3-nlp/typical_methods.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 代表的な自然言語処理（日本語を例に）\n",
    "- [補足説明](./typical_methods.md)\n",
    "- mecabによる形態素解析\n",
    "    - 出力オプションの例\n",
    "    - ノード単位で処理し、原型取得する例\n",
    "- NLTKの利用\n",
    "    - KWIC（もしくはKWIC index）\n",
    "    - コロケーション（collocations）\n",
    "    - 単語の出現頻度順\n",
    "    - 単語の条件付き出現頻度\n",
    "- KNPによる係り受け解析\n",
    "    - 文節単位での処理例\n",
    "    - 係り受けを利用したクエリ検索例"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "すもも\t名詞,一般,*,*,*,*,すもも,スモモ,スモモ\n",
      "も\t助詞,係助詞,*,*,*,*,も,モ,モ\n",
      "もも\t名詞,一般,*,*,*,*,もも,モモ,モモ\n",
      "も\t助詞,係助詞,*,*,*,*,も,モ,モ\n",
      "もも\t名詞,一般,*,*,*,*,もも,モモ,モモ\n",
      "の\t助詞,連体化,*,*,*,*,の,ノ,ノ\n",
      "うち\t名詞,非自立,副詞可能,*,*,*,うち,ウチ,ウチ\n",
      "EOS\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# 形態素解析\n",
    "import MeCab\n",
    "tagger = MeCab.Tagger()\n",
    "print(tagger.parse(\"すもももももももものうち\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "すもも\tスモモ\tすもも\t名詞-一般\t\t\n",
      "も\tモ\tも\t助詞-係助詞\t\t\n",
      "もも\tモモ\tもも\t名詞-一般\t\t\n",
      "も\tモ\tも\t助詞-係助詞\t\t\n",
      "もも\tモモ\tもも\t名詞-一般\t\t\n",
      "の\tノ\tの\t助詞-連体化\t\t\n",
      "うち\tウチ\tうち\t名詞-非自立-副詞可能\t\t\n",
      "EOS\n",
      "\n",
      "----\n",
      "すもも も もも も もも の うち \n",
      "\n"
     ]
    }
   ],
   "source": [
    "# MeCab.Tagger() 時の出力オプションで、パース結果が変わる。\n",
    "# parse() は単に文字列出力する。オプションは mecab --help で調べよう。\n",
    "tagger = MeCab.Tagger('-Ochasen')\n",
    "print(tagger.parse(\"すもももももももものうち\"))\n",
    "\n",
    "print('----')\n",
    "\n",
    "tagger = MeCab.Tagger('-Owakati')\n",
    "print(tagger.parse(\"すもももももももものうち\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "node.surface=, node.feature=BOS/EOS,*,*,*,*,*,*,*,* => original = *\n",
      "node.surface=おじいさん, node.feature=名詞,一般,*,*,*,*,おじいさん,オジイサン,オジーサン => original = おじいさん\n",
      "node.surface=は, node.feature=助詞,係助詞,*,*,*,*,は,ハ,ワ => original = は\n",
      "node.surface=山, node.feature=名詞,一般,*,*,*,*,山,ヤマ,ヤマ => original = 山\n",
      "node.surface=へ, node.feature=助詞,格助詞,一般,*,*,*,へ,ヘ,エ => original = へ\n",
      "node.surface=しば, node.feature=名詞,一般,*,*,*,*,しば,シバ,シバ => original = しば\n",
      "node.surface=刈, node.feature=名詞,一般,*,*,*,*,刈,カリ,カリ => original = 刈\n",
      "node.surface=かりに, node.feature=副詞,一般,*,*,*,*,かりに,カリニ,カリニ => original = かりに\n",
      "node.surface=、, node.feature=記号,読点,*,*,*,*,、,、,、 => original = 、\n",
      "node.surface=おばあさん, node.feature=名詞,一般,*,*,*,*,おばあさん,オバアサン,オバーサン => original = おばあさん\n",
      "node.surface=は, node.feature=助詞,係助詞,*,*,*,*,は,ハ,ワ => original = は\n",
      "node.surface=川, node.feature=名詞,一般,*,*,*,*,川,カワ,カワ => original = 川\n",
      "node.surface=へ, node.feature=助詞,格助詞,一般,*,*,*,へ,ヘ,エ => original = へ\n",
      "node.surface=洗濯, node.feature=名詞,サ変接続,*,*,*,*,洗濯,センタク,センタク => original = 洗濯\n",
      "node.surface=せ, node.feature=動詞,自立,*,*,サ変・スル,未然ヌ接続,する,セ,セ => original = する\n",
      "node.surface=ん, node.feature=助動詞,*,*,*,不変化型,基本形,ん,ン,ン => original = ん\n",
      "node.surface=たく, node.feature=助動詞,*,*,*,特殊・タイ,連用テ接続,たい,タク,タク => original = たい\n",
      "node.surface=に, node.feature=助詞,格助詞,一般,*,*,*,に,ニ,ニ => original = に\n",
      "node.surface=行き, node.feature=動詞,自立,*,*,五段・カ行促音便,連用形,行く,イキ,イキ => original = 行く\n",
      "node.surface=まし, node.feature=助動詞,*,*,*,特殊・マス,連用形,ます,マシ,マシ => original = ます\n",
      "node.surface=た, node.feature=助動詞,*,*,*,特殊・タ,基本形,た,タ,タ => original = た\n",
      "node.surface=。, node.feature=記号,句点,*,*,*,*,。,。,。 => original = 。\n",
      "node.surface=, node.feature=BOS/EOS,*,*,*,*,*,*,*,* => original = *\n"
     ]
    }
   ],
   "source": [
    "# parseToNode() を使うと、ノード単位で処理しやすい。\n",
    "import MeCab\n",
    "tagger = MeCab.Tagger()\n",
    "\n",
    "sentence = \"おじいさんは山へしば刈かりに、おばあさんは川へ洗濯せんたくに行きました。\"\n",
    "node = tagger.parseToNode(sentence)\n",
    "while node:\n",
    "   features = node.feature.split(\",\")\n",
    "   original = features[6]\n",
    "   print('node.surface={}, node.feature={} => original = {}'.format(node.surface,node.feature,original))\n",
    "   node = node.next\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "# nltk.Text.concordance with \"おじいさん\"\n",
      "Displaying 17 of 17 matches:\n",
      "むかし 、 むかし 、 ある ところ に 、 おじいさん と おばあさん が あり まし た 。 まいにち 、 おじいさん は 山\n",
      "に 、 おじいさん と おばあさん が あり まし た 。 まいにち 、 おじいさん は 山 へ しば刈り に 、 おばあさん は 川 へ 洗濯 に 行き ま\n",
      " 来 まし た 。 おやおや 、 これ は みごと な 桃 だ こと 。 おじいさん へ の おみやげ に 、 どれどれ 、 うち へ 持っ て 帰り ましょ\n",
      "い ながら 、 おばあさん の 前 へ 流れ て 来 まし た 。 早く おじいさん と 二人 で 分けて 食べ ましょ う 。 と 言っ て 、 桃 を ひ\n",
      " て お うち へ 帰り まし た 。 夕方 に なっ て やっ と 、 おじいさん は 山 から しば を 背負っ て 帰っ て 来 まし た 。 おばあさ\n",
      "っ て 来 まし た 。 おばあさん 、 今 帰っ た よ 。 おや 、 おじいさん 、 おか いん なさい 。 待っ て い まし た よ 。 さあ 、 早\n",
      "だ ね 、 その いい もの という の は 。 こう いい ながら 、 おじいさん は わらじ を ぬい で 、 上 に 上がり まし た 。 ほら 、 ご\n",
      "赤さん が 元気 よく とび出 し まし た 。 おやおや 、 まあ 。 おじいさん も 、 おばあさん も 、 びっくり し て 、 二人 いっしょ に 声\n",
      " 神さま が この 子 を さずけ て 下さっ た に ちがい ない 。 おじいさん も 、 おばあさん も 、 うれし がっ て 、 こう 言い まし た \n",
      " 、 うれし がっ て 、 こう 言い まし た 。 そこで あわて て おじいさん が お湯 を わかす やら 、 おばあさん が むつき を そろえる や\n",
      "いう ので 、 この 子 に 桃太郎 という 名 を つけ まし た 。 おじいさん と おばあさん は 、 それ は それ は だいじ に し て 桃太郎 \n",
      "くらい でし た が 、 その くせ 気だて はごく やさしくっ て 、 おじいさん と おばあさん に よく 孝行 を し まし た 。 桃太郎 は 十五 \n",
      " それ は いさましい こと だ 。 じゃあ 行っ て おい で 。 と おじいさん は 言い まし た 。 まあ 、 そんな 遠方 へ 行く の で は 、\n",
      "。 じゃあ 、 りっぱ に 鬼 を 退治 し て くる が いい 。 と おじいさん は 言い まし た 。 気 を つけ て 、 けが を し ない よう \n",
      "う 。 と 元気 な 声 を のこし て 、 出 て いき まし た 。 おじいさん と おばあさん は 、 門 の 外 に 立っ て 、 いつまでも 、 い\n",
      " もと くい らしい 様子 を し て 帰っ て 来 まし た ので 、 おじいさん も おばあさん も 、 目 も 鼻 も なくし て 喜び まし た 。 \n",
      " た 。 えらい ぞ 、 えらい ぞ 、 それ こそ 日本一 だ 。 と おじいさん は 言い まし た 。 まあ 、 まあ 、 けが が なくっ て 、 何\n"
     ]
    }
   ],
   "source": [
    "# concordance\n",
    "\n",
    "import nltk\n",
    "\n",
    "# データ準備\n",
    "# https://newtechnologylifestyle.net/711-2/\n",
    "# momotaro.txtは、上記を使ってルビを除外したもの。\n",
    "filename = \"./corpus/momotaro.txt\"\n",
    "with open(filename, \"r\") as fh:\n",
    "    sentences = \"\"\n",
    "    for line in fh.readlines():\n",
    "        sentences += line + \" \"\n",
    "\n",
    "def tokenize(sentences):\n",
    "    \"\"\"文章を分かち書きする。\n",
    "\n",
    "    :param sentences(str): 複数の文章を含む文字列。日本語想定。\n",
    "    :return(list): 分かち書きした単語をlistとしてまとめたもの。\n",
    "    \"\"\"\n",
    "\n",
    "    # 「。」、「！」、「？」で終わる一連の文字列を文として認識し分割する。\n",
    "    jp_sent_tokenizer = nltk.RegexpTokenizer(u'[^　「」！？。]*[！？。]')\n",
    "    tagger = MeCab.Tagger('-Owakati -d /usr/local/lib/mecab/dic/mecab-ipadic-neologd')\n",
    "\n",
    "    sents = jp_sent_tokenizer.tokenize(sentences)\n",
    "    tokens = []\n",
    "    for sent in sents:\n",
    "        for word in tagger.parse(sent).split(\" \"):\n",
    "            if word == \" \" or word == \"\\n\":\n",
    "                continue\n",
    "            tokens.append(word)\n",
    "    return tokens\n",
    "\n",
    "tokens = tokenize(sentences)\n",
    "text = nltk.Text(tokens)\n",
    "print('\\n# nltk.Text.concordance with \"おじいさん\"')\n",
    "text.concordance(\"おじいさん\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "# nltk.Text.collocations\n",
      "## default\n",
      "ドンブラ コッコ; いかめしい くろがね; さけび ながら\n",
      "\n",
      "## collocations(window_size=5)\n",
      "コッコ スッコッコ; ドンブラ コッコ; ドンブラ スッコッコ; おじいさん おばあさん; どちら おいで; 鬼が島 せいばつ; 日本一\n",
      "きびだんご; ござい 日本一; こっこ こっこ; いつまでも 見送っ; ぎゅうぎゅう 押さえつけ; たずね 鬼が島; きびだんご もらっ;\n",
      "いかめしい くろがね; 下さい ましょ; きびだんご 下さい; 見える 見える; キャッ キャッ; コッコ ドンブラ; スッコッコ コッコ\n"
     ]
    }
   ],
   "source": [
    "# collocation\n",
    "\n",
    "print(\"\\n# nltk.Text.collocations\")\n",
    "print(\"## default\")\n",
    "text.collocations()\n",
    "\n",
    "print(\"\\n## collocations(window_size=5)\")\n",
    "text.collocations(window_size=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "# FreqDist\n",
      "[('、', 252), ('。', 178), ('た', 119), ('て', 116), ('まし', 92), ('に', 87), ('を', 82), ('と', 76), ('の', 75), ('は', 73), ('が', 46), ('で', 29), ('桃太郎', 29), ('し', 27), ('も', 27), ('へ', 21), ('から', 20), ('だ', 19), ('おじいさん', 17), ('おばあさん', 16), ('う', 16), ('ます', 16), ('鬼', 16), ('言い', 15), ('来', 13), ('ながら', 13), ('お', 12), ('い', 12), ('な', 10), ('もの', 10)]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 単語の出現頻度順\n",
    "\n",
    "print(\"\\n# FreqDist\")\n",
    "fdist = nltk.FreqDist(tokens)\n",
    "check_num = 30\n",
    "print(fdist.most_common(check_num))\n",
    "\n",
    "%matplotlib inline\n",
    "import matplotlib\n",
    "import matplotlib.font_manager as font_manager\n",
    "#TTFファイル(フォント)のアドレスを指定\n",
    "font_path = '/Library/Fonts/Arial Unicode.ttf'\n",
    "#フォントの詳細情報を取得\n",
    "font_prop = font_manager.FontProperties(fname = font_path)\n",
    "#フォントの名前を使い、matplotのデフォルトフォントに指定\n",
    "matplotlib.rcParams['font.family'] = font_prop.get_name()\n",
    "\n",
    "fdist.plot(check_num)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "# FreqDist with bigram (2-gram)\n",
      "[(('まし', 'た'), 92), (('た', '。'), 91), (('て', '、'), 48), (('。', 'と'), 31), (('は', '、'), 24), (('と', '、'), 17), (('し', 'て'), 14), (('。', '桃太郎'), 14), (('に', '、'), 13), (('ながら', '、'), 13), (('う', '。'), 11), (('言い', 'まし'), 11), (('桃太郎', 'は'), 11), (('て', 'い'), 10), (('ます', '。'), 10), (('て', '来'), 9), (('ましょ', 'う'), 9), (('で', '、'), 9), (('来', 'まし'), 8), (('を', 'し'), 8), (('、', '桃太郎'), 8), (('、', 'おじいさん'), 7), (('と', 'おばあさん'), 7), (('言っ', 'て'), 7), (('も', '、'), 7), (('から', '、'), 7), (('に', 'なり'), 7), (('だ', '。'), 7), (('よし', '、'), 7), (('桃太郎', 'さん'), 7)]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# bigram\n",
    "print(\"\\n# FreqDist with bigram (2-gram)\")\n",
    "bigrams = nltk.bigrams(tokens)\n",
    "bigrams_fdist = nltk.FreqDist(bigrams)\n",
    "print(bigrams_fdist.most_common(check_num))\n",
    "\n",
    "# conditional FreqDist\n",
    "cfd = nltk.ConditionalFreqDist(bigrams_fdist)\n",
    "cfd.plot(conditions=[\"桃太郎\", \"鬼\"])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# S-ID:1 KNP:4.19-CF1.1 DATE:2019/10/31 SCORE:-1054.00784\r\n",
      "お爺さんは─────┐　　　　　　　　　<体言>\r\n",
      "      山へ─────┤　　　　　　　　　<体言>\r\n",
      "        芝刈りに、<P>──┐　　　　　<体言>\r\n",
      "お婆さんは─────┐　　　 │　　　　　<体言>\r\n",
      "      川へ─────┤　　　 │　　　　　<体言>\r\n",
      "            洗濯に<P>─PARA─────┐　<体言><用言:動><格解析結果:ガ/婆さん;ニ/川>\r\n",
      "                      行きました。<用言:動><格解析結果:ガ/-;ヲ/-;ニ/-;デ/-;カラ/-;マデ/-;ヘ/-;時間/-;ノ/->\r\n",
      "EOS\r\n"
     ]
    }
   ],
   "source": [
    "# KNP on terminal command\n",
    "\n",
    "!echo \"お爺さんは山へ芝刈りに、お婆さんは川へ洗濯に行きました。\" | juman | knp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "文節単位での処理\n",
      "\tID:0, 見出し:お爺さんは, 係り受けタイプ:D, 親文節ID:2, 素性:<文頭><ハ><助詞><体言><係:未格><提題><区切:3-5><主題表現><格要素><連用要素><正規化代表表記:爺さん/じいさん><主辞代表表記:爺さん/じいさん>\n",
      "\tID:1, 見出し:山へ, 係り受けタイプ:D, 親文節ID:2, 素性:<ヘ><助詞><体言><一文字漢字><係:ヘ格><区切:0-0><格要素><連用要素><正規化代表表記:山/やま?山/さん><主辞代表表記:山/やま?山/さん>\n",
      "\tID:2, 見出し:芝刈りに、, 係り受けタイプ:P, 親文節ID:5, 素性:<サ変><ニ><読点><助詞><体言><係:ニ格><並キ:名:&ST:4.0&&ニ><区切:0-4><読点並キ><並列タイプ:？><格要素><連用要素><正規化代表表記:芝刈り/しばかり><主辞代表表記:芝刈り/しばかり><並列類似度:4.545><並結句数:2><並結文節数:3>\n",
      "\tID:3, 見出し:お婆さんは, 係り受けタイプ:D, 親文節ID:5, 素性:<SM-主体><SM-人><人名><ハ><助詞><体言><係:未格><提題><区切:3-5><主題表現><格要素><連用要素><正規化代表表記:御祖母/おばあ><主辞代表表記:御祖母/おばあ>\n",
      "\tID:4, 見出し:川へ, 係り受けタイプ:D, 親文節ID:5, 素性:<ヘ><助詞><体言><一文字漢字><係:ヘ格><区切:0-0><格要素><連用要素><正規化代表表記:川/かわ><主辞代表表記:川/かわ>\n",
      "\tID:5, 見出し:洗濯に, 係り受けタイプ:D, 親文節ID:6, 素性:<サ変><ニ><助詞><体言><係:連用><レベル:A><区切:0-5><ID:〜に><用言:動><サ変動詞化><連用要素><連用節><動態述語><正規化代表表記:洗濯/せんたく><主辞代表表記:洗濯/せんたく>\n",
      "\tID:6, 見出し:行きました。, 係り受けタイプ:D, 親文節ID:-1, 素性:<文末><時制-過去><句点><用言:動><レベル:C><区切:5-5><ID:（文末）><係:文末><提題受:30><主節><格要素><連用要素><動態述語><敬語:丁寧表現><正規化代表表記:行く/いく><主辞代表表記:行く/いく>\n"
     ]
    }
   ],
   "source": [
    "# KNP：文節単位での処理\n",
    "\n",
    "from pyknp import KNP\n",
    "\n",
    "knp = KNP()\n",
    "sentence = \"お爺さんは山へ芝刈りに、お婆さんは川へ洗濯に行きました。\"\n",
    "result = knp.parse(sentence)\n",
    "\n",
    "print(\"文節単位での処理\")\n",
    "for bnst in result.bnst_list(): # 各文節へのアクセス\n",
    "    print(\"\\tID:%d, 見出し:%s, 係り受けタイプ:%s, 親文節ID:%d, 素性:%s\" \\\n",
    "            % (bnst.bnst_id, \"\".join(mrph.midasi for mrph in bnst.mrph_list()), bnst.dpndtype, bnst.parent_id, bnst.fstring))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "# user_query = 芝刈り\n",
      "child: id=0, surface=お爺さんは\n",
      "child: id=1, surface=山へ\n"
     ]
    }
   ],
   "source": [
    "# 係り受けを利用したクエリ検索例\n",
    "\n",
    "def find_children(parse_result, parent_id):\n",
    "    \"\"\"KNPパース結果に対し、parent_id以下の子ノード情報を取得する。\n",
    "\n",
    "    :param parse_result(pyknp.knp.blist.BList): KNP.parse()した結果。\n",
    "    :param parent_id(int): 出発点となるID。\n",
    "    :return:\n",
    "      children_ids(list): parent_idを親とする子ノードのID一覧。\n",
    "      children_surfaces(list): children_idsに対応したsurface一覧。\n",
    "    \"\"\"\n",
    "    children_ids = []\n",
    "    children_surfaces = []\n",
    "    for child_id, bnst in enumerate(parse_result.bnst_list()):\n",
    "        if bnst.parent_id == parent_id:\n",
    "            children_ids.append(child_id)\n",
    "            surface = \"\".join(mrph.midasi for mrph in bnst.mrph_list())\n",
    "            children_surfaces.append(surface)\n",
    "    return children_ids, children_surfaces\n",
    "\n",
    "def search(parse_result, user_query):\n",
    "    \"\"\"KNPパース結果に対し、ユーザクエリに該当するノード情報を取得する。\n",
    "    複数該当ノードがある場合には、最初に見つかったノードのみを返す。\n",
    "\n",
    "    :param parse_result(pyknp.knp.blist.BList): KNP.parse()した結果。\n",
    "    :param user_query(str): 探したい単語。\n",
    "    :return:\n",
    "      children_ids(list): parent_idを親とする子ノードのID一覧。\n",
    "      children_surfaces(list): children_idsに対応したsurface一覧。\n",
    "    \"\"\"\n",
    "    for id, bnst in enumerate(result.bnst_list()):\n",
    "        surface = \"\".join(mrph.midasi for mrph in bnst.mrph_list())\n",
    "        if user_query in surface:\n",
    "            children_ids, children_surfaces = find_children(result, id)\n",
    "            return children_ids, children_surfaces\n",
    "    return None, None\n",
    "\n",
    "user_query = \"芝刈り\"\n",
    "print('\\n# user_query = {}'.format(user_query))\n",
    "\n",
    "children_ids, children_surfaces = search(result, user_query)\n",
    "for i in range(len(children_ids)):\n",
    "    print(\"child: id={}, surface={}\".format(children_ids[i], children_surfaces[i]))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "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": "3.7.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}