails & any punctuation\n",
        "    - converts to lower case\n",
        "    - removes N/A responses\n",
        "    \"\"\"\n",
        "    text = text.lower() # lower case the text\n",
        "    text = re.sub(\"((\\S+)?(http(s)?)(\\S+))|((\\S+)?(www)(\\S+))|((\\S+)?(\\@)(\\S+)?)\", \" \", text)\n",
        "    text = re.sub(\"[^a-zA-Z ]\", \"\", text)\n",
        "    text = re.sub(\"nan\", \"\", text)\n",
        "    text = re.sub(\"question not applicable\", \"\", text)\n",
        "    text = nltk.word_tokenize(text)\n",
        "    return text\n",
        "\n",
        "stop_words = stopwords.words('english')\n",
        "def remove_stop_words(text):\n",
        "    \"\"\"\n",
        "    Removes all stopwords from tokens\n",
        "    \"\"\"\n",
        "    return [word for word in text if word not in stop_words]\n",
        "\n",
        "stemmer = PorterStemmer()\n",
        "def stem_words(text):\n",
        "    \"\"\"\n",
        "    Converts tokens to stem words, so plural and singular tokens are treated the same\n",
        "    \"\"\"\n",
        "    try:\n",
        "        text = [stemmer.stem(word) for word in text]\n",
        "        text = [word for word in text if len(word) > 1] # exclude 1 letter words\n",
        "    except IndexError: \n",
        "        pass\n",
        "    return text\n",
        "\n",
        "def apply_all(text):\n",
        "    \"\"\"\n",
        "    Applies all cleaning steps to text response\n",
        "    \"\"\"\n",
        "    return stem_words(remove_stop_words(initial_clean(text)))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[nltk_data] Downloading package stopwords to /root/nltk_data...\n",
            "[nltk_data]   Unzipping corpora/stopwords.zip.\n",
            "[nltk_data] Downloading package punkt to /root/nltk_data...\n",
            "[nltk_data]   Unzipping tokenizers/punkt.zip.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "y5YODIJ9oEQa",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 221
        },
        "outputId": "08032bf5-1026-4799-edcb-0ddf90d22bf2"
      },
      "source": [
        "t1 = time.time()\n",
        "df2['Tokenized Response Answer'] = df2['Response Answer'].apply(apply_all)\n",
        "t2 = time.time()\n",
        "print(\"Time to clean and tokenize\", len(df2), \"accounts:\", round((t2-t1)/60, 3), \"min\")\n",
        "\n",
        "df2[['Tokenized Response Answer', 'Response Answer']].head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Time to clean and tokenize 988 accounts: 0.224 min\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>Tokenized Response Answer</th>\n",
              "      <th>Response Answer</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>[benchmark, retrofit, build, larger, sqft, com...</td>\n",
              "      <td>Benchmarking and retrofits for buildings large...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>[target, netzero, emiss, allow, baselin, emiss...</td>\n",
              "      <td>Our target is net-zero emissions by 2050 and i...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>[public, facil, commerci, build, residenti, bu...</td>\n",
              "      <td>Public facility. Commercial buildings. Residen...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>[key, find, sector, provid, guidanc, strategi,...</td>\n",
              "      <td>Key findings by sector provide guidance for wh...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>[public, facil, industri, facil, residenti, bu...</td>\n",
              "      <td>Public facility. Industrial facilities. Reside...</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                           Tokenized Response Answer                                    Response Answer\n",
              "0  [benchmark, retrofit, build, larger, sqft, com...  Benchmarking and retrofits for buildings large...\n",
              "1  [target, netzero, emiss, allow, baselin, emiss...  Our target is net-zero emissions by 2050 and i...\n",
              "2  [public, facil, commerci, build, residenti, bu...  Public facility. Commercial buildings. Residen...\n",
              "3  [key, find, sector, provid, guidanc, strategi,...  Key findings by sector provide guidance for wh...\n",
              "4  [public, facil, industri, facil, residenti, bu...  Public facility. Industrial facilities. Reside..."
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 10
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "WjmL2fe50Pyo"
      },
      "source": [
        "Convert singular tokens to bigrams\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vxFYjHNP7og_",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 238
        },
        "outputId": "4154d590-aeab-4c35-c9c5-ad9b7ab5de11"
      },
      "source": [
        "df2['Response Answer Bigrams'] = df2['Tokenized Response Answer'].apply(lambda x: ['_'.join(i) for i in ngrams(x, 2)])\n",
        "df2.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:1: DeprecationWarning: generator 'ngrams' raised StopIteration\n",
            "  \"\"\"Entry point for launching an IPython kernel.\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>Account Type</th>\n",
              "      <th>Account Number</th>\n",
              "      <th>Account Name</th>\n",
              "      <th>Response Answer</th>\n",
              "      <th>Tokenized Response Answer</th>\n",
              "      <th>Response Answer Bigrams</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>City</td>\n",
              "      <td>1093</td>\n",
              "      <td>City of Atlanta</td>\n",
              "      <td>Benchmarking and retrofits for buildings large...</td>\n",
              "      <td>[benchmark, retrofit, build, larger, sqft, com...</td>\n",
              "      <td>[benchmark_retrofit, retrofit_build, build_lar...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>City</td>\n",
              "      <td>1184</td>\n",
              "      <td>City of Austin</td>\n",
              "      <td>Our target is net-zero emissions by 2050 and i...</td>\n",
              "      <td>[target, netzero, emiss, allow, baselin, emiss...</td>\n",
              "      <td>[target_netzero, netzero_emiss, emiss_allow, a...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>City</td>\n",
              "      <td>35268</td>\n",
              "      <td>City of Boston</td>\n",
              "      <td>Public facility. Commercial buildings. Residen...</td>\n",
              "      <td>[public, facil, commerci, build, residenti, bu...</td>\n",
              "      <td>[public_facil, facil_commerci, commerci_build,...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>City</td>\n",
              "      <td>35853</td>\n",
              "      <td>City of Baltimore</td>\n",
              "      <td>Key findings by sector provide guidance for wh...</td>\n",
              "      <td>[key, find, sector, provid, guidanc, strategi,...</td>\n",
              "      <td>[key_find, find_sector, sector_provid, provid_...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>City</td>\n",
              "      <td>37241</td>\n",
              "      <td>City of Berkeley</td>\n",
              "      <td>Public facility. Industrial facilities. Reside...</td>\n",
              "      <td>[public, facil, industri, facil, residenti, bu...</td>\n",
              "      <td>[public_facil, facil_industri, industri_facil,...</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "  Account Type  ...                            Response Answer Bigrams\n",
              "0         City  ...  [benchmark_retrofit, retrofit_build, build_lar...\n",
              "1         City  ...  [target_netzero, netzero_emiss, emiss_allow, a...\n",
              "2         City  ...  [public_facil, facil_commerci, commerci_build,...\n",
              "3         City  ...  [key_find, find_sector, sector_provid, provid_...\n",
              "4         City  ...  [public_facil, facil_industri, industri_facil,...\n",
              "\n",
              "[5 rows x 6 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 11
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Gg-dKKZWphox"
      },
      "source": [
        "Filter corpus by word frequency"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CY_oiejvphUT",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        },
        "outputId": "fd06bf7c-71c7-42a5-bc7e-d6678c6a72bf"
      },
      "source": [
        "# Use NLTK fdist to get a frequency distribution of all words\n",
        "all_words = [word for item in list(df2['Tokenized Response Answer']) for word in item]\n",
        "fdist = FreqDist(all_words)\n",
        "print(\"Number of unique words: \", len(fdist)) \n",
        "\n",
        "all_bigrams = [word for item in list(df2['Response Answer Bigrams']) for word in item]\n",
        "# use nltk fdist to get a frequency distribution of all words\n",
        "fdist2 = FreqDist(all_bigrams)\n",
        "print(\"Number of unique bigrams: \", len(fdist2)) "
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Number of unique words:  15256\n",
            "Number of unique bigrams:  176569\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "l8r44YDLpvG3",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 187
        },
        "outputId": "242c7055-a25e-4b4e-ba65-d85b916a4014"
      },
      "source": [
        "# choose k and visually inspect the bottom 10 words of the top k\n",
        "k = 50000\n",
        "top_k_words = fdist2.most_common(k)\n",
        "top_k_words[-10:]"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[('amount_conting', 1),\n",
              " ('conting_fuel', 1),\n",
              " ('requir_flight', 1),\n",
              " ('flight_air', 1),\n",
              " ('intern_air', 1),\n",
              " ('collect_achiev', 1),\n",
              " ('canada_via', 1),\n",
              " ('via_nation', 1),\n",
              " ('endors_collect', 1),\n",
              " ('collect_target', 1)]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 13
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VyOOFZNvpyzS",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 187
        },
        "outputId": "d2af808a-169f-4ac0-bf17-ead6e093ca4b"
      },
      "source": [
        "# choose k and visually inspect the bottom 10 words of the top k\n",
        "k = 6000\n",
        "top_k_words = fdist2.most_common(k)\n",
        "top_k_words[-10:]"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[('technolog_industri', 6),\n",
              " ('purchas_percent', 6),\n",
              " ('th_year', 6),\n",
              " ('annual_absolut', 6),\n",
              " ('sourc_compani', 6),\n",
              " ('help_employe', 6),\n",
              " ('offic_provid', 6),\n",
              " ('organ_around', 6),\n",
              " ('countri_worldwid', 6),\n",
              " ('two_oper', 6)]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 14
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Hv3Sl7xYq1Er"
      },
      "source": [
        "top_k_words,_ = zip(*fdist2.most_common(k))\n",
        "top_k_words = set(top_k_words)\n",
        "def keep_top_k_words(text):\n",
        "  \"\"\"Returns only the top k words.\"\"\"\n",
        "  return [word for word in text if word in top_k_words]"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UD-OhkcWq5QM"
      },
      "source": [
        "df2['Response Answer Bigrams'] = df2['Response Answer Bigrams'].apply(keep_top_k_words)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "q_f7h7nqrBIi",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 85
        },
        "outputId": "4c27d5a0-cbd8-4be8-9fb9-7a2b1f884d2a"
      },
      "source": [
        "# Analyze document length\n",
        "df2['Response Length'] = df2['Response Answer Bigrams'].apply(lambda x: len(x))\n",
        "doc_lengths = list(df2['Response Length'])\n",
        "df2.drop(labels='Response Length', axis=1, inplace=True)\n",
        "\n",
        "print(\"length of list:\",len(doc_lengths),\n",
        "      \"\\naverage document length\", np.average(doc_lengths),\n",
        "      \"\\nminimum document length\", min(doc_lengths),\n",
        "      \"\\nmaximum document length\", max(doc_lengths))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "length of list: 988 \n",
            "average document length 108.42307692307692 \n",
            "minimum document length 0 \n",
            "maximum document length 941\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4GGrpbjnrOWa",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 441
        },
        "outputId": "aef3070d-7312-48b4-b4ea-2b357994e361"
      },
      "source": [
        "# plot a histogram of document length\n",
        "num_bins = 50\n",
        "fig, ax = plt.subplots(figsize=(12,6));\n",
        "# the histogram of the data\n",
        "n, bins, patches = ax.hist(doc_lengths, num_bins)\n",
        "ax.set_xlabel('Document Length (bigrams)', fontsize=15)\n",
        "ax.set_ylabel('Normed Frequency', fontsize=15)\n",
        "ax.grid()\n",
        "ax.set_xticks(np.logspace(start=np.log10(10),stop=np.log10(1000),num=8, base=10.0))\n",
        "plt.xlim(0,1000)\n",
        "ax.plot([np.average(doc_lengths) for i in np.linspace(0.0,0.0035,100)], np.linspace(0.0,0.0035,100), '-',\n",
        "        label='average doc length')\n",
        "ax.legend()\n",
        "ax.grid()\n",
        "fig.tight_layout()\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAA1gAAAGoCAYAAABbkkSYAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOzdeXgUVd7+/7vTSSBAQiCEDkRWZRME1EHMgCBBUITIJsqAqIgLuDAQBcGFEZBFQWGE54ssOg8izIAYgmNAAkFWQcQZjTCggkQDkg5kYUkgS6d+f+RHP2ay0HSqk3R8v66L60qq6lR9KpVO9805dcpiGIYhAAAAAEC5+VR2AQAAAABQXRCwAAAAAMAkBCwAAAAAMAkBCwAAAABMQsACAAAAAJP4VnYBZisoKJDDUfrEiFarpdT1Za0ri7vt3FGRx6oo1fGcvBnXAyjEawEAqic/P6tH91/tApbDYSgzM7vU9cHBtUpdX9a6srjbzh0VeayKUh3PyZtxPYBCvBYAoHoKDQ306P4ZIggAAAAAJiFgAQAAAIBJCFgAAAAAYJIKvQcrJydHI0eOVG5urhwOh+6++26NHz9eycnJio6OVmZmptq3b68333xT/v7+ys3N1eTJk3X48GEFBwdrwYIFuu666yqyZAAAAHgJhyNfGRlnlJ+fW9mloArw9fVXvXqhslordtqJCj2av7+/Vq5cqdq1aysvL08jRoxQjx499Le//U2PPvqo+vfvr2nTpmn9+vUaMWKEPvroIwUFBWnr1q2Ki4vT/PnztXDhwoosGQAAAF4iI+OMataspdq1w2SxWCq7HFQiwzCUlXVeGRln1KBBowo9doUOEbRYLKpdu7YkKT8/X/n5+bJYLNq/f7/uvvtuSdLgwYOVkJAgSdq+fbsGDx4sSbr77ru1b98+GQZT5gIAAKC4/Pxc1a4dRLjC/587giqlN7PCp2l3OBwaMmSIfvnlF40YMUJNmjRRUFCQfH0LSwkLC5Pdbpck2e12NWpUmDh9fX0VGBiojIwM1a9fv9T9W60WBQfXKmO9T6nry1pXFnfbuaMij1VRquM5eTOuB1CI1wLgfex2i3x9PfuMI3gXi6XsbOAJFR6wrFarNm7cqPPnz+uZZ57RTz/9ZOr+eQ6W96mO5+TNuB5AIV4LgPcxDEMOR0Fll4EqxDCKZ4Nq+xysoKAgde3aVd98843Onz+v/Px8SVJKSopsNpskyWaz6fTp05IKhxReuHBB9erVq6ySAQAAgGrrvfeWas2aVeXeT58+d5hQTVE//vi99u3b4/zerFo9oUIDVnp6us6fPy9Junz5sr744gtdf/316tq1q7Zs2SJJ2rBhgyIjIyVJkZGR2rBhgyRpy5Ytuv322xlTCwAAgN8Vh8NR2SVUuh9//EH79u2t7DJcUqFDBFNTUzVlyhQ5HA4ZhqF77rlHvXr10g033KCJEydq4cKFateunYYNGyZJuv/++zVp0iT16dNHdevW1YIFCyqyXAAAAOCaTJ36vOx2u3JzczVs2HANHDhEsbHrderUKT3zzJ8lSZs2/VNHj/5H0dEvasuWTVq//h/Ky8vXjTe21/PPT5HValWfPnfovvuG6ODBA4qOflH/+tdX2rt3t3JyLqtDh06aPPklWSwWHTlyWHPnzpTF4qMuXbpq//69WrVqnRwOh959d7H+/e+vlZeXq8GDh2nQoKHF6l258j1t3hynevXqqWFDm9q0aSepsMdo3rw5ysm5rMaNr9PUqdMUFBSkkyeTNW/eHGVmZshq9dHMmW8oPLz0xyitWfOBtm/fpry8XPXo0Utjxjyl06d/1QsvjFfHjp313XeJCg0N1dy5b6lGjZolns/776/WihXvKjc3R4mJ32rUqEclSUlJP+nZZ5+U3W7XAw/8ScOGDTf/grqhQgNW27ZtFRsbW2x5kyZNtH79+mLLa9SooXfeeaciSgMAAEA1UuPoetU88g9T93m53XDltL2/zG0Kg0hd5eRc1uOPP6w774xUz569NXbsaGfASkjYqocffkxJSSeUkLBVS5a8L19fX82fP1fx8ZvVr98AXbp0STfe2EHPPTdRktSiRQuNHv2EJGnmzFe1d+9ude/eQ7NnT9eLL76iDh06asmSRc46Pv10o2rXrq0VKz5Qbm6uxo0bo9tuu12NG4c7tzl69IgSEuL1v/+7Rg5Hvh577CFnwHr99b9owoRJuvnmW7Vixbv629+W689/fl7Tp7+ihx56VD179lJOTk6ZM3wfOLBfycnJWr58pQzD0JQp0frmm3/JZgvTyZPJeu21WXrxxVf06qtTtGPHdt19970lno+fn58ef3ysM5RKhUMEf/nlZ73zzrvKzs7WiBFDNXjw/c6J8ypT5VcAAAAAVBMfffQP7dq1Q5KUmmpXcnKyOnS4SY0bh+vQoe/UpEkT/fJLkjp27KSYmHX6/vsjevzxhyVJOTmXnfMNWK1W3XlnpHO///rXQa1e/YFyci7r/Pnzat78enXqdLOys7PVoUNHSVKfPvfoiy92S5K++mq/jh07ph07tkuSsrIu6uTJ5CIBKzHx3+rRo5dq1qwpSerevYck6eLFi7pw4YJuvvlWSVK/fgP06qsvKjs7S2fPnlHPnr0kFXaGlOXAgf366qv9Gj16pCTp0qVsnTz5i2y2MDVq1FitWrWRJLVp01anT/+qCxculHo+JYmI6CZ/f3/5+/urXr16Sk9PU8OGtjJrqggELAAAAFQ7OW3vv2pvk9n+9a+DOnjwgJYu/Ztq1qypZ599Urm5OZKk3r376vPPt6pp0+bq0eNOWSwWGYahfv0GaOzYZ4vty9/fX1Zr4ZTzOTk5euutN7RixQey2cL03ntLnfstjWEYmjhxkrp2jTD/RF1kGIYeeujRYkMTT5/+VX5+fs7vfXyscjjKPp+S+Pn5/2YfPlXmXrVKm0UQAAAAqE6ysi4qMDBINWvW1M8/J+k//znkXNejRy/t3r1T27ZtUe/efSVJt956m3bsSFBGRrok6fz5c0pJOV1sv7m5hQ/LDQ4OVnZ2tnbsSJAkBQYGqlatWjp8uPA4CQnxzja33Rah2Nj1zpm6f/nlZ126dKnIfjt1ukW7d+9QTs5lZWdnae/ewt6iOnXqKDAwSN9++29J0mefxalz51tUq1ZthYY2dPbQ5ebm6vLly6X+PLp2jVBc3CfKzi6cJv3MmVTnuZakrPOpVauWcz9VHT1YAAAAgAm6dv2jYmNjNHLk/WratJluvLGDc11QUJCaN2+hEydOOJe3aNFSTzwxThMnPivDKJDV6qvo6BcVFtaoyH4DAwMVFTVIo0Y9qJCQELVr1965bsqUaXrzzddlsfioc+dbVKdOHUlSVNQgpaSc1mOPjZRhGAoOrqc5c94qst82bdoqMrKPHnlkhOrVq6e2bW90rnvlldd+M8lFuKZO/Ysk6dVXZ2jevNl67713ZbX6aubMuaVOcnHbbbcrKemExo4dLUkKCKiladNmysen9D6e0s7nllv+oA8/XKlHHx3hnOSiqrIYZd2Z5oUKCgz5+Lg3lfvlPIcu8KDhClcdz8mbcT2AQrwWAO+TkvKzwsKaVXYZFSo7O1u1atWSJK1a9b9KSzurCRNeqOSq3Gf2+ZT0O+HpBw1Xux4sHx+Lmk+Jc6tt0tz+umByPQAAAICn7Nu3R6tW/a8cjnyFhTXSSy+9VtkllUt1OJ9qF7AAAACA34vevfs67+mqDqrD+TDJBQAAAKqNanb3C8qhsn4XCFgAAACoFnx9/ZWVdZ6QBRmGoays8/L19b/6xiZjiCAAAACqhXr1QpWRcUYXL2ZWdimoAnx9/VWvXmjFH7fCjwgAAAB4gNXqqwYNGl19Q8CDGCIIAAAAACYhYAEAAACASQhYAAAAAGASAhYAAAAAmISABQAAAAAmIWABAAAAgEkIWAAAAABgEgIWAAAAAJiEgAUAAAAAJiFgAQAAAIBJCFgAAAAAYBICFgAAAACYhIAFAAAAACYhYAEAAACASQhYAAAAAGASAhYAAAAAmISABQAAAAAmIWABAAAAgEkIWAAAAABgEgIWAAAAAJiEgAUAAAAAJiFgAQAAAIBJCFgAAAAAYBICFgAAAACYhIAFAAAAACYhYAEAAACASQhYAAAAAGASAhYAAAAAmISABQAAAAAmIWABAAAAgEkIWAAAAABgEgIWAAAAAJiEgAUAAAAAJiFgAQAAAIBJCFgAAAAAYBICFgAAAACYhIAFAAAAACYhYAEAAACASQhYAAAAAGASAhYAAAAAmISABQAAAAAmqdCAdfr0aY0aNUr33nuv+vfvr5UrV0qSFi1apDvuuEMDBw7UwIEDtXPnTmebpUuXqk+fPrr77ru1e/fuiiwXAAAAAK6Jb0UezGq1asqUKWrfvr0uXryooUOHqlu3bpKkRx99VGPGjCmy/bFjxxQXF6e4uDjZ7XaNHj1aW7ZskdVqrciyAQAAAMAlFdqD1bBhQ7Vv316SVKdOHbVs2VJ2u73U7RMSEtS/f3/5+/urSZMmatasmRITEyuqXAAAAAC4JpV2D9bJkyd15MgRderUSZK0evVqRUVFaerUqTp37pwkyW63KywszNnGZrOVGcgAAAAAoDJV6BDBK7KysjR+/Hi99NJLqlOnjv70pz/p6aeflsVi0V//+lfNnTtXc+bMqYzSFBxc65rbWK0+brVzR0Ueq6JUx3PyZlwPoBCvBQCAOyo8YOXl5Wn8+PGKiopS3759JUkNGjRwrh82bJjGjh0rqbDHKiUlxbnObrfLZrN5tL7MzOxrbhMcXMutdu6oyGNVlOp4Tt6M6wEU4rUAANVTaGigR/dfoUMEDcPQyy+/rJYtW2r06NHO5ampqc6vt23bplatWkmSIiMjFRcXp9zcXCUnJyspKUkdO3asyJIBAAAAwGUV2oP19ddfa+PGjWrdurUGDhwoSYqOjtann36qo0ePSpLCw8M1Y8YMSVKrVq3Ur18/3XvvvbJarZo2bRozCAIAAACosiyGYRiVXYTZmk+Jc6td0tz+OnPmwjW3Y4hg+VTHc/JmXA+gEK8FAKieqtUQQQAAAACozghYAAAAAGASAhYAAAAAmISABQAAAAAmIWABAAAAgEkIWAAAAABgEgIWAAAAAJiEgAUAAAAAJiFgAQAAAIBJCFgAAAAAYBICFgAAAACYhIAFAAAAACYhYAEAAACASQhYAAAAAGASAhYAAAAAmISABQAAAAAmIWABAAAAgEkIWAAAAABgEgIWAAAAAJiEgAUAAAAAJiFgAQAAAIBJCFgAAAAAYBICFgAAAACYhIAFAAAAACYhYAEAAACASQhYAAAAAGASAhYAAAAAmISABQAAAAAmIWABAAAAgEkIWAAAAABgEgIWAAAAAJiEgAUAAAAAJiFgAQAAAIBJCFgAAAAAYBICFgAAAACYhIAFAAAAACYhYAEAAACASQhYAAAAAGASAhYAAAAAmISABQAAAAAmIWABAAAAgEkIWAAAAABgEgIWAAAAAJjEpYDlcDg8XQcAAAAAeD2XAlaPHj305ptv6vjx456uBwAAAAC8lksBa/jw4dqyZYsGDBigYcOGae3atbp48aKnawMAAAAAr+JSwHruueeUkJCg999/Xy1atNDcuXPVvXt3Pf/88/riiy88XSMAAAAAeAXfa9k4IiJCERERysrK0ubNm7VmzRqNGTNGjRo10uDBg/XAAw/IZrN5qlYAAAAAqNLcmkXw0KFDOnjwoE6cOKGgoCDdeuut+uijj9S3b19t3LjR7BoBAAAAwCu43IN16tQpbdiwQbGxsTp16pT++Mc/atasWbrrrrvk7+8vh8OhN954Q/PmzdPAgQM9WTMAAAAAVEkuBaxRo0bp66+/ls1m05AhQzRkyBCFh4cX2cZqtWrAgAH64IMPPFIoAAAAAFR1LgWskJAQLVu2TN26dZPFYil1u3bt2ikhIcG04gAAAADAm7gUsBYuXOjSzvz8/Ir1bP3W6dOnNXnyZKWlpcliseiBBx7QI488oszMTE2cOFGnTp1SeHi4Fi5cqLp168owDM2aNUs7d+5UzZo1NXfuXLVv3961MwMAAACACubSJBcffPCB5s+fX+K6t956Sx9++KFLB7NarZoyZYo2bdqktWvXas2aNTp27JiWLVumiIgIxcfHKyIiQsuWLZMk7dq1S0lJSYqPj9fMmTP12muvuXZWAAAAAFAJXApYa9asUbNmzUpc17x5c61Zs8algzVs2NDZA1WnTh21bNlSdrtdCQkJGjRokCRp0KBB2rZtmyQ5l1ssFnXu3Fnnz59XamqqS8cCAAAAgIrmUsD69ddf1bRp0xLXXXfddTp16tQ1H/jkyZM6cuSIOnXqpLS0NDVs2FCSFBoaqrS0NEmS3W5XWFiYs01YWJjsdvs1HwsAAAAAKoJL92AFBQXpxIkT6tq1a7F1J06cUJ06da7poFlZWRo/frxeeumlYm0tFkuZE2l4WnBwrWtuY7X6uNXOHRV5rIpSHc/Jm3E9gEK8FgAA7nApYEVGRmrx4sW6+eab1aZNG+fyH374QYsXL1bv3r1dPmBeXp7Gjx+vqKgo9e3bV1LhLIWpqalq2LChUlNTVb9+fUmSzWZTSkqKs21KSopsNpvLx3JHZmb2NbcJDq7lVjt3VOSxKkp1PCdvxvUACvFaAIDqKTQ00KP7d2mIYHR0tOrVq6fBgwdr6NChGjt2rIYOHapBgwYpJCREzz//vEsHMwxDL7/8slq2bKnRo0c7l0dGRio2NlaSFBsb6wxsV5YbhqFvvvlGgYGBzqGEAAAAAFDVuNSDFRwcrPXr12vDhg368ssvlZmZqaZNm+rBBx/UoEGD5O/v79LBvv76a23cuFGtW7fWwIEDJRWGtyeffFITJkzQ+vXr1bhxY+e08D179tTOnTvVp08fBQQEaPbs2W6eJgAAAAB4nsUwDKOyizBb8ylxbrVLmttfZ85cuOZ2DBEsn+p4Tt6M6wEU4rUAANWTp4cIutSD9VsOh0O5ubnFlgcEBJhSEAAAAAB4K5cC1sWLF/X2228rPj5e6enpKqnT68iRI6YXBwAAAADexKWANW3aNH3++ecaNmyYbrjhBvn5+Xm6LgAAAADwOi4FrD179uill17SsGHDPF0PAAAAAHgtl6ZpDwgI8PjzpwAAAADA27kUsEaPHq01a9aooKDA0/UAAAAAgNdyaYig3W7X999/r3vuuUddu3ZVYGDRqQ0tFosmTZrkkQIBAAAAwFu4FLC2bNkii8Wi/Px87d27t9h6AhYAAAAAuBiwtm/f7uk6AAAAAMDruXQPFgAAAADg6lwOWEePHtWECRN01113qUOHDjp8+LAkacGCBdq5c6fHCgQAAAAAb+FSwNq5c6eGDh2qs2fPatCgQcrPz3eu8/Pz04cffuixAgEAAADAW7gUsN5++20NHjxYH374ocaOHVtkXbt27XTkyBGPFAcAAAAA3sSlgPXTTz/p3nvvlVQ4Y+Bv1alTR+fOnTO/MgAAAADwMi4FrJCQECUnJ5e47tixY2rcuLGpRQEAAACAN3IpYN1777165513dPDgQecyi8WiEydOaPny5YqKivJYgQAAAADgLVx6DtaECRN0/PhxjRo1Sg0aNJAkPf300zp79qy6deump556yqNFAgAAAIA3cClg+fv7a+nSpdq3b5/27dunjIwM1a1bVxEREerWrZunawQAAAAAr+BSwLoiIiJCERERnqoFAAAAALyaSwHr2LFjV93mhhtuKHcxAAAAAODNXApYAwYMKDY9+3/jWVgAAAAAfu9cClgffPBBsWXnz5/X7t27tWfPHr3yyiumFwYAAAAA3salgHXbbbeVuPyuu+7SggULtHnzZvXq1cvUwgAAAADA27j0HKyy3H777UpISDCjFgAAAADwauUOWDt27FBQUJAZtQAAAACAV3NpiOCf//znYsvy8vL0008/6eeff9bEiRNNLwwAAAAAvI1LASs9Pb3Ysho1augPf/iDpk6dqp49e5peGAAAAAB4G5cC1qpVqzxdBwAAAAB4vXLfgwUAAAAAKORSD9bUqVNd3qHFYtHs2bPdLggAAAAAvJVLAeuHH35QSkqK0tLSFBISovr16ys9PV1paWmqX7++GjVq5NzWYrF4rFgAAAAAqMpcCljPPPOMZs+erTVr1uiWW25xLv/66681ZcoUPf300+rdu7fHigQAAAAAb+DSPVhvvfWWJkyYUCRcSdKtt96q8ePHa/78+R4pDgAAAAC8iUsBKzk5WTVr1ixxXUBAgE6dOmVqUQAAAADgjVwKWDfeeKMWL16s1NTUIsvtdrsWLVqk9u3be6Q4AAAAAPAmLt2DNXPmTD322GOKjIxUhw4dnJNcHDp0SMHBwZo3b56n6wQAAACAKs+lgNWqVStt27ZN69ev16FDh3TmzBm1aNFCUVFRGjp0aKnDBwEAAADg98SlgCVJNWrU0MiRIz1ZCwAAAAB4NZcDliTt3LlThw4dUkpKisaNG6fGjRvrq6++UtOmTWWz2TxVIwAAAAB4BZcC1tmzZzVu3DgdPnxY4eHhOnnypIYPH67GjRsrJiZG/v7+mj59uqdrBQAAAIAqzaVZBGfOnKns7Gxt3rxZmzdvlmEYznURERHav3+/xwoEAAAAAG/hUsDavXu3JkyYoGbNmslisRRZFxYWJrvd7pHiAAAAAMCbuBSwJMlqtZa4PCMjQzVq1DCtIAAAAADwVi4FrD/84Q9atWqVHA6Hc9mVnqyPP/5Yt99+u2eqAwAAAAAv4tIkFy+88IJGjBihAQMGqE+fPrJYLFq3bp2OHTumH374QWvXrvV0nQAAAABQ5bnUg9W6dWvFxMSoQ4cOiomJkdVq1datW2Wz2bRu3Tq1aNHC03UCAAAAQJV31R6s3Nxcvffee+rVq5fmzZtXETUBAAAAgFe6ag+Wv7+/3n33XZ0/f74i6gEAAAAAr+XSEMGOHTvqP//5j6drAQAAAACv5tIkF5MmTdILL7wgX19f9ezZUyEhIcWehxUQEOCRAgEAAADAW7gUsB544AFJ0uuvv65Zs2aVuM2RI0fMqwoAAAAAvJBLAWv27NnFeqwAAAAAAEWVGrDeffddDRkyRA0bNtSQIUMkSQUFBfLxcem2rRJNnTpVO3bsUEhIiD799FNJ0qJFi7Ru3TrVr19fkhQdHa2ePXtKkpYuXar169fLx8dHr7zyiu644w63jw0AAAAAnlZqWvrrX/+qlJQU5/cOh0Pt27fX4cOH3T7YkCFDtGLFimLLH330UW3cuFEbN250hqtjx44pLi5OcXFxWrFihaZPny6Hw+H2sQEAAADA00oNWIZhuLTsWnTp0kV169Z1aduEhAT1799f/v7+atKkiZo1a6bExMRyHR8AAAAAPMn98X4mWr16taKiojR16lSdO3dOkmS32xUWFubcxmazyW63V1aJAAAAAHBVLk1y8VtmT3bxpz/9SU8//bQsFov++te/au7cuZozZ46px7gWwcG1rrmN1erjVjt3VOSxKkp1PCdvxvUACvFaAAC4o8yA9fbbbzuH9F0ZHjhv3jwFBQUV2c5isWjhwoVuFdCgQQPn18OGDdPYsWMlFfZY/fYeMLvdLpvN5tYxrkVmZvY1twkOruVWO3dU5LEqSnU8J2/G9QAK8VoAgOopNDTQo/svdYhgly5d5HA4lJ6ervT0dGVkZKhLly7Kz893LrvyLy0tze0CUlNTnV9v27ZNrVq1kiRFRkYqLi5Oubm5Sk5OVlJSkjp27Oj2cQAAAADA00rtwVq1apXpB4uOjtaBAweUkZGhHj166LnnntOBAwd09OhRSVJ4eLhmzJghSWrVqpX69eune++9V1arVdOmTZPVajW9JgAAAAAwi8Uo79SAVVDzKXFutUua219nzly45nYMESyf6nhO3ozrARTitQAA1VOlDREEAAAAAFwbAhYAAAAAmISABQAAAAAmIWABAAAAgEkIWAAAAABgklKnaZ86deo17WjOnDnlLgYAAAAAvFmpAeuHH34o8v3p06eVnp6ukJAQ1a9f3/mA4fr166tx48YeLxQAAAAAqrpSA9bHH3/s/Hr79u2aPXu2Fi9erFtuucW5/Ouvv9aUKVM0btw4z1YJAAAAAF7ApXuw3nrrLU2YMKFIuJKkW2+9VePHj9f8+fM9UhwAAAAAeBOXAlZycrJq1qxZ4rqAgACdOnXK1KIAAAAAwBu5FLBuvPFGLV68WKmpqUWW2+12LVq0SO3bt/dIcQAAAADgTUq9B+u3Zs6cqccee0yRkZHq0KGDc5KLQ4cOKTg4WPPmzfN0nQAAAABQ5bkUsFq1aqVt27Zp/fr1OnTokM6cOaMWLVooKipKQ4cOLXX4IAAAAAD8nrgUsCSpRo0aGjlypCdrAQAAAACv5nLAkqSdO3fq0KFDSklJ0bhx49S4cWN99dVXatq0qWw2m6dqBAAAAACv4FLAOnv2rMaNG6fDhw8rPDxcJ0+e1PDhw9W4cWPFxMTI399f06dP93StAAAAAFCluTSL4MyZM5Wdna3Nmzdr8+bNMgzDuS4iIkL79+/3WIEAAAAA4C1cCli7d+/WhAkT1KxZM1ksliLrwsLCZLfbPVIcAAAAAHgTlwKWJFmt1hKXZ2RkqEaNGqYVBAAAAADeyqWA9Yc//EGrVq2Sw+FwLrvSk/Xxxx/r9ttv90x1AAAAAOBFXJrk4oUXXtCIESM0YMAA9enTRxaLRevWrdOxY8f0ww8/aO3atZ6uEwAAAACqPJd6sFq3bq2YmBh16NBBMTExslqt2rp1q2w2m9atW6cWLVp4uk4AAAAAqPJcfg5W06ZNNW/ePE/WAgAAAABezeVJLgAAAAAAZXO5B2vTpk3atm2b7Ha7cnJyiq1fv369qYUBAAAAgLdxKWDNnz9fK1as0E033aSmTZvK39/f03UBAAAAgNdxKWB9/PHHmjhxop566ilP1wMAAAAAXsule7B8fX3Vvn17T9cCAAAAAF7NpYD18MMP66OPPpJhGJ6uBwAAAAC8lktDBJ944gm98cYbuueee3TbbbcpMDCwyHqLxaJJkyZ5pEAAAAAA8BYuBaxPPvlEK1eulI+Pj3bs2CE/P78i6wlYAAAAAOBiwCApissAACAASURBVHrrrbfUr18/TZ8+XXXq1PF0TQAAAADglVy6B+vixYu6//77CVcAAAAAUAaXAlbfvn21f/9+T9cCAAAAAF7NpSGCd9xxh+bPn6+zZ8/q9ttvV1BQULFtevbsaXpxAAAAAOBNXApY0dHRkgofOPzxxx8XW2+xWHTkyBFzKwMAAAAAL+NSwEpISPB0HQAAAADg9a4asHJycvTqq6/qqaeeUteuXSuiJgAAAADwSled5KJGjRr67rvvVFBQUBH1AAAAAIDXcmkWwcjISG3bts3TtQAAAACAV3PpHqzu3bvrzTff1JkzZ9SjRw81aNBAFoulyDbMIggAAADg986lgDVp0iRJUnx8vOLj44utZxZBAAAAAGAWQQAAAAAwjUsBKzw83NN1AAAAAIDXcylgSVJ+fr7i4+P19ddfKzMzU8HBwbr11lvVt29f+fq6vBsAAAAAqLZcSkZpaWl67LHH9P333ys8PFwNGjTQN998o9WrV6tt27Z6//33Vb9+fU/XCgAAAABVmksBa86cOcrMzNS6devUsWNH5/LExESNHz9ec+bM0bx58zxWJAAAAAB4A5eeg7Vr1y698MILRcKVJHXs2FHR0dHauXOnR4oDAAAAAG/iUsDKzc1V7dq1S1xXu3Zt5eXlmVoUAAAAAHgjlwJWp06dtHz5cmVnZxdZnp2dreXLl6tTp04eKQ4AAAAAvIlL92BNmTJFDz/8sO68805169ZNISEhSk9P1549e2QYhlatWuXpOgEAAACgynOpB6tdu3basmWLHnjgAaWnp+uLL75QWlqahg8fri1btqht27aerhMAAAAAqjyXH2BVv359vfDCC+U62NSpU7Vjxw6FhITo008/lSRlZmZq4sSJOnXqlMLDw7Vw4ULVrVtXhmFo1qxZ2rlzp2rWrKm5c+eqffv25To+AAAAAHiSSz1YZhkyZIhWrFhRZNmyZcsUERGh+Ph4RUREaNmyZZIKZy5MSkpSfHy8Zs6cqddee60iSwUAAACAa1ZqD9bDDz/s8k4sFotWrlx51e26dOmikydPFlmWkJDgvIdr0KBBGjVqlCZNmqSEhAQNGjRIFotFnTt31vnz55WamqqGDRu6XBcAAAAAVKRSA1ZwcPBVG585c0b//ve/ZbFY3C4gLS3NGZpCQ0OVlpYmSbLb7QoLC3NuFxYWJrvdTsACAAAAUGWVGrDeeeedUhv9+uuvWr58uXbs2KF69erp0UcfNaUYi8VSrrBmhuDgWtfcxmr1caudOyryWBWlOp6TN+N6AIV4LQAA3OHyJBeS9PPPP2vp0qX65JNPFBISoujoaA0fPlw1a9Z0u4CQkBDn0L/U1FTVr19fkmSz2ZSSkuLcLiUlRTabze3juCozM/vqG/2X4OBabrVzR0Ueq6JUx3PyZlwPoBCvBQConkJDAz26f5cmufjxxx/1/PPP695779WXX36pl19+WVu3btWjjz5arnAlSZGRkYqNjZUkxcbGqnfv3kWWG4ahb775RoGBgQwPBAAAAFClldmDdejQIb377rtKSEhQs2bN9Prrr+u+++6T1Wp162DR0dE6cOCAMjIy1KNHDz333HN68sknNWHCBK1fv16NGzfWwoULJUk9e/bUzp071adPHwUEBGj27NluHRMAAAAAKorFMAyjpBWPP/649u7dq9atW2vs2LHq169fRdfmtuZT4txqlzS3v86cuXDN7RgiWD7V8Zy8GdcDKMRrAQCqJ08PESw1YLVt21aSVLduXfn4XH0k4b59+8ytrBzcDVhHZ96jmn7u9c5dysnXxfOX3Gp7LarjG351PCdvxvUACvFaAIDqydMBq9Qhgs8++6xHD1wV1fSzlqv366LJ9QAAAADwLgQsAAAAADCJS7MIAgAAAACujoAFAAAAACYhYAEAAACASQhYAAAAAGASAhYAAAAAmISABQAAAAAmIWABAAAAgEkIWAAAAABgEgIWAAAAAJiEgAUAAAAAJiFgAQAAAIBJCFgAAAAAYBICFgAAAACYhIAFAAAAACYhYAEAAACASQhYAAAAAGASAhYAAAAAmISABQAAAAAmIWABAAAAgEl8K7uA6uJynkOhoYHX3O5STr4unr/kgYoAAAAAVDQClklq+lnVfErcNbdLmttfFz1QDwAAAICKxxBBAAAAADAJAQsAAAAATELAAgAAAACTELAAAAAAwCQELAAAAAAwCQELAAAAAExCwAIAAAAAkxCwAAAAAMAkBCwAAAAAMAkBCwAAAABMQsACAAAAAJMQsAAAAADAJAQsAAAAADAJAQsAAAAATELAAgAAAACTELAAAAAAwCQELAAAAAAwiW9lFwD31QkKUEAN9y7hpZx8XTx/yeSKAAAAgN83ApYXC6jhq+ZT4txqmzS3vy6aXA8AAADwe8cQQQAAAAAwCQELAAAAAExCwAIAAAAAkxCwAAAAAMAkTHLxO3U5z6HQ0MBrbsfsgwAAAEDpCFi/UzX9rG7NQMjsgwAAAEDpGCIIAAAAACahB6uSuTNUz52hfQAAAAA8j4BVydwdqicVDtcDAAAAUHUwRBAAAAAATFJlerAiIyNVu3Zt+fj4yGq1KiYmRpmZmZo4caJOnTql8PBwLVy4UHXr1q3sUgEAAACgRFWqB2vlypXauHGjYmJiJEnLli1TRESE4uPjFRERoWXLllVyhQAAAABQuioVsP5bQkKCBg0aJEkaNGiQtm3bVskVAQAAAEDpqswQQUkaM2aMLBaLHnzwQT344INKS0tTw4YNJUmhoaFKS0ur5AohScHBtUzdn9XqY/o+4T6uB1CI1wIAwB1VJmD9/e9/l81mU1pamkaPHq2WLVsWWW+xWGSxWCqpOvxWZma2qfsLDq5l+j7hPq4HUIjXAgBUT55+5FGVGSJos9kkSSEhIerTp48SExMVEhKi1NRUSVJqaqrq169fmSUCAAAAQJmqRA9Wdna2CgoKVKdOHWVnZ2vv3r16+umnFRkZqdjYWD355JOKjY1V7969K7tUlEOdoAAF1Cj5V+5q/5NwKSdfF89f8kRZAAAAgGmqRMBKS0vTM888I0lyOBwaMGCAevTooZtuukkTJkzQ+vXr1bhxYy1cuLCSK0V5BNTwLddDlS+aXA8AAABgtioRsJo0aaJPPvmk2PJ69epp5cqVlVARAAAAAFy7KhGw4D0u5zk8fmMgAAAA4K0IWLgmNf2s5RrmBwAAAFRnVWYWQQAAAADwdgQsAAAAADAJAQsAAAAATELAAgAAAACTELAAAAAAwCQELAAAAAAwCQELAAAAAExCwAIAAAAAkxCwAAAAAMAkBCwAAAAAMIlvZRcAuOJynkOhoYHX3O5STr4unr/kgYoAAACA4ghY8Ao1/axqPiXumtslze2vix6oBwAAACgJQwQBAAAAwCQELAAAAAAwCQELAAAAAEzCPVhAFVInKEABNdx7WTKhBwAAQOUjYAFVSEANX7cm85CY0AMAAKAqYIggAAAAAJiEgAUAAAAAJiFgAQAAAIBJuAcL1drlPIdCQwPdasukEQAAALhWBCxUazX9rEwaAQAAgArDEEEAAAAAMAkBCwAAAABMQsACAAAAAJNwDxbgAXWCAhRQg5cXAADA7w2fAAEPCKjh69bkGklz+3ugGgAAAFQUhggCAAAAgEnowQKqCZ75BQAAUPkIWEA1wTO/AAAAKh9DBAEAAADAJPRgAaUoz5A7AAAA/D4RsIBSlHfIHQAAAH5/CFgAivXWudpzx+QYAAAARRGwALjdW8fkGAAAAEUxyQUAAAAAmISABQAAAAAmIWABAAAAgEm4BwuA16kTFKCAGtf+54tJOQAAgKcRsABUCndD0hVMygEAAKoiAhaAShFQw5fnjAEAgGqHe7AAAAAAwCQELAAAAAAwCUMEAbjtcp5DoaGBlV0GAABAlUHAAuC2mn5Wr7qPqjyBkBkIAQCAKwhYAH43yhMIj868x6vCWXlmaSRMAgDgPgIWALigvL11FT09fHlnaWQ6ewAA3EPAAgAPK8/QxMt5DtX0s5pcEQAA8BQCFgB4WHl7v9x9qDIAAKh4BCwAQBGV0ePmbfd9cY8bAKA0XhGwdu3apVmzZqmgoEDDhg3Tk08+WdklAUC1VRk9bpU1icjVglJZNVXGPW7uBrvyDDVlkhYAuDZVPmA5HA7NmDFDf/vb32Sz2XT//fcrMjJSN9xwQ2WXBgAwSWVNIuLuZCCVNQSzPPV606QnTNICwJtV+YCVmJioZs2aqUmTJpKk/v37KyEhgYAFAEAVV56eKHf9np535+7Pt7LOszLq9bbe0Mqq19t+l6o6i2EYRmUXUZbPPvtMu3fv1qxZsyRJsbGxSkxM1LRp0yq5MgAAAAAoyqeyCwAAAACA6qLKByybzaaUlBTn93a7XTabrRIrAgAAAICSVfmAddNNNykpKUnJycnKzc1VXFycIiMjK7ssAAAAACimyk9y4evrq2nTpunxxx+Xw+HQ0KFD1apVq8ouCwAAAACKqfKTXAAAAACAt6jyQwQBAAAAwFsQsAAAAADAJFX+HixXTZ06VTt27FBISIg+/fRTSVJmZqYmTpyob7/9Vjk5OWratKk2b94sSTp69Kj+8pe/KDs7Wz4+PsrKypKPj4/GjRunwYMHS5LOnz+vHj16qKCgQIZhqGPHjlq9erVGjBihrKwsSVJaWpo6duyo//f//l+Rek6dOqVnn31WBQUFys/P10MPPaQ//elPpdb/008/aeLEic7vk5OTNX78eNntdn3++efy8/NT06ZNNWfOHAUFBRVrv2vXLs2aNUsFBQUaNmyYnnzyyfL9QN1Q0jVYuHChEhIS5OPjo5CQEM2ZM0c2m00XLlzQpEmT9Ouvv8rhcOixxx7T0KFDi+1z06ZNWrJkiQoKCnTnnXdq0qRJFX1aXu306dOaPHmy0tLSZLFY9MADD+iRRx6RJK1atUqrV6+W1WpVz549NXnyZOXl5emVV17Rf/7zH+Xn52vQoEF66qmniu33pZde0qFDh2QYhlq0aKE5c+aodu3aFX16wDWJjIxU7dq15ePjI6vVqpiYGG3evFmLFy/W8ePH9dFHH+mmm26SJH3yySd67733nG2///57bdiwQe3atSuyz9++l4SHh2v+/PmqU6dOhZ4XAPzelPW5/9SpUwoPD9fChQtVt25dGYahWbNmaefOnapZs6bmzp2r9u3bS5I2bNigJUuWSFKRDPBbpe23TEY1ceDAAePQoUNG//79ncveeOMNY+nSpcaBAweMGTNmGF27dnWuGzJkiPHll18aGRkZRteuXY05c+YYmZmZRmRkpJGZmWkYhmE4HA4jNTXVMAzDyMrKMjp16mSsXbu2yHGfffZZY8OGDcXqycnJMXJycgzDMIyLFy8avXr1MlJSUlw6l/z8fOOPf/yjcfLkSWP37t1GXl6eYRiG8eabbxpvvvlmidv37t3b+OWXX4ycnBwjKirK+PHHH106lplKugYXLlxwfr1y5Urj1VdfNQzDMJYsWeI8l7S0NKNLly7On9cV6enpRs+ePY20tDTDMAxj8uTJxhdffOHp06hW7Ha7cejQIcMwCq9F3759jR9//NHYt2+f8cgjjzh/5mfPnjUMwzA++eQTY8KECYZhGEZ2drbRq1cvIzk5udh+f3tdZ8+ebSxdutTTpwKUW69evZx/T644duyYcfz4ceOhhx4yEhMTS2x39OhRo3fv3iWuu/JeYhiG8dFHHxkLFiwwt2gAQDFlfe43DMNYunSp83Pmjh07jDFjxhgFBQXGv//9b+P+++83DMMwMjIyjMjISCMjI6NYBvit0vZblmozRLBLly7F0mRCQoIGDRqkLl266L777tPFixed65KSktSlSxft2bNH3bt3165du1S3bl1169ZNu3fvliT5+PgoNDRUknT58mUZhiEfn//7kV28eFH79+/XXXfdVawef39/+fv7S5Jyc3NVUFDg8rns27dPTZo0UXh4uLp37y5f38KOxs6dOxd5JtgViYmJatasmZo0aSJ/f3/1799fCQkJLh/PLCVdg9/+T+6lS5dksVgkSRaLRVlZWTIMQ1lZWapbt67zPK9ITk5Ws2bNVL9+fUlSRESEtmzZ4uGzqF4aNmzo/F+aOnXqqGXLlrLb7fr73/+uJ5980vk7GhISIqnwuly6dEn5+fm6fPmy/Pz8Svzf+CvLDMPQ5cuXK+hsAPNdf/31atmyZZnbxMXFqX///iWuu/JeIkndunVTfHy86TUCAIoq63O/JA0aNEjbtm0rstxisahz5846f/68UlNTtWfPHnXr1k3BwcHFMoAr+y1LtQlYJUlLS1PDhg0lSfXr15fD4XCua9WqlRISEmS325WZmanTp09LKnywsd1ud26Xm5urjh07KiIiQq1bt9b999/vXLdt2zZFRESUOhzk9OnTioqK0p133qknnnjC5Qckx8XFacCAAcWWf/zxx+rRo0ex5Xa7XWFhYc7v//scKtuCBQvUs2dP/fOf/9Sf//xnSdLIkSN1/Phx3XHHHbrvvvv08ssvFwmvktSsWTOdOHFCJ0+eVH5+vhISEkoMmHDNyZMndeTIEXXq1ElJSUk6ePCghg0bpoceekiJiYmSpLvvvlsBAQHq3r27evXqpccee0zBwcEl7m/q1Knq1q2bfvrpJ40aNaoiTwVw25gxYzRkyBCtXbvW5TabNm0qNWBdeS+RpM8++8z5XgIAqFi//dwfGhqqtLQ0ScU/J4eFhclut7v8+bm0/ZalWges37rSc3LFrFmztGbNGq1cuVK5ubnO/8n/b/7+/kpMTNT27dt14sQJbd261bnu008/LfVNV5IaNWqkf/7zn4qPj9eGDRt09uzZq9aZm5ur7du365577imyfMmSJbJarbrvvvuuuo+qZuLEidq5c6eioqL04YcfSpL27Nmjdu3aaffu3YqNjdWMGTOK9DBKUt26dfXaa69p4sSJGjlypMLDw4uFMLgmKytL48eP10svvaQ6derI4XDo3LlzWrdunSZPnqwJEybIMAwlJibKx8dHu3fvVkJCgt5//30lJyeXuM85c+Zo9+7duv7667Vp06YKPiPg2v3973/Xhg0btHz5cq1evVpfffXVVdt8++23CggIUOvWrUtcf+W9ZMiQIcrKyir1vQQAUHEsFkuxz/4Vud9q/Wk1JCREqampkgrTp9Vqda67/vrr9f7772vy5MkKCQlRkyZNJBWm3JJ6msLDw3XjjTdq/fr1kqT09HR99913uvPOO69ah81mU6tWrXTw4MGrbrtr1y61b99eDRo0cC6LiYnRjh07NH/+/BIvqs1mK9KzU9o5VLaoqCjn8JmYmBj17dtXFotFzZo103XXXaeffvqpWJvIyEh99NFHWrt2rVq0aKHmzZtXcNXeLy8vT+PHj1dUVJT69u0rqfB3pk+fPrJYLOrYsaN8fHyUkZGhTz/9VHfccYf8/PwUEhKiW265Rd99912p+7Zarerfvz/DouAVrvxdDAkJUZ8+fZw9t2Upa3ig9H/vJTExMerfv7/zvQQAULF++7k/NTXVeYvJf39OTklJkc1mc/nzc2n7LUu1DliRkZGKjY2VJMXHxxeZ5exK994f//hHff755xo4cKDOnTvnvCdLko4fP65Tp05JKpxB5PDhw7rxxhslSVu2bNGdd96pGjVqlHjslJQU570p586d07/+9S+1aNHiqjX/95v5rl27tGLFCi1ZskQBAQEltrnpppuUlJSk5ORk5ebmKi4uTpGRkVc9VkVISkpyfp2QkOC816FRo0bat2+fJOns2bM6ceKErrvuumLtr1ync+fOac2aNRo2bJjni65GDMPQyy+/rJYtW2r06NHO5XfddZe+/PJLSdKJEyeUl5enevXqqVGjRs7l2dnZ+vbbb4vdn2IYhn7++Wfn19u3b7/qPSxAZcvOznb2kmdnZ2vv3r1q1apVmW0KCgq0efPmMgPWlb9RBQUFWrJkiYYPH25e0QAAl/32c39sbKx69+5dZLlhGPrmm28UGBiohg0bqnv37tqzZ4/OnTtXLAO4st+yWAzDMEw8t0oTHR2tAwcOKCMjQyEhIXruued01113acKECUpMTFReXp4KCgrUoEEDDRkyRF988YXOnTsnSWratKlOnDghi8WiESNG6IsvvtDy5cv12WefafLkybryI7r55pv1wQcfSJJGjRqlJ554osg9Ud99953+8Y9/aNasWdq7d6/mzp0ri8UiwzD00EMP6cEHHyzzHLKzs9WrVy9t27ZNgYGBkqQ+ffooNzfXeR9Mp06dNGPGDNntdr3yyitavny5JGnnzp2aPXu2HA6Hhg4dqnHjxpn7A3ZBSddg165dzp9teHi4pk+f7hzjOnXqVJ05c0aGYeiJJ57QwIEDJUkDBw7Uxo0bnfs8evSoJOmZZ54p84MOijt48KBGjhyp1q1bO4dXRkdHKyIiQi+99JKOHj0qPz8/TZ48WREREcrKytLUqVN1/PhxGYahIUOG6PHHH5ckPfHEE3r99dcVGhrqfFSBYRhq06aNpk+fztTUqNKSk5P1zDPPSJIcDocGDBigcePGaevWrZo5c6bS09MVFBSkdu3aOadn//LLL/XWW29p3bp1Rfb18ssva/jw4brpppu0cuVKrVmzRlLh3+vnn3/eI8NSAAD/p6zP/adPn1bjxo21cOFCBQcHyzAMzZgxQ7t371ZAQIBmz57tfCTH+vXrtXTpUknS2LFjnY8M+u3f+YyMjBL3W5ZqE7AAAAAAoLJV6yGCAAAAAFCRCFgAAAAAYBICFgAAAACYhIAFAAAAACYhYAEAAACASQhYAFAOixYtUps2bdSmTRu1bdtWXbp00dChQ7VgwQKdOXOmssurdJs2bVJMTIxL20ZGRuqNN97wcEWuS0xM1KJFi4otX7Rokbp27er2fuPj49WnTx85HA5JhQ9eb9OmjbKysspsN2rUKI0fP97t41YlhmEoKirK+WwZAKhOCFgAUE6BgYFau3at/vGPf+jtt99W3759tXHjRkVFRenQoUOVXV6l+uyzz7Rhw4bKLsMtiYmJWrx4san7LCgo0DvvvKMxY8bIarVeU9u//OUvio6ONrWeymKxWDRmzBj9z//8j/Lz8yu7HAAwFQELAMrJarWqc+fO6ty5s+644w499dRT+uSTTxQaGqro6GhnTwWwb98+/fLLLxowYMA1t73hhhvUvHnzctdw+fLlcu/DDP369dPZs2e1a9euyi4FAExFwAIADwgKCtKkSZP0888/a+/evc7l6enpevHFF9W1a1d16vT/tXfnQVVWbwDHv1AIIgoIiAWihqGj7AqBILKE5r4NOqBYuYtEoaWouZYMrtclNMTdUlm0LBtyYTHLJE1TQ2fIcBm1UFAQ0IsC9/eHc98fl8VEcX7+7PnMMHPvfc973uece/+4D+d5z3UhPDycs2fP1jo/OTmZAQMG4OTkRPfu3YmKiqKkpASou1QsOzubjh07kpubC8DVq1fp2LEj3333HTNnzsTd3R0/Pz/27t0LQGJiIr6+vnh5ebF06VKqqqp0+svNzWXChAm4ubnh5uZGVFSUTsmj9nrZ2dlERUXh5uZGUFAQX375pdImJiaG/fv388svvyhllHWV3DVESkoK/fr1w9HRkYCAABITE3WOx8TEMHToUH766ScGDBiAq6sroaGh/PHHHzrtiouLiY6OxtXVFV9fX9avX8/ixYsJDAwEHpbtffLJJwBK7OHh4Tp9nDt3juHDh+Pi4sLgwYM5ceLEP8b/1Vdf4ePjg4mJSa1jf/75J2FhYTg7O9O7d28OHjyoc7yu9z0tLY1evXrh7OxMeHg4586do2PHjjplmYGBgcTFxREfH4+fnx9du3YF4NSpU0yaNAlfX19cXV0ZNGgQ33zzjU7/2vLFnJwcwsPDcXFxYdCgQeTk5HD37l1mzpxJ165dCQoKYt++fTrnnjhxgrCwMNzd3XF3d2fQoEGkpaUpxw0NDenZs6eUCQohXjgv/68DEEKIF9Ubb7zByy+/zOnTp/Hz8wNgypQpXLlyhenTp2Nubs7GjRsZPXo0X3/9NW3btgVg7dq1rF69mrCwMD766CPUajVZWVncvXuX5s2bNyiGZcuWMWDAANasWcPu3buJiYnh/PnzXL9+ndjYWHJycli5ciWdO3emX79+AFy+fJnQ0FAcHR1ZunQplZWVrFq1ikmTJpGamoqenp7S/5w5cxg8eDAjRoxg3759LFy4ECcnJ5ydnYmIiOD69euUlJQwb948AFq3bv3E87lhwwZUKhXjxo3D09OTnJwcVq1aRdOmTRk1apTS7q+//mLJkiVMnjwZQ0NDlixZQnR0NN9++60Se0xMDCdPnmT27NlYWlqyZcsWLl26pJTt+fv7M2bMGDZt2kRSUhKATlKkVquZMWMG77zzDpaWlsTHxxMZGUlmZiZNmzatdwzHjh1j9OjRdR6Ljo4mLCyMiRMnkpqayvvvv8+ePXvo1KlTne3Pnj3L1KlT6d27Nx9//DF5eXlER0fX2Xbfvn106NCBefPmKSuq169fx93dndDQUJo0acLJkyeZNWsW+vr6tVbYYmJiGDlyJOPGjWP58uVERUXh7OyMnZ0dq1evZvfu3cyYMYNu3brRunVrSktLmTRpEkFBQUyZMgWNRkNubq7yTwItNzc34uPj0Wg0Op8rIYT4fyYJlhBCPCOGhoaYm5tTUFAAwA8//MDJkyfZvn07np6eAHh5eREYGMjGjRtZuHAhd+7cISEhgbfffpuZM2cqffXq1euJYvDy8lLu23FxcWH//v1kZGSQlpbGSy+9hJ+fH+np6Rw8eFBJsD777DMsLS1JTEykSZMmwMNVnD59+nD48GH8/f2V/vv160dERAQAnp6eZGZmcuDAAeXLt5mZGRqNBldX1yeKX6u0tJT4+HgmT55MZGQkAD4+Pty7d49169YRGhqqJEfFxcXs0NKIkAAACM9JREFU3LlTKafTaDRMmTKFvLw87O3tyc3NJSMjg5UrV9KnTx8AvL296dmzJ82aNQOgZcuW2NjYANQZu1qtZtasWXh7ewPQqlUrBg8ezPHjx5Vkuqb8/Hxu3ryJg4NDncdDQkIYO3YsAD169KBv374kJCSgUqnqbJ+YmIi9vT0qlQo9PT38/Px48OABy5Ytq7N9QkIChoaGynPt+62dIw8PD/Lz80lOTq6VYI0ZM4YhQ4YozydMmICnp6eS0Dk7OyufrbCwMC5evEhJSQlz5sxRElNfX99aMXXq1Ini4mIuX77cKOWPQgjxPJASQSGEeIY0Go3y+MyZM1hYWCjJFYCxsTEBAQH8+uuvwMOyLbVazdChQxvl+l5eXspjExMTzM3N8fDw0NlgoW3btuTn5yvPf/75Z4KDg9HX16eiooKKigpsbW2xsbGptWmHj4+P8tjAwIB27drx999/N0rs1Z06dYq7d+/y1ltvKTFVVFTg5eVFQUGBzjVtbGx0vqzb29sDKGPUjkFbDghgZGRE9+7dHzseAwMDnZ0Ea16jLtpE29zcvM7jwcHBymN9fX2CgoI4c+ZMvf39/vvvBAQE6Kz8VB9TdV5eXjrJFTxMRD/99FMCAgLo0qULXbp0ISkpiUuXLtU6X5tIAtjZ2Sl9ajVv3hxzc3Nl/HZ2dhgbG/Phhx9y6NAh7ty5U2dc2rnQzo0QQrwIZAVLCCGekfLycoqKirC0tATg5s2bWFhY1GpnYWFBcXExAEVFRQBYWVk1SgwtWrTQed6kSZNarxkYGFBeXq48v337NomJibXub4KH5XeP6t/AwID79+8/bdi13L59G9BddakZl3bFqWYZpYGBAYAyxoKCApo1a1Yr4WjZsuVjx9OsWTP09f/7P0rtSl/1eaxJe0zbtqaa17ewsHjkVv83b96sdU59Y9B+BquLiYnh9OnTREREYG9vj4mJCTt37iQ9Pb1W2+pzqp3Puj5b2vfe1NSUzZs3s2bNGj744AM0Gg0+Pj7MmTOHNm3a6JwDj543IYT4fyMJlhBCPCPHjh2joqJCKTGzsrKisLCwVrvCwkJMTU0BMDMzA+r+8qxV/YusVn0rBE/C1NSUN998k5CQkFrH6lt9eda085OQkFBnktq+ffvH7svS0pKysjLKy8t1kqxbt249faCPoB1Dfe/VrVu3dOa3sLDwkYm2lZVVrZjrG0PN+5vKy8vJyspi7ty5hIaGKq/v2LHj0YNoAFdXVzZu3Iharebo0aPExcUxbdo0kpOTlTbaudDOjRBCvAikRFAIIZ6BO3fusGzZMtq2bauUnrm4uFBYWMjx48eVdvfu3SMrK0vZ2c3NzQ0jI6NH7qzWunVrLl68qPPajz/+2Gixe3t7c+HCBRwdHXFyctL5s7W1bVBfNVfHnpR2Xm7cuFErJicnpzp35auPo6MjgM5KjTYJqBk7NN7qiq2tLQYGBly9erXO49V3DayqqiI9PR1nZ+d6+3N0dCQzM1OnDDUjI+OxYrl//z5VVVU6q2mlpaWPfX5DGBkZERgYyLBhw7hw4YLOsWvXrqGvr69s8CKEEC8CWcESQoinVFlZyW+//QZAWVkZOTk57Ny5k3v37rFhwwblfqcePXrg5uZGdHQ006ZNw8zMjE2bNqFWq5XNDVq0aEFERAQqlYoHDx7g5+fH/fv3OXz4MJGRkVhbWxMcHExqaiqxsbH4+/uTnZ3NkSNHGm08kZGRhISEMGHCBIYNG6bcW3P06FGGDBmic+/RP2nfvj3p6ekcOnQIa2trWrVqhbW1db3tL168yPfff6/zWtOmTenZsyeRkZEsWrSIa9eu4eHhQVVVFZcuXSI7O5v4+PjHjsnBwYGAgAAWLFhAWVkZVlZWbN68GSMjI52Vntdeew2ArVu34uXlhYmJifLakzA0NMTR0ZGcnByGDRtW63hKSgoGBga8/vrrpKamcuXKFVasWFFvf+PHj2f48OFER0czdOhQ8vLySElJAdApX6xL8+bNcXJyIj4+HhMTE/T19Vm/fj0mJiaUlpY+8Ri1srKy2L17N0FBQbz66qvk5+eTlJSkc98WPLyPrEOHDg3eHVMIIZ5nkmAJIcRTKikpYcSIEejp6WFiYoKdnR0DBw5k1KhRtUq81q5dS1xcHLGxsZSXl+Ps7MzWrVt1/oM/ceJETE1N2bZtG7t27cLU1JRu3bopO9z5+/szdepUduzYQUpKCkFBQcyePVvZze9ptW/fnqSkJFatWsXcuXNRq9VYW1vj7e3d4JWGsLAwzp8/z6xZsyguLiYyMpL33nuv3vaZmZlkZmbqvGZjY0NGRgbjx4+nVatWbN26lc2bN2NoaEi7du3o27dvg8cYFxfH/PnzWbRoEcbGxoSFhdGmTRud3yTr1q0bY8eOZdu2baxYsQIPDw+2b9/e4GtVFxwczK5du+o8plKpiI2NZeXKlbzyyiuoVCo6d+5cb19OTk4sX74clUpFeno6jo6OzJ8/n3ffffexVvSWL1/O3LlzmTFjBmZmZowcORK1Ws0XX3zxxOPTsrOzQ09PD5VKRWFhIS1btlQ+t9UdOXKE3r17P/X1hBDieaKnqV5bIIQQQvwLVVRU0L9/f1xcXFi8ePEzu05BQQH+/v7s2LHjkeV/T2rv3r1Mnz6dQ4cO6Wwm8TzKy8ujf//+HDhwoMGlp0II8TyTFSwhhBD/Omlpady4cQMHBwfKyspITk7m8uXLLFmy5Jle19LSkpCQELZt21bv71U1xLx58/Dx8aFFixacO3eOdevW4e/v/9wnVwBbtmxh4MCBklwJIV44kmAJIYT41zE2NmbPnj1cuXKFyspKHBwc+Pzzz5/JqlJNERERpKamUllZqfN7ZE+iqKiIBQsWUFRUhJmZGX369GH69OmNFOmzo9FosLW1rfWDxkII8SKQEkEhhBBCCCGEaCSyTbsQQgghhBBCNBJJsIQQQgghhBCikUiCJYQQQgghhBCNRBIsIYQQQgghhGgkkmAJIYQQQgghRCP5D5deOIFLsswGAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 864x432 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "crIdpgUmrsEl",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "3df98ca2-55dd-45e4-d599-05b143bbabb0"
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('drive')\n",
        "df2.to_csv('text_responses.csv')\n",
        "!cp text_responses.csv \"drive/My Drive/\""
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Mounted at drive\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jdQQUG1isAk7",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "c25b3078-2258-4e3e-f5af-09a14e953a76"
      },
      "source": [
        "# only keep articles with more than 25 tokens\n",
        "df2 = df2[df2['Response Answer Bigrams'].map(len) >= 25]\n",
        "# ensure all tokenized items are lists\n",
        "df2 = df2[df2['Response Answer Bigrams'].map(type) == list]\n",
        "df2.reset_index(drop=True,inplace=True)\n",
        "print(\"After cleaning and excluding short aticles, the dataframe now has:\", len(df2), \"accounts\")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "After cleaning and excluding short aticles, the dataframe now has: 673 accounts\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EypIvV5aspJ8",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        },
        "outputId": "dae08fdd-11a3-4737-8ffa-e34dae3c86ce"
      },
      "source": [
        "df2.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>Account Type</th>\n",
              "      <th>Account Number</th>\n",
              "      <th>Account Name</th>\n",
              "      <th>Response Answer</th>\n",
              "      <th>Tokenized Response Answer</th>\n",
              "      <th>Response Answer Bigrams</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>City</td>\n",
              "      <td>35853</td>\n",
              "      <td>City of Baltimore</td>\n",
              "      <td>Key findings by sector provide guidance for wh...</td>\n",
              "      <td>[key, find, sector, provid, guidanc, strategi,...</td>\n",
              "      <td>[futur_emiss, ghg_emiss, emiss_energi, energi_...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>City</td>\n",
              "      <td>37241</td>\n",
              "      <td>City of Berkeley</td>\n",
              "      <td>Public facility. Industrial facilities. Reside...</td>\n",
              "      <td>[public, facil, industri, facil, residenti, bu...</td>\n",
              "      <td>[public_facil, industri_facil, residenti_build...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>City</td>\n",
              "      <td>49787</td>\n",
              "      <td>City of Benicia</td>\n",
              "      <td>nan. nan. nan. Benicia's commitment to a healt...</td>\n",
              "      <td>[benicia, commit, healthi, environ, resid, goe...</td>\n",
              "      <td>[activ_involv, environment_stewardship, health...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>City</td>\n",
              "      <td>52897</td>\n",
              "      <td>City of Aspen</td>\n",
              "      <td>As a part of the community-wide climate action...</td>\n",
              "      <td>[part, communitywid, climat, action, plan, asp...</td>\n",
              "      <td>[climat_action, action_plan, set_target, targe...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>City</td>\n",
              "      <td>54092</td>\n",
              "      <td>City of Ann Arbor</td>\n",
              "      <td>This is an ambitious target for our city. Ther...</td>\n",
              "      <td>[ambiti, target, citi, target, subnat, level, ...</td>\n",
              "      <td>[ambiti_target, level_target, unit_state, comm...</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "  Account Type  ...                            Response Answer Bigrams\n",
              "0         City  ...  [futur_emiss, ghg_emiss, emiss_energi, energi_...\n",
              "1         City  ...  [public_facil, industri_facil, residenti_build...\n",
              "2         City  ...  [activ_involv, environment_stewardship, health...\n",
              "3         City  ...  [climat_action, action_plan, set_target, targe...\n",
              "4         City  ...  [ambiti_target, level_target, unit_state, comm...\n",
              "\n",
              "[5 rows x 6 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 21
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kyL9fQ4rs1tq"
      },
      "source": [
        "## Build Topic Models"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-18CDxxwtB6F"
      },
      "source": [
        "def train_lda(data, num_topics, chunksize=None, alpha='symmetric', eta=None):\n",
        "    \"\"\"\n",
        "    This function trains the lda model\n",
        "    We setup parameters like number of topics, the chunksize to use in Hoffman method\n",
        "    We also do 2 passes of the data since this is a small dataset, so we want the distributions to stabilize\n",
        "\n",
        "    * low alpha means each document is only represented by a small number of topics, and vice versa\n",
        "    * low eta means each topic is only represented by a small number of words, and vice versa\n",
        "    \"\"\"\n",
        "    dictionary = corpora.Dictionary(data['Response Answer Bigrams'])\n",
        "    corpus = [dictionary.doc2bow(doc) for doc in data['Response Answer Bigrams']]\n",
        "    t1 = time.time()\n",
        "    lda = LdaModel(corpus=corpus, num_topics=num_topics, id2word=dictionary,\n",
        "                   alpha=alpha, eta=eta, chunksize=chunksize, minimum_probability=0.0, passes=2)\n",
        "    t2 = time.time()\n",
        "    print(\"Time to train LDA model on \", len(data), \"articles: \", (t2-t1)/60, \"min\")\n",
        "    return dictionary,corpus,lda"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SAm2zO9muFxx",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "59c0ccf1-1e0d-42f8-b4b6-7e501ce4369c"
      },
      "source": [
        "#dictionary,corpus,lda = train_lda(train_df, 5, chunksize = 30, eta = 0.5e-2, alpha = 1e-2)\n",
        "dictionary,corpus,lda = train_lda(df2, 20, chunksize=100)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Time to train LDA model on  673 articles:  0.024739352862040202 min\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2OZlRcn9vXz6",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 717
        },
        "outputId": "d6e011b3-042b-4168-c59a-355888dfcac2"
      },
      "source": [
        "# show_topics method shows the the top num_words contributing to num_topics number of random topics\n",
        "lda.show_topics(num_topics=20, num_words=10)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[(0,\n",
              "  '0.038*\"energi_effici\" + 0.029*\"effici_build\" + 0.027*\"build_servic\" + 0.012*\"real_estat\" + 0.012*\"emiss_reduct\" + 0.008*\"cost_save\" + 0.007*\"monetari_save\" + 0.007*\"payback_period\" + 0.007*\"pleas_specifi\" + 0.006*\"employe_commut\"'),\n",
              " (1,\n",
              "  '0.019*\"lowcarbon_energi\" + 0.017*\"solar_pv\" + 0.014*\"energi_instal\" + 0.012*\"purchas_agreement\" + 0.012*\"custom_servic\" + 0.012*\"solar_panel\" + 0.012*\"renew_sourc\" + 0.011*\"renew_energi\" + 0.010*\"purchas_electr\" + 0.009*\"power_purchas\"'),\n",
              " (2,\n",
              "  '0.023*\"emiss_reduct\" + 0.014*\"ghg_emiss\" + 0.010*\"energi_effici\" + 0.007*\"carbon_footprint\" + 0.007*\"process_emiss\" + 0.007*\"climat_chang\" + 0.007*\"north_america\" + 0.007*\"reduct_target\" + 0.006*\"greenhous_ga\" + 0.005*\"per_unit\"'),\n",
              " (3,\n",
              "  '0.016*\"margin_abat\" + 0.016*\"develop_margin\" + 0.016*\"report_program\" + 0.016*\"ghg_intens\" + 0.016*\"instrument_air\" + 0.015*\"evalu_emiss\" + 0.015*\"use_intern\" + 0.014*\"ghg_emiss\" + 0.014*\"capit_power\" + 0.013*\"intens_product\"'),\n",
              " (4,\n",
              "  '0.019*\"energi_effici\" + 0.011*\"pleas_specifi\" + 0.009*\"metric_ton\" + 0.009*\"energi_use\" + 0.009*\"emiss_reduct\" + 0.008*\"renew_energi\" + 0.008*\"effici_build\" + 0.007*\"build_servic\" + 0.007*\"squar_feet\" + 0.007*\"reduc_energi\"'),\n",
              " (5,\n",
              "  '0.029*\"fuel_effici\" + 0.015*\"emiss_reduct\" + 0.015*\"nation_park\" + 0.013*\"water_dispos\" + 0.012*\"cdp_includ\" + 0.012*\"pleas_specifi\" + 0.011*\"fuel_save\" + 0.010*\"wholli_own\" + 0.010*\"carbon_per\" + 0.010*\"social_impact\"'),\n",
              " (6,\n",
              "  '0.014*\"climat_chang\" + 0.011*\"scope_emiss\" + 0.011*\"energi_effici\" + 0.011*\"ghg_emiss\" + 0.008*\"renew_energi\" + 0.007*\"emiss_reduct\" + 0.007*\"ficial_servic\" + 0.007*\"co_ltd\" + 0.006*\"asset_manag\" + 0.006*\"head_offic\"'),\n",
              " (7,\n",
              "  '0.026*\"group_delphi\" + 0.018*\"ficial_feasibl\" + 0.018*\"paper_use\" + 0.017*\"us_food\" + 0.015*\"emiss_oper\" + 0.015*\"climat_chang\" + 0.013*\"aim_maintain\" + 0.013*\"offic_space\" + 0.013*\"chang_busi\" + 0.012*\"esg_report\"'),\n",
              " (8,\n",
              "  '0.045*\"renew_energi\" + 0.018*\"fiscal_year\" + 0.017*\"renew_electr\" + 0.013*\"energi_purchas\" + 0.012*\"purchas_renew\" + 0.011*\"environment_impact\" + 0.009*\"dell_technolog\" + 0.008*\"energi_certif\" + 0.008*\"lowcarbon_energi\" + 0.008*\"inform_technolog\"'),\n",
              " (9,\n",
              "  '0.031*\"energi_effici\" + 0.010*\"effici_build\" + 0.010*\"scope_emiss\" + 0.010*\"build_servic\" + 0.010*\"emiss_reduct\" + 0.009*\"renew_energi\" + 0.008*\"pleas_specifi\" + 0.007*\"base_year\" + 0.005*\"energi_consumpt\" + 0.005*\"metric_ton\"'),\n",
              " (10,\n",
              "  '0.037*\"forwardlook_statement\" + 0.015*\"risk_uncertainti\" + 0.010*\"power_plant\" + 0.009*\"effici_program\" + 0.009*\"electr_gener\" + 0.009*\"hydro_one\" + 0.009*\"electr_vehicl\" + 0.009*\"renew_energi\" + 0.008*\"paperbas_packag\" + 0.008*\"differ_materi\"'),\n",
              " (11,\n",
              "  '0.012*\"oper_control\" + 0.012*\"energi_effici\" + 0.011*\"innov_challeng\" + 0.011*\"lowest_ghg\" + 0.011*\"billion_asset\" + 0.011*\"train_program\" + 0.010*\"hotel_resort\" + 0.009*\"project_develop\" + 0.009*\"new_emiss\" + 0.009*\"move_forward\"'),\n",
              " (12,\n",
              "  '0.084*\"natur_ga\" + 0.044*\"emiss_reduct\" + 0.021*\"tonn_coe\" + 0.012*\"methan_emiss\" + 0.011*\"process_emiss\" + 0.010*\"fugit_emiss\" + 0.010*\"metric_tonn\" + 0.009*\"oil_ga\" + 0.007*\"elk_hill\" + 0.006*\"approxim_million\"'),\n",
              " (13,\n",
              "  '0.046*\"energi_effici\" + 0.032*\"effici_process\" + 0.015*\"process_optim\" + 0.013*\"compress_air\" + 0.012*\"greenhous_ga\" + 0.010*\"effici_build\" + 0.009*\"energi_intens\" + 0.008*\"build_servic\" + 0.007*\"manufactur_facil\" + 0.007*\"design_manufactur\"'),\n",
              " (14,\n",
              "  '0.027*\"ghg_emiss\" + 0.016*\"emiss_reduct\" + 0.010*\"reduct_target\" + 0.009*\"reduct_measur\" + 0.009*\"emiss_normal\" + 0.009*\"energi_effici\" + 0.008*\"longterm_goal\" + 0.008*\"initi_drive\" + 0.007*\"reduct_goal\" + 0.007*\"scope_ghg\"'),\n",
              " (15,\n",
              "  '0.028*\"suppli_chain\" + 0.023*\"manag_servic\" + 0.021*\"fuch_group\" + 0.020*\"servic_includ\" + 0.019*\"servic_provid\" + 0.016*\"product_compani\" + 0.014*\"provid_servic\" + 0.013*\"year_lowcarbon\" + 0.013*\"unit_state\" + 0.012*\"project_manag\"'),\n",
              " (16,\n",
              "  '0.020*\"water_wast\" + 0.018*\"target_appli\" + 0.017*\"energi_water\" + 0.014*\"minim_environment\" + 0.012*\"leas_asset\" + 0.012*\"light_system\" + 0.011*\"custom_relationship\" + 0.010*\"site_includ\" + 0.010*\"busi_develop\" + 0.009*\"sand_eco\"'),\n",
              " (17,\n",
              "  '0.013*\"ghg_emiss\" + 0.013*\"emiss_reduct\" + 0.011*\"energi_use\" + 0.008*\"coe_per\" + 0.008*\"per_squar\" + 0.008*\"scope_emiss\" + 0.007*\"mt_coe\" + 0.007*\"next_two\" + 0.007*\"emiss_per\" + 0.007*\"intens_target\"'),\n",
              " (18,\n",
              "  '0.016*\"distribut_center\" + 0.015*\"reduct_activ\" + 0.014*\"energi_effici\" + 0.012*\"global_oper\" + 0.011*\"emiss_reduct\" + 0.011*\"carbon_neutral\" + 0.010*\"drive_invest\" + 0.010*\"invest_emiss\" + 0.010*\"own_leas\" + 0.008*\"includ_cdp\"'),\n",
              " (19,\n",
              "  '0.068*\"energi_effici\" + 0.047*\"effici_build\" + 0.043*\"build_servic\" + 0.014*\"effici_process\" + 0.009*\"servic_light\" + 0.008*\"per_metric\" + 0.008*\"servic_energi\" + 0.007*\"pleas_specifi\" + 0.007*\"process_optim\" + 0.006*\"light_energi\"')]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 45
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Z11H4O_8Jyq7"
      },
      "source": [
        "def plot_topic_dist(idx):\n",
        "\n",
        "  bow = dictionary.doc2bow(df2.iloc[idx,5])\n",
        "  doc_distribution = np.array([tup[1] for tup in lda.get_document_topics(bow=bow)])\n",
        "\n",
        "  # bar plot of topic distribution for this document\n",
        "  fig, ax = plt.subplots(figsize=(12,6));\n",
        "\n",
        "  # the histogram of the data\n",
        "  patches = ax.bar(np.arange(len(doc_distribution)), doc_distribution)\n",
        "  ax.set_xlabel('Topic ID', fontsize=15)\n",
        "  ax.set_ylabel('Topic Contribution', fontsize=15)\n",
        "  ax.set_title(\"Topic Distribution for: \" + df2.iloc[idx, 2], fontsize=20)\n",
        "  ax.set_xticks(np.linspace(0,20,5))\n",
        "  fig.tight_layout()\n",
        "  plt.show()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DhidxEq3TOUL",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 441
        },
        "outputId": "f4c867dc-8061-43b4-a7cd-8d820c2383da"
      },
      "source": [
        "rand_idx = np.random.randint(len(df2))\n",
        "plot_topic_dist(rand_idx)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 864x432 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Tsbpog84KVKY",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "27af799e-86df-413b-f5a7-5776836e5e90"
      },
      "source": [
        "# matrix of topic distributions for all \n",
        "doc_topic_dist = np.array([[tup[1] for tup in lst] for lst in lda[corpus]])\n",
        "doc_topic_dist.shape"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(673, 20)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 54
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "50rTwkYgTwZ3"
      },
      "source": [
        "## Compute Similarity Scores between Corporate & City Topic Distributions\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2cKHynYrKZSa"
      },
      "source": [
        "from scipy.spatial.distance import jensenshannon\n",
        "\n",
        "def jensen_shannon(query, matrix):\n",
        "  \"\"\" Finds Jensen-Shannon similarity between query account & every account in matrix \"\"\" \n",
        "\n",
        "  return np.array([jensenshannon(p,query) for p in matrix])\n",
        "\n",
        "def get_most_similar_documents(query,matrix,k=10):\n",
        "    \"\"\"\n",
        "    Implements the Jensen-Shannon distance\n",
        "    and returns the top k indices of the smallest jensen shannon distances\n",
        "    \"\"\"\n",
        "    sims = jensen_shannon(query,matrix) # list of jensen shannon distances\n",
        "    return np.argsort(sims)[:k], np.sort(sims)[:k] # the top k positional index of the smallest Jensen Shannon distances"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "03oWouhoKhYi"
      },
      "source": [
        "idx = 0\n",
        "new_bow = dictionary.doc2bow(df2.iloc[idx,5])\n",
        "new_doc_distribution = np.array([tup[1] for tup in lda.get_document_topics(bow=new_bow)])\n",
        "\n",
        "sims = jensen_shannon(new_doc_distribution, doc_topic_dist)\n",
        "most_sim_ids, most_sims_score = get_most_similar_documents(new_doc_distribution, doc_topic_dist)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fESu62i7K0Qv",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "8acc1ad3-292e-4652-d9de-5ce90af66461"
      },
      "source": [
        "most_similar_df = df2[df2.index.isin(most_sim_ids[0:3])]\n",
        "most_similar_df['Account Name']\n",
        "most_sims_score[4]"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.5243355747356068"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 91
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zGbmifcLRqtJ",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 441
        },
        "outputId": "df44ace1-b8c1-4768-992f-71fb8676702d"
      },
      "source": [
        "plot_topic_dist(0)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 864x432 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3u-0BxCDSIft",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 441
        },
        "outputId": "5529ad0e-a277-4f76-c406-579206ad21a5"
      },
      "source": [
        "plot_topic_dist(92)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 864x432 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qWCAk727Ul3A"
      },
      "source": [
        "## Build Free Text Similarity Matrix between Cities & Corporations"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZlwH3aESUlGw",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 272
        },
        "outputId": "cea2d7f7-99f9-43e6-b039-6007989b3676"
      },
      "source": [
        "city_idx = df2.index[df2['Account Type'] == 'City'].tolist()\n",
        "corp_idx = df2.index[df2['Account Type'] == 'Corporation'].tolist()\n",
        "\n",
        "sim_matrix = pd.DataFrame({'account_1' : [], \n",
        "                           'account_1_type': [],\n",
        "                           'account_2':[], \n",
        "                           'account_2_type': [],\n",
        "                           'sim_score': []})\n",
        "\n",
        "for i in city_idx:\n",
        "  b = dictionary.doc2bow(df2.iloc[i,5])\n",
        "  d = np.array([tup[1] for tup in lda.get_document_topics(bow=b)])\n",
        "  sim_scores = jensen_shannon(d, doc_topic_dist) # array for sim scores for city `i`\n",
        "\n",
        "  append_df = pd.DataFrame({'account_1': df2.iloc[i,1],\n",
        "                            'account_1_type': df2.iloc[i,0],\n",
        "                            'account_2': df2['Account Number'],\n",
        "                            'account_2_type': df2['Account Type'],\n",
        "                            'sim_score': sim_scores})\n",
        "  \n",
        "  sim_matrix = sim_matrix.append(append_df, ignore_index=True)\n",
        "\n",
        "sim_matrix = sim_matrix[sim_matrix['account_2_type'] == 'Corporation']\n",
        "sim_matrix['account_1'] = sim_matrix['account_1'].astype('int')\n",
        "sim_matrix['account_2'] = sim_matrix['account_2'].astype('int')\n",
        "  \n",
        "print(sim_matrix.shape)\n",
        "print(len(city_idx) * len(corp_idx))\n",
        "\n",
        "sim_matrix.head()\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/scipy/spatial/distance.py:1296: RuntimeWarning: invalid value encountered in sqrt\n",
            "  return np.sqrt(js / 2.0)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "(5320, 5)\n",
            "5320\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>account_1</th>\n",
              "      <th>account_1_type</th>\n",
              "      <th>account_2</th>\n",
              "      <th>account_2_type</th>\n",
              "      <th>sim_score</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>8</th>\n",
              "      <td>35853</td>\n",
              "      <td>City</td>\n",
              "      <td>58</td>\n",
              "      <td>Corporation</td>\n",
              "      <td>0.603660</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>9</th>\n",
              "      <td>35853</td>\n",
              "      <td>City</td>\n",
              "      <td>64</td>\n",
              "      <td>Corporation</td>\n",
              "      <td>0.730616</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>10</th>\n",
              "      <td>35853</td>\n",
              "      <td>City</td>\n",
              "      <td>119</td>\n",
              "      <td>Corporation</td>\n",
              "      <td>0.739552</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>11</th>\n",
              "      <td>35853</td>\n",
              "      <td>City</td>\n",
              "      <td>285</td>\n",
              "      <td>Corporation</td>\n",
              "      <td>0.685070</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>12</th>\n",
              "      <td>35853</td>\n",
              "      <td>City</td>\n",
              "      <td>291</td>\n",
              "      <td>Corporation</td>\n",
              "      <td>0.790877</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "    account_1 account_1_type  account_2 account_2_type  sim_score\n",
              "8       35853           City         58    Corporation   0.603660\n",
              "9       35853           City         64    Corporation   0.730616\n",
              "10      35853           City        119    Corporation   0.739552\n",
              "11      35853           City        285    Corporation   0.685070\n",
              "12      35853           City        291    Corporation   0.790877"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 78
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "FHIEA5NEBU2_",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "074cad6b-9d44-4f83-81e8-761b334dd0d4"
      },
      "source": [
        "drive.mount('drive')\n",
        "sim_matrix.to_csv('text_response_similarity_scores.csv')\n",
        "!cp text_response_similarity_scores.csv \"drive/My Drive/\""
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Drive already mounted at drive; to attempt to forcibly remount, call drive.mount(\"drive\", force_remount=True).\n"
          ],
          "name": "stdout"
        }
      ]
    }
  ]
}