korean_texttiling.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import os\n",
    "import glob\n",
    "import re\n",
    "import textiling\n",
    "import segmentation\n",
    "import nltk\n",
    "from sentence_transformers import SentenceTransformer\n",
    "\n",
    "import konlpy\n",
    "from konlpy.tag import Okt\n",
    "import matplotlib.pylab as plt\n",
    "%matplotlib inline\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "dir_path = \"./texts/\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_data(name, version):\n",
    "    path = dir_path + name + \"_\" + version + \"_data.txt\"\n",
    "    text = \"\"\n",
    "    with open(path, \"r\") as f:\n",
    "        lines = f.readlines()\n",
    "\n",
    "    for line in lines:\n",
    "        text += line\n",
    "\n",
    "    return text\n",
    "\n",
    "kakao_origin = load_data(\"kakao\", \"origin\")\n",
    "kakao_stem = load_data(\"kakao\", \"stem\")\n",
    "band_origin = load_data(\"band\", \"origin\")\n",
    "band_stem = load_data(\"band\", \"stem\")\n",
    "nate_origin = load_data(\"nate\", \"origin\")\n",
    "nate_stem = load_data(\"nate\", \"stem\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "tt = textiling.TextTilingTokenizer()\n",
    "kakao_cos_tt = tt.tokenize(kakao_origin, mode = \"cos\")\n",
    "kakao_origin_tt = tt.tokenize(kakao_stem, mode = \"origin\")\n",
    "band_cos_tt = tt.tokenize(band_origin, mode = \"cos\")\n",
    "band_origin_tt = tt.tokenize(band_stem, mode = \"origin\")\n",
    "nate_cos_tt = tt.tokenize(nate_origin, mode = \"cos\")\n",
    "nate_origin_tt = tt.tokenize(nate_stem, mode = \"origin\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "희 집 근처에는 병원 있어?\n",
      "응, 난 엄청 근처에 있어.\n",
      "나는 차 타고 20분 거리에 있어.\n",
      "집 근처에 병원이 있으면 편리한 거 같아.\n",
      "왜 그렇게 생각하는 거야?\n",
      "언제 사고가 발생할지 모르잖아.\n",
      "사고가 나면 병원을 빨리 가는 게 중요해.\n",
      "나는 집 근처에 병원이 있어서 다행이야.\n",
      "20분 정도 거리면 나도 나름 가까운 거 같아.\n",
      "다음에 이사 가면 근처에 병원 있는지 꼭 확인해야겠어.\n",
      "------------------------------------------------------------------------------------------------------------------------------------------------------\n",
      "나도 참고해야지.\n",
      "미리 대비하는 모습 좋은 거 같아.\n",
      "나는 경유차가 첫차였어\n",
      "아 노후 경유차였구나\n",
      "노후 경유차면 운행이 쉽진 않지\n",
      "서울은 여기저기 감시 카메라가 많아가지고 끌고 다니기 힘들었어\n",
      "그러니깐 요즘은 블랙박스로 찍어서 신고도 많이하니깐\n",
      "요즘은 진짜 차 끌고 다닐때 조심해야돼\n",
      "맞어 블랙박스가 있어서 너무 좋으면서도 조심 또 조심해야돼\n",
      "너네는 단속카메라에 걸린적 없어?\n",
      "나 너무 많지 딱지 날라올때마다 피눈물이\n",
      "나도 주정차 단속이랑 과속 한번씩 있는데 진짜 눈물나\n",
      "나도 어디 갔다가 주차할때가 없어서 잠깐 세웠는데 과태료 날라오더라\n",
      "차가 편하기도 하면서 더 불편한 부분도 있어\n",
      "근데 단속 중이면 문자오는건가 ? 그런것도 있어\n",
      "그게 뭐야? 단소을 알려준다고?\n",
      "그런게 있어? 찾아봐야겠다\n",
      "너희 일할 때 mbti 뭐야?\n",
      "나는 해봤는데 ENTJ야.\n",
      "J들이 꼼꼼하고 계획적이긴 하지\n",
      "회사에서 칭찬 받겠다\n",
      "나는 덤벙대는 성격이라 부럽다\n",
      "우리 회사도 그런 사람 몇 명 있어\n",
      "------------------------------------------------------------------------------------------------------------------------------------------------------\n",
      "꼼꼼함이 장점인 사람이 존경스럽긴 하지\n",
      "열린 마인드라 좋았어.\n",
      "요즘은 꼰대 기준 나이도 낮더라\n",
      "내가 꼰대인가? 하는 순간 꼰대래\n",
      "모두가 기성세대가 되니까\n",
      "특히 회사에서 세대차이는 극복하기 어려운 것 같아\n",
      "아무래도 그렇게 모일 일이 없으니까\n",
      "맞아 인위적으로 모인 조직이지 회사는.\n",
      "어떻게 하면 덜 꼰대가 될 수 있을지 고민이다 참\n",
      "군대는 몇살부터 가는 걸까요?\n",
      "만 19세 이상부터 가는 걸로 알고 있는데\n",
      "20세부터 입영통지서 오는 거 아닌가요? 정확히는 모르겠네요.\n",
      "그럼 몇살까지 갈수 있어요?\n",
      "그렇게 세세하게 알고 싶지는 않아요 군대란 곳은\n",
      "28세 아닌가요. 모르겠네요. 알 수 없어요.\n",
      "갑자기 몇살부터 몇살까지 갈수 있는지 궁금했어요\n",
      "나이 맞으면 군대 갈 생각 있으신가 봐요\n",
      "누가 갈 사람이 있으신가봐요. 저도 가고 싶어요.\n",
      "네 여자이지만 가보고 싶습니다.\n",
      "키가 작어도 가기가 어려울 걸요\n",
      "키제한도 있군요. 몸은 힘들겠지만 아무 생각 안 할수 있을 것 같아요.\n",
      "이번에 수학 수행평가 뭔지 알아?\n",
      "들었었는데 까먹었어.\n",
      "코사인 법칙 증명하는 거일 거야.\n",
      "아, 나 이런 거 잘 못하는데.\n",
      "아 맞다, 증명하는 거였지.\n",
      "나도 이런 거 잘 못해.\n",
      "그냥 문제 푸는 게 더 나은데.\n",
      "맞아, 나도 그렇게 생각해.\n",
      "수학은 너무 어려운 거 같아.\n",
      "수행평가부터 너무 어려워 죽겠어.\n",
      "수학 선생님께 문제 푸는 걸로 바꿀 수 있는지 한 번 물어볼까?\n",
      "응, 물어봐. 제발 바뀌었으면 좋겠어.\n",
      "인정해, 증명하는 거는 너무 힘들고 어려워.\n",
      "------------------------------------------------------------------------------------------------------------------------------------------------------\n",
      "갑자기 할머니가 너무 보고 싶어요\n",
      "살아 계실때 잘할껄 후회중 입니다.\n",
      "저는 우리 어머니가 너무 보고 싶네요\n",
      "저도 너무 보고 싶어요. 할아버지도 보고 싶고요.\n",
      "살아 계실때는 평생 계실 것만 같아서 모르고 살았네요\n",
      "돌아가신 지 10년 정도 됐는데 너무 너무 보고 싶네요\n",
      "아직도 할머니가 해주신 김치가 남아있어요.\n",
      "네 저도 너무너무 그분들이 그립습니다.\n",
      "살아계실 때 사랑한단 말 좀 많이 할 걸 그랬어요\n",
      "할머니 음식도 너무 그리워요.\n",
      "------------------------------------------------------------------------------------------------------------------------------------------------------\n",
      "저도 저희 엄마 아빠께 사랑한다 말을 자주 해야 할 것 같아요\n",
      "그래서 아버지한테 많이 하고 있어요\n",
      "저도 자주 하고 있답니다. 오늘도 하고 왔어요.\n",
      "연애할 때 자주 보지 않아?\n",
      "우리 장거리라서 주말 밖에 못 봐\n",
      "헉 힘들겠다 장거리연애…\n",
      "주말 한 주 라도 일있으면 2주 만에 보네?\n",
      "응 맞아 얼마나 길게 느껴지는지\n",
      "헉 나도 장거리 해봐서 그 기분 알지\n",
      "얼마나 장거리 였는데?\n",
      "------------------------------------------------------------------------------------------------------------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "# Segment example 1 <= n1 < n2 <= 20\n",
    "def print_seg(tt, n1, n2):\n",
    "    n2 = min(len(tt), n2)\n",
    "    for i in range(n1-1, n2+1):\n",
    "        print(tt[i])\n",
    "        print(\"-\"*150)\n",
    "\n",
    "print_seg(band_cos_tt, 1, 4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "def origin_pred(tt):\n",
    "    origin_label = [0]\n",
    "    for origin_t in tt:\n",
    "        origin_label.append(origin_label[-1] + len(origin_t.strip().split(\"\\n\\n\\t\")))\n",
    "\n",
    "    return origin_label[1:]\n",
    "\n",
    "def cos_pred(tt):\n",
    "    cos_label = [0]\n",
    "    for cos_t in tt:\n",
    "        cos_label.append(cos_label[-1] + len(cos_t.split(\"\\n\")))\n",
    "\n",
    "    return cos_label[1:]\n",
    "\n",
    "kakao_origin_pred = origin_pred(kakao_origin_tt)\n",
    "kakao_cos_pred = cos_pred(kakao_cos_tt)\n",
    "band_origin_pred = origin_pred(band_origin_tt)\n",
    "band_cos_pred = cos_pred(band_cos_tt)\n",
    "nate_origin_pred = origin_pred(nate_origin_tt)\n",
    "nate_cos_pred = cos_pred(nate_cos_tt)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_label(data_name):\n",
    "    with open(dir_path + data_name+\"_label.txt\") as f:\n",
    "        label = [int(x) for x in f.readline().strip().split(\" \")]\n",
    "    return label\n",
    "\n",
    "kakao_label = load_label(\"kakao\")\n",
    "band_label = load_label(\"band\")\n",
    "nate_label = load_label(\"nate\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "def draw_graph(label, cos_pred, origin_pred):\n",
    "    plt.figure(figsize=(20,2))\n",
    "    for i in (label):\n",
    "        plt.axvline(x = i, color=\"r\", linewidth = 1)\n",
    "    for i in (cos_pred):\n",
    "        plt.axvline(x = i, color=\"b\", linewidth = 1)\n",
    "    for i in (origin_pred):\n",
    "        plt.axvline(x = i, color=\"y\", linewidth = 1)\n",
    "    plt.show()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Red line -> label  \n",
    "Blue line -> cosine similarity method  \n",
    "Yellow line -> original method"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 2000x200 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "draw_graph(kakao_label, kakao_cos_pred, kakao_origin_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 2000x200 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "draw_graph(band_label, band_cos_pred, band_origin_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 2000x200 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "draw_graph(nate_label, nate_cos_pred, nate_origin_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "original texttiling kakao data score 0.30783582089552236\n",
      "cos similarity texttiling kakao data score 0.2891791044776119\n",
      "original texttiling band data scroe 0.3208955223880597\n",
      "cos similarity texttiling band data score 0.2947761194029851\n",
      "original texttiling nate data scroe 0.38076923076923075\n",
      "cos similarity texttiling nate data score 0.35384615384615387\n"
     ]
    }
   ],
   "source": [
    "def evaluate_windowdiff(label, pred, n):\n",
    "    label_s = \"\"\n",
    "    pred_s = \"\"\n",
    "    for i in range(max(label[-1], pred[-1])+1):\n",
    "        if i in label:\n",
    "            label_s += \"1\"\n",
    "        else:\n",
    "            label_s += \"0\"\n",
    "\n",
    "        if i in pred:\n",
    "            pred_s += \"1\"\n",
    "        else:\n",
    "            pred_s += \"0\"\n",
    "\n",
    "    return segmentation.windowdiff(label_s, pred_s, n)\n",
    "\n",
    "print(f'original texttiling kakao data score {evaluate_windowdiff(kakao_label, kakao_origin_pred, 3)}')\n",
    "print(f'cos similarity texttiling kakao data score {evaluate_windowdiff(kakao_label, kakao_cos_pred, 3)}')\n",
    "print(f'original texttiling band data scroe {evaluate_windowdiff(band_label, band_origin_pred, 3)}')\n",
    "print(f'cos similarity texttiling band data score {evaluate_windowdiff(band_label, band_cos_pred, 3)}')\n",
    "print(f'original texttiling nate data scroe {evaluate_windowdiff(nate_label, nate_origin_pred, 3)}')\n",
    "print(f'cos similarity texttiling nate data score {evaluate_windowdiff(nate_label, nate_cos_pred, 3)}')"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.8.2 64-bit ('3.8.2')",
   "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": "3.8.2"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "b53045254e8a726d67e9aedd93b4694c7b0ba470b0e3efc589a91bac98bd4be9"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}