src/Wikipedia Navigation Vectors - Getting Started.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Wikipedia Navigation Vectors - Getting Started"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In this project, we learned embeddings for Wikipedia articles and [Wikidata](https://www.wikidata.org/wiki/Wikidata:Main_Page) items by applying [Word2vec](https://en.wikipedia.org/wiki/Word2vec) models to a corpus of reading sessions.\n",
    "\n",
    "Although Word2vec models were developed to learn word embeddings from a corpus of sentences, they can be applied to any kind of sequential data. The learned embeddings have the property that items with similar neighbors in the training corpus have similar representations (as measured by the [cosine similarity](https://en.wikipedia.org/wiki/Cosine_similarity), for example). Consequently, applying Wor2vec to reading sessions results in article embeddings, where articles that tend to be read in close succession have similar representations. Since people usually generate sequences of semantically related articles while reading, these embeddings also capture semantic similarity between articles.\n",
    "\n",
    "There have been several approaches to learning vector representations of Wikipedia articles that capture semantic similarity by using the article text or the links between articles. An advantage of training Word2vec models on reading sessions, is that they learn from the actions of millions of humans who are using a diverse array of signals, including the article text, links, third-party search engines, and their existing domain knowledge, to determine what to read next in order to learn about a topic.\n",
    "\n",
    "An additional feature of not relying on text or links, is that we can learn representations for [Wikidata Items](https://www.wikidata.org/wiki/Help:Items) by simply mapping article titles within each session to Wikidata items using [Wikidata sitelinks](https://www.wikidata.org/wiki/Help:Sitelinks). As a result, these Wikidata vectors are jointly trained over reading sessions for all Wikipedia language editions, allowing the model to learn from people across the globe. This approach also overcomes data sparsity issues for smaller Wikipedias, since the representations for articles in smaller Wikipedias are shared across many other potentially larger ones. Finally, instead of needing to generate a separate embedding for each Wikipedia in each language, we have a single model that gives a vector representation for any article in any language, provided the article has been mapped to a Wikidata item."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Documentation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "More detailed documentation can be found on [here](https://meta.wikimedia.org/wiki/Research:Wikipedia_Navigation_Vectors)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Downloading the Vectors"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The vectors are available for download at [figshare](https://dx.doi.org/10.6084/m9.figshare.3146878)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### A Simple Embedding Class"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The `WikiEmbedding` class takes care of loading the vector data into a matrix and exposes a method for finding the nearest neighbors of an article in the embedding."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from sklearn.preprocessing import normalize\n",
    "import requests\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')\n",
    "%matplotlib inline\n",
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "\n",
    "\n",
    "class WikiEmbedding:\n",
    "\n",
    "    def __init__(self, fname):\n",
    "        \n",
    "        self.w2idx = {}\n",
    "        self.idx2w = []\n",
    "    \n",
    "        with open(fname, 'rb') as f:\n",
    "            \n",
    "            m, n = next(f).decode('utf8').strip().split(' ')\n",
    "            self.E = np.zeros((int(m), int(n)))\n",
    "\n",
    "            for i, l in enumerate(f):\n",
    "                l = l.decode('utf8').strip().split(' ')\n",
    "                w = l[0]\n",
    "                self.E[i] = np.array(l[1:])\n",
    "                self.w2idx[w] = i\n",
    "                self.idx2w.append(w)\n",
    "                \n",
    "        self.E = normalize(self.E)\n",
    "        self.idx2w = np.array(self.idx2w)\n",
    "\n",
    "    def most_similar(self, w, n=10, min_similarity=0.5):\n",
    "        \"\"\"\n",
    "        Find the top-N most similar words to w, based on cosine similarity.\n",
    "        As a speed optimization, only consider neighbors with a similarity\n",
    "        above min_similarity\n",
    "        \"\"\"\n",
    "        \n",
    "        if type(w) is str:\n",
    "            w = self.E[self.w2idx[w]]\n",
    "\n",
    "        scores = self.E.dot(w)\n",
    "        # only consider neighbors above threshold\n",
    "        min_idxs = np.where(scores > min_similarity)\n",
    "        ranking = np.argsort(-scores[min_idxs])[1:(n+1)]\n",
    "        nn_ws = self.idx2w[min_idxs][ranking]\n",
    "        nn_scores = scores[min_idxs][ranking]\n",
    "        return list(zip(list(nn_ws), list(nn_scores)))\n",
    "    "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Exploring Nearest Neighbors"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Here we load an embedding over English Wikipedia articles trained on reading sessions in enwiki for the first week of March 2016 (Note: we will get to using Wikidata embeddings in a bit). We then show the nearest neighbors for some sample articles."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "en_embedding = WikiEmbedding('../data/2016-03-01_2016-03-07/2016-03-01_2016-03-07_en_200')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('Word_embedding', 0.95622320591391496),\n",
       " ('Gensim', 0.9269277744544846),\n",
       " ('Brown_clustering', 0.90669111256072077),\n",
       " ('Bag-of-words_model', 0.8997812984904453),\n",
       " ('Distributional_semantics', 0.89210129481761236),\n",
       " ('Bag-of-words_model_in_computer_vision', 0.89153302212057661),\n",
       " ('Topic_model', 0.89111001451403138),\n",
       " ('Feature_learning', 0.88864836681622683),\n",
       " ('Vector_space_model', 0.88702724731990568),\n",
       " ('Document-term_matrix', 0.88491820738885474)]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "en_embedding.most_similar('Word2vec')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('Wiki', 0.83135166035258534),\n",
       " ('English_Wikipedia', 0.80031730995900063),\n",
       " ('Wikimedia_Foundation', 0.72032968647518936),\n",
       " ('Larry_Sanger', 0.6898701159733629),\n",
       " ('Nupedia', 0.68647996980941084),\n",
       " ('History_of_Wikipedia', 0.63363057247932586),\n",
       " ('Encyclopedia', 0.61845099080085975),\n",
       " ('Jimmy_Wales', 0.61750657289550304),\n",
       " ('Wikipedia_community', 0.58988242628270737),\n",
       " ('List_of_wikis', 0.58899013182893245)]"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "en_embedding.most_similar('Wikipedia')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Visualizing the Embedding using TSNE\n",
    "\n",
    "\n",
    "The [t-SNE](https://en.wikipedia.org/wiki/T-distributed_stochastic_neighbor_embedding) algorithm tries to take a collection of high dimensional vectors and find lower dimensional representations such that the relative distances between points stay preserved as well as possible. This allows us to visualize a part of the embedding in 2D. In the plot below you see the 2D t-SNE projection for the 20 most popular articles in the training data. (Note: The code for getting a readable scatter plot is a bit hairy, so the code has been factored out to a helper file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from getting_started_helpers import get_tsne, plot_tsne"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAsUAAAFrCAYAAAA5NfQbAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1520c6470>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "tsne_en_embedding = get_tsne(en_embedding, pca_dim = 10, n_items=5000)\n",
    "plot_tsne(en_embedding, tsne_en_embedding, n = 20)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "When looking at a single week of data, the most popular articles tend to be about things trending in the media. We can see that articles about similar topics and events cluster together."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Working with Wikidata Vectors\n",
    "\n",
    "One difficulty with working with the Wikidata embedding is that item ids are hard to interpret. We can, however, use the Wikidata API to map back and forth between Wikipedia articles and Wikidata ids. Below are some helper funtions for doing the mapping."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def items_to_titles(items, lang):\n",
    "    \"\"\"\n",
    "    Input: a list of Wikidata item ids\n",
    "    Output: a dictionary mapping from ids to title in lang\n",
    "    \n",
    "    Note: items without an articlce in lang are not included in the output\n",
    "    \"\"\"\n",
    "    lang += 'wiki'\n",
    "    payload = {'action': 'wbgetentities',\n",
    "               'props': 'sitelinks/urls',\n",
    "               'format': 'json',\n",
    "               'ids': '|'.join(items),               \n",
    "              }\n",
    "    r = requests.get('https://www.wikidata.org/w/api.php', params=payload).json()\n",
    "    \n",
    "    return parse_wikidata_sitelinks(r, lang, True)\n",
    "    \n",
    "    \n",
    "def titles_to_items(titles, lang):\n",
    "    \"\"\"\n",
    "    Input: a list of article titles in lang\n",
    "    Output: a dictionary mapping from titles in lang to Wikidata ids\n",
    "    \n",
    "    Note: articles in lang without a Wikidata id are not included in the output\n",
    "\n",
    "    \"\"\"\n",
    "    lang += 'wiki'\n",
    "    payload = {'action': 'wbgetentities',\n",
    "               'props': 'sitelinks/urls',\n",
    "               'format': 'json',\n",
    "               'sites': lang,\n",
    "               'titles': '|'.join(titles),\n",
    "              }\n",
    "    r = requests.get('https://www.wikidata.org/w/api.php', params=payload).json()\n",
    "    \n",
    "    return parse_wikidata_sitelinks(r, lang, False)\n",
    "\n",
    "\n",
    "def parse_wikidata_sitelinks(response, lang, item_to_title):\n",
    "    \"\"\"\n",
    "    Helper function for parsing sitelinks from Wikidata Api\n",
    "    \"\"\"\n",
    "    d = {}\n",
    "    if 'entities' not in response:\n",
    "        print ('No entities in reponse')\n",
    "        return d\n",
    "\n",
    "    for item, v in response['entities'].items():\n",
    "        if 'sitelinks' in v:\n",
    "            if lang in v['sitelinks']:\n",
    "                title = v['sitelinks'][lang]['title'].replace(' ', '_')\n",
    "                if item_to_title:\n",
    "                    d[item] = title\n",
    "                else:\n",
    "                    d[title] = item\n",
    "    return d"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Given these helper functions, we can now look for nearest neighbors for a Wikipedia article in any language by first mapping the article to its Wikidata item, then finding the nearest neighbors in the Wikidata embedding and finally mapping the Wikidata items back to article names."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def most_similar(embedding, title, lang, n=10, min_similarity=0.7):\n",
    "    item = titles_to_items([title,], lang)[title]\n",
    "    nn = embedding.most_similar(item, n=n, min_similarity=min_similarity)\n",
    "    nn_items = [x[0] for x in nn]\n",
    "    nn_items_to_titles = items_to_titles(nn_items, lang)\n",
    "    return [(nn_items_to_titles[x[0]], x[1]) for x in nn if x[0] in nn_items_to_titles]    "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Lets try it out! First we load the wikidata embedding."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "wikidata_embedding = WikiEmbedding('../data/2016-03-01_2016-03-07/2016-03-01_2016-03-07_wikidata_200')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Then we can query nearest neighbors for 'Q1', the 'Universe' Wikidata item."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('Q229206', 0.84477422002858837),\n",
       " ('Q318', 0.81135711307867753),\n",
       " ('Q323', 0.80632248705977405),\n",
       " ('Q1059081', 0.74788564287585446),\n",
       " ('Q4169', 0.74763828248109099),\n",
       " ('Q16652009', 0.74135312707844747),\n",
       " ('Q221392', 0.74040813066583788),\n",
       " ('Q321', 0.73970201130507318),\n",
       " ('Q5646104', 0.72849855030646837),\n",
       " ('Q4113683', 0.72136321375851276)]"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "wikidata_embedding.most_similar('Q1')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Or we can get query the Wikidata embedding for the nearest neighbors for enwiki 'Universe' article."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('Universe_(disambiguation)', 0.84477422002858837),\n",
       " ('Galaxy', 0.81135711307867753),\n",
       " ('Big_Bang', 0.80632248705977405),\n",
       " ('Cosmos', 0.74788564287585446),\n",
       " ('Outer_space', 0.74763828248109099),\n",
       " ('Observable_universe', 0.74040813066583788),\n",
       " ('Milky_Way', 0.73970201130507318),\n",
       " ('Black_hole', 0.71752815613001775),\n",
       " ('Chronology_of_the_universe', 0.71578467494239983),\n",
       " (\"Earth's_location_in_the_Universe\", 0.71243763700781371),\n",
       " ('Physical_cosmology', 0.70309098687221405),\n",
       " ('Galaxy_formation_and_evolution', 0.70290893139795907),\n",
       " ('Timeline_of_cosmological_theories', 0.70288585448097241),\n",
       " ('Space', 0.70268257298798587),\n",
       " ('Dark_energy', 0.69347522693715991),\n",
       " ('Cosmology', 0.69242340032770677),\n",
       " ('Dark_matter', 0.6857252701601938),\n",
       " ('Observational_cosmology', 0.68504308638673128),\n",
       " ('Metric_expansion_of_space', 0.68413646823318397)]"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "most_similar(wikidata_embedding, 'Universe', 'en', n=30, min_similarity=0.6)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('Universe_(disambiguation)', 0.84477422002858837),\n",
       " ('Galaxy', 0.81135711307867753),\n",
       " ('Big_Bang', 0.80632248705977405),\n",
       " ('Cosmos', 0.74788564287585446),\n",
       " ('Outer_space', 0.74763828248109099),\n",
       " ('Observable_universe', 0.74040813066583788),\n",
       " ('Milky_Way', 0.73970201130507318)]"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "most_similar(wikidata_embedding, 'Universe', 'en')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As a comparison, here are the nearest neighbors for 'Universe' in the enwiki embedding, which was only trained on enwiki reading sessions. The top neighbors are pretty different."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Application Ideas\n",
    "\n",
    "Here are some ideas for how to use these embeddings to improve Wikipedia.\n",
    "\n",
    "#### Translation Recommendations\n",
    "\n",
    "We recently created a [tool](http://recommend.wmflabs.org/#Recommend) for recommending articles for translation between Wikipedias. Users choose a source language to translate from and a target language to translate to. Then they choose a seed article in the source language that represents their interests. We can find articles in the source language missing in the target language using Wikidata sitelinks, and then rank the missing articles according to the similarity between their vectors and the seed vector. \n",
    "\n",
    "#### Reading Recommendations\n",
    "The Reading Team at WMF recently introduced a \"Related Pages\" feature that gives readers 3 recommendations for further reading. The current recommendations are generated by the [More Like This Query](https://www.elastic.co/guide/en/elasticsearch/reference/current/query-dsl-mlt-query.html) feature in Elastic Search.\n",
    "\n",
    "Instead, we could generate recommendations for further reading by looking up the nearest neighbors of the current article the reader is on in an embedding. The beauty of this approach is that the nearest neighbors are by definition articles that tend to be read together. Furthermore, the Wikidata embedding would allow us to use a single model to generate recommendations across all languages!\n",
    "\n",
    "#### Link Recommendations\n",
    "If articles are frequently read within the same session, you might be able to make Wikipedia easier to navigate if you were to create a link between them. For a given article you could generate recommendations for links to add by finding the nearest neighbors that are not already linked and adding a link if the original article has a suitable anchor text. Again, the Wikidata embedding would allow you to build a model for all languages."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('Deborah_Orr', 0.91539459281436186),\n",
       " ('Jonathan_Self', 0.91529695359415975),\n",
       " ('Umbrella_(novel)', 0.88568083130588349),\n",
       " ('Shark_(novel)', 0.88144038836349925),\n",
       " ('Peter_Self', 0.88066891173737261),\n",
       " ('Cock_and_Bull', 0.8700406005395005),\n",
       " ('Dorian,_an_Imitation', 0.85718259988297951),\n",
       " ('Jonathan_Coe', 0.8134598307405887),\n",
       " ('What_a_Carve_Up!_(novel)', 0.80775135057346681),\n",
       " ('The_Book_of_Dave', 0.79721259734964156)]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "en_embedding.most_similar('Will_Self')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('Kissing_traditions', 0.72906072754159257),\n",
       " ('French_kiss', 0.72400639405281908),\n",
       " ('Kissing,_Bavaria', 0.68524468918114323),\n",
       " ('Air_kiss', 0.66284194730918222),\n",
       " ('Hug', 0.6409782198490559),\n",
       " ('Kiss_(band)', 0.62808919174690359),\n",
       " ('Foreplay', 0.62006375083990484),\n",
       " ('Cheek_kissing', 0.5881288871976873),\n",
       " ('Making_out', 0.58465149766396551),\n",
       " ('Post_Office_(game)', 0.5804142777106831)]"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "en_embedding.most_similar('Kiss')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('Phyllis_Newman', 0.87788281050268258),\n",
       " ('Adolph_Green', 0.86793390643667045),\n",
       " ('Bambi_Linn', 0.83127314845549538),\n",
       " ('List_of_Broadway_musicals_stars', 0.82739932678976524),\n",
       " ('Plain_and_Fancy', 0.823583427500724),\n",
       " ('Betty_Comden', 0.82328068158088841),\n",
       " ('Bells_Are_Ringing_(musical)', 0.82092651923716164),\n",
       " ('Marin_Mazzie', 0.82055069146017445),\n",
       " ('I_Love_My_Wife', 0.8199105465494283),\n",
       " ('The_Most_Happy_Fella', 0.81986287660477519)]"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "en_embedding.most_similar('Amanda_Green')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('Lucknow', 0.729475983488725),\n",
       " ('List_of_districts_of_Uttar_Pradesh', 0.71982895178235284),\n",
       " ('Madhya_Pradesh', 0.70149555349502302),\n",
       " ('Varanasi_district', 0.70075913022256842),\n",
       " ('History_of_Uttar_Pradesh', 0.69928461973048661),\n",
       " ('Pradesh', 0.6957367273967574),\n",
       " ('Western_Uttar_Pradesh', 0.68628632668005041),\n",
       " ('Moradabad_district', 0.68589711236395601),\n",
       " ('Varanasi_division', 0.6843656500536035),\n",
       " ('Purvanchal', 0.68426054961414429)]"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "en_embedding.most_similar('Uttar_Pradesh')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}