handwriting_ewd.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-03T14:05:37.765541Z",
     "start_time": "2019-06-03T14:05:37.000249Z"
    }
   },
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "%reload_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-03T14:05:41.237342Z",
     "start_time": "2019-06-03T14:05:37.795090Z"
    },
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/adamschiller/anaconda3/envs/fastai-cpu/lib/python3.6/site-packages/sklearn/ensemble/weight_boosting.py:29: DeprecationWarning: numpy.core.umath_tests is an internal NumPy module and should not be imported. It will be removed in a future NumPy release.\n",
      "  from numpy.core.umath_tests import inner1d\n"
     ]
    }
   ],
   "source": [
    "from fastai.text import *\n",
    "# import pyperclip"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-03T14:05:41.357962Z",
     "start_time": "2019-06-03T14:05:41.304994Z"
    }
   },
   "outputs": [],
   "source": [
    "PATH = Path('data/handwriting_images/')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "heading_collapsed": true
   },
   "source": [
    "# EWD speech"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-03T14:07:24.159811Z",
     "start_time": "2019-06-03T14:07:24.011485Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "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>filename</th>\n",
       "      <th>text</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>102</th>\n",
       "      <td>EWD_103.png</td>\n",
       "      <td>Reluctantly admitting that it could perhaps\\nb...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>103</th>\n",
       "      <td>EWD_104.png</td>\n",
       "      <td>Needless to say, that\\nkite won't fly either. ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>104</th>\n",
       "      <td>EWD_105.png</td>\n",
       "      <td>that very soon thereafter he makes the excitin...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>105</th>\n",
       "      <td>EWD_106.png</td>\n",
       "      <td>Teaching to unsuspecting youngsters the\\neffec...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>106</th>\n",
       "      <td>EWD_107.png</td>\n",
       "      <td>To my taste\\nand style, that is what education...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        filename                                               text\n",
       "102  EWD_103.png  Reluctantly admitting that it could perhaps\\nb...\n",
       "103  EWD_104.png  Needless to say, that\\nkite won't fly either. ...\n",
       "104  EWD_105.png  that very soon thereafter he makes the excitin...\n",
       "105  EWD_106.png  Teaching to unsuspecting youngsters the\\neffec...\n",
       "106  EWD_107.png  To my taste\\nand style, that is what education..."
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "src_path = PATH/'ewd'\n",
    "\n",
    "CSV = PATH/'ewd.csv'\n",
    "df = pd.read_csv(CSV)\n",
    "df.tail()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-30T17:10:37.616745Z",
     "start_time": "2019-05-30T17:10:37.564214Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "def yrange(n):\n",
    "    i = 14\n",
    "    while i < n:\n",
    "        yield i\n",
    "        i += 1\n",
    "        \n",
    "y = yrange(107)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 187,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-30T18:39:03.578761Z",
     "start_time": "2019-05-30T18:39:03.144296Z"
    },
    "hidden": true,
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "100\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x1c3d2e84a8>"
      ]
     },
     "execution_count": 187,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1296x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "i = next(y)\n",
    "print(i)\n",
    "row = df.iloc[i]\n",
    "# !printf '{row.text}' | pbcopy\n",
    "pyperclip.copy(row.text)\n",
    "im = Image.open(src_path/row.filename)\n",
    "fig,ax = plt.subplots(figsize=(18,5))\n",
    "ax.imshow(im)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 186,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-30T18:39:01.294338Z",
     "start_time": "2019-05-30T18:39:01.241522Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "row.text = '''effective solutions. So, let us try again.\n",
    "You don't need to take my proposal seriously\n",
    "because it is utterly unrealistic to try to teach\n",
    "such material to college freshmen. Wouldn't that\n",
    "be an easy way out? You just postulate that\n",
    "this would be far too difficult.'''"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 237,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-31T22:06:31.430423Z",
     "start_time": "2019-05-31T22:06:31.371576Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "df.to_csv(CSV, columns=['filename', 'text'], index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 235,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-31T22:06:22.033828Z",
     "start_time": "2019-05-31T22:06:21.979731Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "text = [row.replace('—','-').replace('[','(').replace(']',')') for row in df.text.values]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 236,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-31T22:06:29.324106Z",
     "start_time": "2019-05-31T22:06:29.268496Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "df.text = text"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-28T22:41:42.088012Z",
     "start_time": "2019-05-28T22:41:41.925006Z"
    },
    "hidden": true
   },
   "source": [
    "  \n",
    "\n",
    "  \n",
    "\n",
    " "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-28T22:33:12.961993Z",
     "start_time": "2019-05-28T22:33:12.907529Z"
    },
    "heading_collapsed": true
   },
   "source": [
    "# Resize"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 145,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-31T17:58:28.579550Z",
     "start_time": "2019-05-31T17:58:28.528195Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "def resize_max(im, size=1000):\n",
    "    \"Resize an image so that the largest dimension is of specified size\"\n",
    "    r,c = im.size\n",
    "    ratio = size/max(r,c)\n",
    "    return im.resize((int(r*ratio), int(c*ratio)), Image.ANTIALIAS)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-31T17:58:28.905547Z",
     "start_time": "2019-05-31T17:58:28.846760Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "def square_max(im, top_left=False, size=None):\n",
    "    '''\n",
    "    Add whitespace to square an image by its largest dimension or specified size.\n",
    "    Args:\n",
    "        top_left: image is aligned with the top_left corner\n",
    "        size: size of final squared image.  If left blank size = largest dimension\n",
    "    '''\n",
    "    \n",
    "    r,c = im.size\n",
    "    if size is not None and size > max(r,c):\n",
    "        sz = size\n",
    "    else:\n",
    "        sz = max(r,c)\n",
    "        \n",
    "    new_im = Image.new('RGB', (sz, sz), color=(255,255,255))  # new white image\n",
    "    \n",
    "    # box logic\n",
    "    if top_left:\n",
    "        box = (0,0)\n",
    "    else:\n",
    "        if sz == r:\n",
    "            box = (0,random.randint(0,sz-c)) \n",
    "        elif sz == c:\n",
    "            box = (random.randint(0,sz-r),0)\n",
    "        else:\n",
    "            box = (random.randint(0,sz-r),random.randint(0,sz-c))\n",
    "            \n",
    "    new_im.paste(im, box=box)\n",
    "    return new_im"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 147,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-31T17:58:29.655321Z",
     "start_time": "2019-05-31T17:58:29.600063Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "def resize_dir(fn, src, targ=None):\n",
    "    if targ is None: targ = src\n",
    "    dirs = os.listdir(src)\n",
    "    for item in tqdm(dirs):\n",
    "        if os.path.isdir(src/item): continue     # skip if src dir\n",
    "#         if os.path.isfile(targ/item): continue   # skip if file exists in targ dir\n",
    "        im = Image.open(src/item)\n",
    "        rsz = fn(im)\n",
    "        rsz.save(targ/item)\n",
    "        rsz.close()\n",
    "        im.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 150,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-31T18:00:05.824969Z",
     "start_time": "2019-05-31T18:00:05.772601Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "def resize_to_square(src_dir, targ_dir, size):\n",
    "    \"Resize and square all images in src_dir and save in targ_dir\"\n",
    "    resize_dir(partial(resize_max, size=size), src_dir, targ_dir)\n",
    "    resize_dir(square_max, targ_dir)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 156,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-31T18:02:29.164119Z",
     "start_time": "2019-05-31T18:01:59.263955Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "100%|██████████| 31/31 [00:16<00:00,  2.60it/s]\n",
      "100%|██████████| 31/31 [00:13<00:00,  3.11it/s]\n"
     ]
    }
   ],
   "source": [
    "targ_path = PATH/'resized_downloads'\n",
    "os.makedirs(targ_path, exist_ok=True)\n",
    "\n",
    "resize_to_square(src_path, targ_path, 1000)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "heading_collapsed": true
   },
   "source": [
    "# SymSpell"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 192,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-30T18:46:35.918302Z",
     "start_time": "2019-05-30T18:46:35.860976Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "from symspellpy.symspellpy import SymSpell, Verbosity"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 193,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-30T18:46:36.352148Z",
     "start_time": "2019-05-30T18:46:36.299679Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "sym_spell = SymSpell()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 194,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-30T18:46:41.511038Z",
     "start_time": "2019-05-30T18:46:37.478070Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 194,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sym_spell.load_dictionary(Path('data/IAM_handwriting/frequency_dictionary_en_82_765.txt'), 0, 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 195,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-30T18:46:43.318334Z",
     "start_time": "2019-05-30T18:46:43.240469Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "input_term = \"Shus is a test letter. F hope this\\nworks but I not gre it will\\nmy hundusiting is not very god\"\n",
    "res = sym_spell.lookup_compound(input_term, 2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 196,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-30T18:46:45.084999Z",
     "start_time": "2019-05-30T18:46:45.030845Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'thus is a test letter of hope this works but i not gre it will my hindu siting is not very god'"
      ]
     },
     "execution_count": 196,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "res[0].term"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "heading_collapsed": true
   },
   "source": [
    "# Downloaded Images"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 153,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-31T18:00:36.247181Z",
     "start_time": "2019-05-31T18:00:36.196556Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "src_path = PATH/'downloads'\n",
    "text = []"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-31T17:32:35.692707Z",
     "start_time": "2019-05-31T17:32:35.627887Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "fname = '31.png'\n",
    "im = Image.open(src_path/fname)\n",
    "fig,ax = plt.subplots(figsize=(18,8))\n",
    "ax.imshow(im)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-31T17:32:31.734317Z",
     "start_time": "2019-05-31T17:32:31.628168Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "text = '''cate- we can not consecrate- we can not\n",
    "hallow- this ground. The brave men, live-\n",
    "ing and dead, who struggled here, have con-\n",
    "secrated it, for above our poor power to add\n",
    "or detract. The world will little note, nor\n",
    "long remember what we say here, but it can\n",
    "never forget what they did here. It is for us\n",
    "the living, rather, to be dedicated here to\n",
    "the unfinished work which they who fou-\n",
    "ght here have thus far so nobly advanced.\n",
    "It is rather for us to be here dedicated to\n",
    "the great task remaining before us -- that\n",
    "from these honored dead we take increased\n",
    "devotion to that cause for which they gave\n",
    "the last full measure of devotion -- that\n",
    "we here highly resolve that these dead shall\n",
    "not have died in vain -- that this nation,'''\n",
    "\n",
    "df = df.append({'filename': fname, 'text': text}, ignore_index=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-31T16:07:43.895076Z",
     "start_time": "2019-05-31T16:07:43.839647Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "df = pd.DataFrame(list(zip(names, text)),  columns =['filename', 'text'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-31T17:32:40.732132Z",
     "start_time": "2019-05-31T17:32:40.674821Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "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>filename</th>\n",
       "      <th>text</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>27.png</td>\n",
       "      <td>This is because I can spell confusion with\\na ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>28.png</td>\n",
       "      <td>Hello reddit, I am having a very tough time de...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>29.png</td>\n",
       "      <td>Hi Reddit,\\nI came across this subreddit\\ntoda...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>30.png</td>\n",
       "      <td>My Weekend\\nThis weekend was awesome!\\nI went ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>31.png</td>\n",
       "      <td>cate- we can not consecrate- we can not\\nhallo...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   filename                                               text\n",
       "26   27.png  This is because I can spell confusion with\\na ...\n",
       "27   28.png  Hello reddit, I am having a very tough time de...\n",
       "28   29.png  Hi Reddit,\\nI came across this subreddit\\ntoda...\n",
       "29   30.png  My Weekend\\nThis weekend was awesome!\\nI went ...\n",
       "30   31.png  cate- we can not consecrate- we can not\\nhallo..."
      ]
     },
     "execution_count": 103,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.tail()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 256,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-31T22:12:15.395104Z",
     "start_time": "2019-05-31T22:12:15.339497Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "CSV = PATH/'downloads.csv'\n",
    "df = pd.read_csv(CSV)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 259,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-31T22:12:39.298360Z",
     "start_time": "2019-05-31T22:12:39.242266Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "df.to_csv(CSV, columns=['filename', 'text'], index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 257,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-31T22:12:30.535886Z",
     "start_time": "2019-05-31T22:12:30.482834Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "text = [row.replace('—','-').replace('[','(').replace(']',')') for row in df.text.values]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 258,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-31T22:12:35.728658Z",
     "start_time": "2019-05-31T22:12:35.676377Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "df.text = text"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Numericalize"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-03T14:16:32.060114Z",
     "start_time": "2019-06-03T14:16:32.003093Z"
    }
   },
   "outputs": [],
   "source": [
    "df = pd.read_csv(PATH/'downloads.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-03T14:07:32.766645Z",
     "start_time": "2019-06-03T14:07:32.712146Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "82"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "itos = pickle.load(open('data/IAM_handwriting/tmp/char_itos.pkl', 'rb'))\n",
    "len(itos)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-03T14:16:35.463256Z",
     "start_time": "2019-06-03T14:16:35.394533Z"
    }
   },
   "outputs": [],
   "source": [
    "joined_labels = list(df.text)\n",
    "\n",
    "stoi = collections.defaultdict(lambda: 6, {v:k for k,v in enumerate(itos)})\n",
    "ids = np.array([np.array([stoi[letter] for letter in word] + [3]) for word in joined_labels])\n",
    "\n",
    "# convert to strings (as labels)\n",
    "str_ids = np.array([' '.join(str(l) for l in w) for w in ids]).reshape(-1,1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-03T14:16:58.700855Z",
     "start_time": "2019-06-03T14:16:58.641680Z"
    }
   },
   "outputs": [
    {
     "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>filename</th>\n",
       "      <th>text</th>\n",
       "      <th>char_ids</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1.png</td>\n",
       "      <td>My mother, Janey Schultz, always\\ntold her dau...</td>\n",
       "      <td>42 80 1 68 70 75 63 60 73 13 1 39 56 69 60 80 ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2.png</td>\n",
       "      <td>Reading a handwritten article about hand-\\nwri...</td>\n",
       "      <td>47 60 56 59 64 69 62 1 56 1 63 56 69 59 78 73 ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3.png</td>\n",
       "      <td>My Teacher told me that I really did well in t...</td>\n",
       "      <td>42 80 1 49 60 56 58 63 60 73 1 75 70 67 59 1 6...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4.png</td>\n",
       "      <td>compare it too. Investigators searched for\\ncl...</td>\n",
       "      <td>58 70 68 71 56 73 60 1 64 75 1 75 70 70 15 1 3...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5.png</td>\n",
       "      <td>Hello Everybody. I'd like to welcome you you t...</td>\n",
       "      <td>37 60 67 67 70 1 34 77 60 73 80 57 70 59 80 15...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  filename                                               text  \\\n",
       "0    1.png  My mother, Janey Schultz, always\\ntold her dau...   \n",
       "1    2.png  Reading a handwritten article about hand-\\nwri...   \n",
       "2    3.png  My Teacher told me that I really did well in t...   \n",
       "3    4.png  compare it too. Investigators searched for\\ncl...   \n",
       "4    5.png  Hello Everybody. I'd like to welcome you you t...   \n",
       "\n",
       "                                            char_ids  \n",
       "0  42 80 1 68 70 75 63 60 73 13 1 39 56 69 60 80 ...  \n",
       "1  47 60 56 59 64 69 62 1 56 1 63 56 69 59 78 73 ...  \n",
       "2  42 80 1 49 60 56 58 63 60 73 1 75 70 67 59 1 6...  \n",
       "3  58 70 68 71 56 73 60 1 64 75 1 75 70 70 15 1 3...  \n",
       "4  37 60 67 67 70 1 34 77 60 73 80 57 70 59 80 15...  "
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df['char_ids'] = str_ids\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-03T14:17:07.738255Z",
     "start_time": "2019-06-03T14:17:07.682628Z"
    }
   },
   "outputs": [],
   "source": [
    "CSV = PATH/'download_ids.csv'\n",
    "df.to_csv(CSV, columns=['filename', 'char_ids'], index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Concatenate CSVs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-03T14:17:12.219153Z",
     "start_time": "2019-06-03T14:17:12.160276Z"
    }
   },
   "outputs": [],
   "source": [
    "a = pd.read_csv(PATH/'ewd_ids.csv')\n",
    "b = pd.read_csv(PATH/'download_ids.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-03T14:17:17.326657Z",
     "start_time": "2019-06-03T14:17:17.273295Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "138"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "new = pd.concat([a,b], ignore_index=True)\n",
    "len(new)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-03T14:17:18.537957Z",
     "start_time": "2019-06-03T14:17:18.480722Z"
    }
   },
   "outputs": [],
   "source": [
    "new.to_csv(PATH/'downloaded_images.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 266,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-05-31T22:16:33.199061Z",
     "start_time": "2019-05-31T22:16:33.034571Z"
    }
   },
   "outputs": [],
   "source": [
    "df = pd.read_csv(PATH/'downloaded_images.csv')"
   ]
  },
  {
   "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": "3.6.6"
  },
  "toc": {
   "nav_menu": {},
   "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": true,
   "toc_window_display": false
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}