2018-07-12 Digital Entrep POC - ML Topic Modelling with Gensim.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Notebook setup"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Import libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import spacy\n",
    "from spacy.lang.en import English\n",
    "import nltk\n",
    "from nltk.corpus import wordnet as wn\n",
    "from nltk.stem.wordnet import WordNetLemmatizer\n",
    "import pandas as pd\n",
    "import random\n",
    "from gensim import corpora\n",
    "import pickle\n",
    "import gensim\n",
    "import pyLDAvis.gensim\n",
    "import os"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 134,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-12T18:25:44.472101Z",
     "start_time": "2018-07-12T18:25:43.736498Z"
    }
   },
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Set notebook-wide defaults"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 135,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-12T18:25:57.480057Z",
     "start_time": "2018-07-12T18:25:57.464445Z"
    }
   },
   "outputs": [],
   "source": [
    "# set\n",
    "sns.set(font=\"Open Sans\", style='whitegrid', rc={\"grid.linewidth\": 0.01, 'figure.figsize':(10,5)})\n",
    "sns.set_context(\"paper\", font_scale=1.75)  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 136,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-12T18:25:57.730029Z",
     "start_time": "2018-07-12T18:25:57.636430Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1aeac0e9048>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#set color palette\n",
    "wbg_seq_pal = [\"#009fda\", '#0071bc', '#004c92', '#002245']\n",
    "sns.palplot(sns.color_palette(wbg_seq_pal))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 137,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-12T18:25:57.871807Z",
     "start_time": "2018-07-12T18:25:57.776905Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1aeac109518>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "wbg_cat_pal = [\"#009fda\", '#002245', '#4cbb88', '#288462', '#f07761', '#cd2c0f', '#d4d4d4', '#333333']\n",
    "sns.palplot(sns.color_palette(wbg_cat_pal))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 138,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-12T18:25:58.059320Z",
     "start_time": "2018-07-12T18:25:57.965562Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1aeac174ac8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "wbg_cat_pal2 = [\"#009fda\", '#4cbb88', '#cd2c0f', '#333333']\n",
    "sns.palplot(sns.color_palette(wbg_cat_pal2))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Create output folders"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "output_folder = \"outputs\"\n",
    "\n",
    "os.makedirs(output_folder, exist_ok=True)\n",
    "for subfolder in ['3 Topic Modelling']:\n",
    "    os.makedirs(os.path.join(output_folder, subfolder), exist_ok=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Download stopwords, etc."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-12T17:34:37.744169Z",
     "start_time": "2018-07-12T17:32:33.883569Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "showing info https://raw.githubusercontent.com/nltk/nltk_data/gh-pages/index.xml\n",
      "showing info http://nltk.org/nltk_data/\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "spacy.load('en')\n",
    "parser = English()\n",
    "nltk.download()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-12T17:35:12.010295Z",
     "start_time": "2018-07-12T17:35:11.979166Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[nltk_data] Downloading package stopwords to\n",
      "[nltk_data]     C:\\Users\\mrpso\\AppData\\Roaming\\nltk_data...\n",
      "[nltk_data]   Package stopwords is already up-to-date!\n"
     ]
    }
   ],
   "source": [
    "nltk.download('stopwords')\n",
    "en_stop = set(nltk.corpus.stopwords.words('english'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 172,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-12T18:58:54.744016Z",
     "start_time": "2018-07-12T18:58:54.728394Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'my', 'below', 'have', 'other', 'them', 'been', 'same', 'into', 'but', 'll', 'they', 'up', 'who', 'weren', 'will', 'both', 'wasn', 'your', 'was', 'that', 'me', \"aren't\", \"weren't\", \"hasn't\", \"needn't\", 'y', 'these', 'off', 'more', 'yourselves', 'out', 'through', 'any', 'then', 'or', 'from', \"haven't\", 'aren', 's', 'after', 'didn', 'shouldn', 'between', 'those', \"you'd\", 'which', 'had', \"you're\", 'very', 'our', 'for', 'it', \"mustn't\", 'can', 'wouldn', 'now', 'during', 'its', 'be', 'of', 'down', 'no', 'were', 'd', 'doesn', \"she's\", \"should've\", 'being', 'an', 'do', 'there', 'are', 'the', 'this', 'shan', 'don', 'couldn', 'in', 'myself', \"didn't\", \"couldn't\", 'their', 'if', 'here', 'where', \"won't\", 'just', 'ours', 'haven', 'what', 'by', 'most', 'does', 't', 'as', 'he', 'itself', 'than', 'and', 'herself', 'mustn', \"doesn't\", 'when', 'all', \"isn't\", 'theirs', 'is', 'own', 'above', \"you've\", 'until', 'not', 'whom', 'm', \"hadn't\", \"shan't\", 'am', 'too', 'again', 'once', 'ma', 'such', 'against', 'under', 'how', 'her', 'so', 'his', \"wasn't\", 'mightn', \"wouldn't\", 'each', 'hadn', 'some', 'himself', 've', 'hers', 'we', 'she', 'at', 'him', \"it's\", 'over', 'to', \"shouldn't\", 're', 'because', 'yourself', 'few', 'a', \"don't\", 'ain', 'themselves', 'i', \"that'll\", 'o', 'isn', 'needn', 'won', 'doing', 'having', 'ourselves', 'nor', 'before', 'on', 'with', 'should', \"you'll\", 'why', 'yours', 'about', \"mightn't\", 'further', 'has', 'while', 'only', 'did', 'hasn', 'you'}\n"
     ]
    }
   ],
   "source": [
    "print(en_stop)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Define helper functions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def tokenize(text):\n",
    "    lda_tokens = []\n",
    "    tokens = parser(text)\n",
    "    for token in tokens:\n",
    "        if token.orth_.isspace():\n",
    "            continue\n",
    "        elif token.like_url:\n",
    "            lda_tokens.append('URL')\n",
    "        elif token.orth_.startswith('@'):\n",
    "            lda_tokens.append('SCREEN_NAME')\n",
    "        else:\n",
    "            lda_tokens.append(token.lower_)\n",
    "    return lda_tokens"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-12T17:28:06.481681Z",
     "start_time": "2018-07-12T17:28:06.450425Z"
    }
   },
   "outputs": [],
   "source": [
    "def get_lemma(word):\n",
    "    lemma = wn.morphy(word)\n",
    "    if lemma is None:\n",
    "        return word\n",
    "    else:\n",
    "        return lemma"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_lemma2(word):\n",
    "    return WordNetLemmatizer().lemmatize(word)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-12T17:35:14.496924Z",
     "start_time": "2018-07-12T17:35:14.481270Z"
    }
   },
   "outputs": [],
   "source": [
    "def prepare_text_for_lda(text):\n",
    "    tokens = tokenize(text)\n",
    "    tokens = [token for token in tokens if len(token) > 4]\n",
    "    tokens = [token for token in tokens if token not in en_stop]\n",
    "    tokens = [get_lemma(token) for token in tokens]\n",
    "    return tokens"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-16T03:55:36.195301Z",
     "start_time": "2019-05-16T03:55:36.192309Z"
    }
   },
   "outputs": [],
   "source": [
    "# check lemmatizer functions\n",
    "for w in ['dogs', 'ran', 'discouraged']:\n",
    "    print(w, get_lemma(w), get_lemma2(w))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 151,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-12T18:30:42.573426Z",
     "start_time": "2018-07-12T18:30:42.557803Z"
    }
   },
   "outputs": [],
   "source": [
    "def assign_cluster(x, thresh = 0.95):\n",
    "    topic = ''\n",
    "    if x['topic1'] > thresh:\n",
    "        topic = 'topic1'\n",
    "    elif x['topic2'] > thresh:\n",
    "        topic = 'topic2'\n",
    "        \n",
    "    elif x['topic3'] > thresh:\n",
    "        topic = 'topic3'\n",
    "    return(topic)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Code"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Prepare the text data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-12T17:35:25.307237Z",
     "start_time": "2018-07-12T17:35:23.822139Z"
    }
   },
   "outputs": [],
   "source": [
    "# import data source. Make sure to run the '2018-07-10 Digital Entrep POC - Web Scraping.ipynb' before this to get the data\n",
    "df_nigeria_startups = pd.read_csv('outputs/1 extract lists/%s - Nigeria Startups List.csv')\n",
    "df_nigeria_startups = df_nigeria_startups[df_nigeria_startups['startup_description'].notnull()].reset_index(drop=True)\n",
    "\n",
    "# combine text columns\n",
    "df_nigeria_startups['text'] = df_nigeria_startups['startup_description'] + df_nigeria_startups['startup_keywords']\n",
    "\n",
    "# create list of text keywords\n",
    "data = df_nigeria_startups['text'].tolist()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-16T03:58:04.882418Z",
     "start_time": "2019-05-16T03:58:04.878427Z"
    }
   },
   "outputs": [],
   "source": [
    "text_data = []\n",
    "\n",
    "for line in data:\n",
    "    tokens = prepare_text_for_lda(line)\n",
    "    if random.random() > .99:\n",
    "        print(tokens)\n",
    "    text_data.append(tokens)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-16T03:58:09.200224Z",
     "start_time": "2019-05-16T03:58:09.196263Z"
    },
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "print(text_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-12T18:11:53.199925Z",
     "start_time": "2018-07-12T18:11:53.168770Z"
    }
   },
   "outputs": [],
   "source": [
    "# create the dictionary\n",
    "dictionary = corpora.Dictionary(text_data)\n",
    "corpus = [dictionary.doc2bow(text) for text in text_data]\n",
    "pickle.dump(corpus, open('corpus.pkl', 'wb'))\n",
    "dictionary.save('dictionary.gensim')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Create LDA model with 5 topics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-16T03:59:12.845404Z",
     "start_time": "2019-05-16T03:59:12.842431Z"
    },
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "NUM_TOPICS = 3\n",
    "ldamodel = gensim.models.ldamodel.LdaModel(corpus, num_topics = NUM_TOPICS, id2word=dictionary, passes=15)\n",
    "ldamodel.save('model5.gensim')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-12T18:12:06.088982Z",
     "start_time": "2018-07-12T18:12:06.073360Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(0, '0.023*\"social\" + 0.015*\"online\" + 0.011*\"network\" + 0.011*\"business\"')\n",
      "(1, '0.019*\"mobile\" + 0.013*\"africa\" + 0.012*\"nigeria\" + 0.012*\"business\"')\n",
      "(2, '0.020*\"social\" + 0.016*\"platform\" + 0.015*\"nigeria\" + 0.012*\"medium\"')\n"
     ]
    }
   ],
   "source": [
    "# view the topics created\n",
    "topics = ldamodel.print_topics(num_words=4)\n",
    "for topic in topics:\n",
    "    print(topic)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Create LDA visualization using `pyLDAvis`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-12T17:50:41.396409Z",
     "start_time": "2018-07-12T17:50:41.380816Z"
    }
   },
   "outputs": [],
   "source": [
    "dictionary = gensim.corpora.Dictionary.load('dictionary.gensim')\n",
    "corpus = pickle.load(open('corpus.pkl', 'rb'))\n",
    "lda = gensim.models.ldamodel.LdaModel.load('model5.gensim')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-12T18:00:29.138246Z",
     "start_time": "2018-07-12T18:00:28.549871Z"
    }
   },
   "outputs": [],
   "source": [
    "lda_display = pyLDAvis.gensim.prepare(lda, corpus, dictionary, sort_topics=False)\n",
    "# pyLDAvis.display(lda_display)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-12T17:52:36.949575Z",
     "start_time": "2018-07-12T17:52:36.918302Z"
    }
   },
   "outputs": [],
   "source": [
    "pyLDAvis.save_html(lda_display, 'outputs/3 Topic Modelling/2018-07-12 - Nigeria Startups - Topic Modelling LDA visualization.html')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Implement Thematic clustering"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-12T18:05:59.711582Z",
     "start_time": "2018-07-12T18:05:59.695961Z"
    }
   },
   "outputs": [],
   "source": [
    "lda_corpus = lda[corpus]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 125,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-12T18:23:39.579803Z",
     "start_time": "2018-07-12T18:23:39.485988Z"
    }
   },
   "outputs": [],
   "source": [
    "collect_vals = []\n",
    "for row, dat in zip(lda_corpus,data):\n",
    "    temp_list = [dat]\n",
    "    for i in row:\n",
    "        temp_list.append(list(i)[1])\n",
    "    collect_vals.append(temp_list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 183,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-12T19:26:55.973466Z",
     "start_time": "2018-07-12T19:26:55.377065Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1aeae5a7438>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1aeae07f978>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1aeae6946a0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_topic_scores = pd.DataFrame(collect_vals, columns = ['text', 'topic1', 'topic2', 'topic3'])\n",
    "for top in df_topic_scores.columns[1:]:\n",
    "    df_topic_scores[top].hist(bins=100, figsize=(7,3))\n",
    "    plt.axvline(0.97, color='black', linestyle='--')\n",
    "    plt.title(top)\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-12T18:28:49.574463Z",
     "start_time": "2018-07-12T18:28:49.558842Z"
    }
   },
   "outputs": [],
   "source": [
    "df_nigeria_startups_final = df_nigeria_startups.merge(df_topic_scores, on='text')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 155,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-12T18:31:08.883009Z",
     "start_time": "2018-07-12T18:31:08.851880Z"
    }
   },
   "outputs": [],
   "source": [
    "df_nigeria_startups_final['cluster_thresh0.97'] = df_nigeria_startups_final.apply(lambda x: assign_cluster(x, 0.97), axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 157,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-12T18:31:25.136137Z",
     "start_time": "2018-07-12T18:31:25.120498Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "          89\n",
       "topic3    60\n",
       "topic1    53\n",
       "topic2    46\n",
       "Name: cluster_thresh0.97, dtype: int64"
      ]
     },
     "execution_count": 157,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_nigeria_startups_final['cluster_thresh0.97'].value_counts(dropna=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-16T04:02:17.187703Z",
     "start_time": "2019-05-16T04:02:17.184710Z"
    }
   },
   "outputs": [],
   "source": [
    "top = 'topic1'\n",
    "df_nigeria_startups_final[df_nigeria_startups_final['cluster_thresh0.97'] == top].sort_values(top, ascending=False)[['startup_name', 'text']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 167,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-12T18:34:00.378993Z",
     "start_time": "2018-07-12T18:34:00.363401Z"
    }
   },
   "outputs": [],
   "source": [
    "df_nigeria_startups_final = df_nigeria_startups_final.drop(['Unnamed: 0'], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 168,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-12T18:34:35.987994Z",
     "start_time": "2018-07-12T18:34:35.941221Z"
    }
   },
   "outputs": [],
   "source": [
    "df_nigeria_startups_final.to_csv('outputs/3 Topic Modelling/2018-07-12 - Nigeria Startups - Clustering.csv')"
   ]
  }
 ],
 "metadata": {
  "hide_input": false,
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.7"
  },
  "latex_envs": {
   "LaTeX_envs_menu_present": true,
   "autoclose": false,
   "autocomplete": true,
   "bibliofile": "biblio.bib",
   "cite_by": "apalike",
   "current_citInitial": 1,
   "eqLabelWithNumbers": true,
   "eqNumInitial": 1,
   "hotkeys": {
    "equation": "Ctrl-E",
    "itemize": "Ctrl-I"
   },
   "labels_anchors": false,
   "latex_user_defs": false,
   "report_style_numbering": false,
   "user_envs_cfg": false
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {
    "height": "68px",
    "width": "252px"
   },
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": "block",
   "toc_window_display": true
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}