evaluate_iwslt-BertChineseEmbSlimCNNlstmBertLSTM.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "from glob import glob\n",
    "from transformers import BertTokenizer\n",
    "import torch\n",
    "from torch import nn\n",
    "%matplotlib inline\n",
    "import json\n",
    "from tqdm import tqdm\n",
    "from sklearn import metrics\n",
    "import os\n",
    "\n",
    "from model_1_to_1 import (\n",
    "    BertChineseLinearPunc,\n",
    "    BertChineseCNNreplaceBert,\n",
    "    BertChineseBigStrideCNN,\n",
    "    BertChineseSlimCNNBert,\n",
    "    BertChineseSlimCNNBertLSTM,\n",
    "    BertChineseEmbSlimCNNBert,\n",
    "    BertChineseEmbSlimCNNlstmBert,\n",
    "    BertChineseEmbSlimCNNlstmBertLSTM,\n",
    ")\n",
    "from data_1_to_1 import load_file, preprocess_data, create_data_loader"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Model path"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['models/bert-base-uncased.tar.gz',\n",
       " 'models/config.json',\n",
       " 'models/pytorch_model.bin',\n",
       " 'models/vocab.txt',\n",
       " 'models/1_albert_dense_noRnn',\n",
       " 'models/1_albert_dense_rnn',\n",
       " 'models/1_albert_small_dense_hidden_UniRNN',\n",
       " 'models/albert_chinese_small',\n",
       " 'models/albert_chinese_tiny',\n",
       " 'models/albert_en',\n",
       " 'models/bart_tiny',\n",
       " 'models/bert_distill_chinese',\n",
       " 'models/distillbert',\n",
       " 'models/distillroberta_tiny',\n",
       " 'models/2-BertChineseLinearPunc',\n",
       " 'models/3-BertChinese4lyrHiddenCNN',\n",
       " 'models/2-ChineseRobertaCNNlinearDivideConcatPunc_1cnn_k3_20',\n",
       " 'models/2-ChineseRobertaCNNlinearDividedPunc',\n",
       " 'models/3-BertChineseBigStrideCNN',\n",
       " 'models/chinese-roberta-wwm-ext',\n",
       " 'models/2-result-bert-chinese-linear',\n",
       " 'models/2_ChineseRobertaNmlCNNLstmPunc',\n",
       " 'models/2_ChineseRobertaCNNnoBnLstmPunc',\n",
       " 'models/bert_base_chinese',\n",
       " 'models/20200708_002614',\n",
       " 'models/2-result-ChineseRobertaRNNnewLinearPunc',\n",
       " 'models/2-ChineseRobertaNoCNNLstmPunc',\n",
       " 'models/3-SegBertChineseLinearPunc',\n",
       " 'models/2-ChineseRobertaCNNLstmPunc',\n",
       " 'models/2-ChineseRobertaCNNlinearDivideConcatPunc',\n",
       " 'models/2-ChineseRobertaCNNlinearDivideConcatPunc_multiLinear',\n",
       " 'models/2-ChineseRobertaCNNlinearDividedBertPunc',\n",
       " 'models/2-BertChineseLSTMLinearPunc',\n",
       " 'models/2-BertChineseLinearPunc_len200_bs_10',\n",
       " 'models/2-BertChineseLinearPunc_len200_bs20',\n",
       " 'models/2-BertChineseLinearPunc_len300_bs10',\n",
       " 'models/2-BertChineseLinearPunc_len150_bs20',\n",
       " 'models/2-BertChineseCNNlinearDivideConcatPunc',\n",
       " 'models/2-BertChineseCNNlinearDividedBertPunc',\n",
       " 'models/2_2_BertChineseLinearPunc_200_10',\n",
       " 'models/2-BertChineseCNNreplaceBert_len200_bs10',\n",
       " 'models/2-BertChineseSlimCNNBert_len200_bs_10',\n",
       " 'models/3_BertChineseSlimCNNBert_len200_bs10',\n",
       " 'models/3-BertChinese4lyrHiddenLinear',\n",
       " 'models/3-BertChineseSegHiddenLinearPunc',\n",
       " 'models/iwslt-BertChineseLinearPunc',\n",
       " 'models/iwslt-BertChineseSlimCNNBert',\n",
       " 'models/iwslt-BertChineseLSTMLinearPunc',\n",
       " 'models/iwslt-BertChineseSlimCNNBertLSTM',\n",
       " 'models/iwslt-BertChineseEmbSlimCNNBert',\n",
       " 'models/iwslt-BertChineseEmbSlimCNNlstmBert',\n",
       " 'models/iwslt-BertChineseEmbSlimCNNlstmBertLSTM',\n",
       " 'models/iwslt-BertChineseEmbSlimCNNlstmBertLSTM2']"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "glob('models/*')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "path = 'models/iwslt-BertChineseEmbSlimCNNlstmBertLSTM2'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Prepare"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_test = load_file('data/zh_iwslt/test_valid')\n",
    "data_test_asr = load_file('data/zh_iwslt/test_valid')\n",
    "# data_test = load_file('data/zh_pfdsj_3punc/test')\n",
    "# data_test_asr = load_file('data/zh_pfdsj_3punc//test')\n",
    "# data_test = load_file('data/NPR-podcasts/test')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'segment_size': 100,\n",
       " 'dropout': 0.3,\n",
       " 'epochs_top': 10,\n",
       " 'iterations_top': 3,\n",
       " 'batch_size_top': 40,\n",
       " 'learning_rate_top': 1e-05,\n",
       " 'epochs_all': 15,\n",
       " 'iterations_all': 3,\n",
       " 'batch_size_all': 10,\n",
       " 'learning_rate_all': 1e-05,\n",
       " 'seq_len': 200}"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "with open(os.path.join(path, 'hyperparameters.json'), 'r') as f:\n",
    "    hyperparameters = json.load(f)\n",
    "hyperparameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "tokenizer = BertTokenizer.from_pretrained(os.path.join(\"models\", 'bert_base_chinese'), do_lower_case=True)\n",
    "\n",
    "punctuation_enc = {\n",
    "    'O': 0,\n",
    "    '，': 1,\n",
    "    '。': 2,\n",
    "    '？': 3\n",
    "}\n",
    "\n",
    "# punctuation_enc = {\n",
    "#     'O': 0,\n",
    "#     'PERIOD': 1\n",
    "# }\n",
    "\n",
    "# segment_size = hyperparameters['segment_size']\n",
    "segment_size = 200"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_test, y_test = preprocess_data(data_test, tokenizer, punctuation_enc, segment_size)\n",
    "X_test_asr, y_test_asr = preprocess_data(data_test_asr, tokenizer, punctuation_enc, segment_size)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fucking code**************\n"
     ]
    }
   ],
   "source": [
    "output_size = len(punctuation_enc)\n",
    "dropout = hyperparameters['dropout']\n",
    "bert_punc = BertChineseEmbSlimCNNlstmBertLSTM(segment_size, output_size, dropout, None).to(device)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Model progress"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "scrolled": false
   },
   "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>time</th>\n",
       "      <th>epoch</th>\n",
       "      <th>iteration</th>\n",
       "      <th>training loss</th>\n",
       "      <th>loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>f1_O</th>\n",
       "      <th>f1_，</th>\n",
       "      <th>f1_。</th>\n",
       "      <th>f1_？</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2020-07-08 00:44:43</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0.2112</td>\n",
       "      <td>0.2015</td>\n",
       "      <td>0.9337</td>\n",
       "      <td>0.9694</td>\n",
       "      <td>0.0226</td>\n",
       "      <td>0.4289</td>\n",
       "      <td>0.0000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2020-07-08 00:54:59</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>0.1448</td>\n",
       "      <td>0.1333</td>\n",
       "      <td>0.9495</td>\n",
       "      <td>0.9806</td>\n",
       "      <td>0.3479</td>\n",
       "      <td>0.6815</td>\n",
       "      <td>0.1283</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2020-07-08 01:05:14</td>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>0.0999</td>\n",
       "      <td>0.1159</td>\n",
       "      <td>0.9541</td>\n",
       "      <td>0.9827</td>\n",
       "      <td>0.4692</td>\n",
       "      <td>0.7050</td>\n",
       "      <td>0.4678</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2020-07-08 01:15:30</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>0.1238</td>\n",
       "      <td>0.1115</td>\n",
       "      <td>0.9557</td>\n",
       "      <td>0.9832</td>\n",
       "      <td>0.5162</td>\n",
       "      <td>0.7234</td>\n",
       "      <td>0.4970</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2020-07-08 01:25:47</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>0.1087</td>\n",
       "      <td>0.1091</td>\n",
       "      <td>0.9573</td>\n",
       "      <td>0.9833</td>\n",
       "      <td>0.4628</td>\n",
       "      <td>0.7312</td>\n",
       "      <td>0.4928</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>2020-07-08 01:36:02</td>\n",
       "      <td>2</td>\n",
       "      <td>3</td>\n",
       "      <td>0.1212</td>\n",
       "      <td>0.1073</td>\n",
       "      <td>0.9572</td>\n",
       "      <td>0.9837</td>\n",
       "      <td>0.5239</td>\n",
       "      <td>0.7358</td>\n",
       "      <td>0.4963</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>2020-07-08 01:46:19</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0958</td>\n",
       "      <td>0.1049</td>\n",
       "      <td>0.9582</td>\n",
       "      <td>0.9835</td>\n",
       "      <td>0.4736</td>\n",
       "      <td>0.7400</td>\n",
       "      <td>0.5728</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>2020-07-08 01:56:36</td>\n",
       "      <td>3</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0931</td>\n",
       "      <td>0.1044</td>\n",
       "      <td>0.9582</td>\n",
       "      <td>0.9839</td>\n",
       "      <td>0.5193</td>\n",
       "      <td>0.7397</td>\n",
       "      <td>0.5964</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>2020-07-08 02:06:36</td>\n",
       "      <td>3</td>\n",
       "      <td>3</td>\n",
       "      <td>0.1031</td>\n",
       "      <td>0.1057</td>\n",
       "      <td>0.9582</td>\n",
       "      <td>0.9839</td>\n",
       "      <td>0.5317</td>\n",
       "      <td>0.7432</td>\n",
       "      <td>0.6276</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>2020-07-08 02:16:36</td>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0945</td>\n",
       "      <td>0.1049</td>\n",
       "      <td>0.9581</td>\n",
       "      <td>0.9837</td>\n",
       "      <td>0.5321</td>\n",
       "      <td>0.7424</td>\n",
       "      <td>0.6503</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>2020-07-08 02:26:37</td>\n",
       "      <td>4</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0943</td>\n",
       "      <td>0.1062</td>\n",
       "      <td>0.9590</td>\n",
       "      <td>0.9838</td>\n",
       "      <td>0.4998</td>\n",
       "      <td>0.7443</td>\n",
       "      <td>0.6426</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>2020-07-08 02:36:35</td>\n",
       "      <td>4</td>\n",
       "      <td>3</td>\n",
       "      <td>0.0936</td>\n",
       "      <td>0.1061</td>\n",
       "      <td>0.9588</td>\n",
       "      <td>0.9836</td>\n",
       "      <td>0.4814</td>\n",
       "      <td>0.7422</td>\n",
       "      <td>0.6387</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>2020-07-08 02:46:36</td>\n",
       "      <td>5</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0945</td>\n",
       "      <td>0.1077</td>\n",
       "      <td>0.9584</td>\n",
       "      <td>0.9838</td>\n",
       "      <td>0.5221</td>\n",
       "      <td>0.7402</td>\n",
       "      <td>0.6496</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>2020-07-08 02:56:37</td>\n",
       "      <td>5</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0943</td>\n",
       "      <td>0.1119</td>\n",
       "      <td>0.9562</td>\n",
       "      <td>0.9829</td>\n",
       "      <td>0.5414</td>\n",
       "      <td>0.7451</td>\n",
       "      <td>0.6618</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>2020-07-08 03:06:36</td>\n",
       "      <td>5</td>\n",
       "      <td>3</td>\n",
       "      <td>0.0866</td>\n",
       "      <td>0.1051</td>\n",
       "      <td>0.9588</td>\n",
       "      <td>0.9838</td>\n",
       "      <td>0.5150</td>\n",
       "      <td>0.7457</td>\n",
       "      <td>0.6622</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>2020-07-08 03:16:36</td>\n",
       "      <td>6</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0777</td>\n",
       "      <td>0.1091</td>\n",
       "      <td>0.9587</td>\n",
       "      <td>0.9837</td>\n",
       "      <td>0.5133</td>\n",
       "      <td>0.7424</td>\n",
       "      <td>0.6633</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>2020-07-08 03:26:37</td>\n",
       "      <td>6</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0762</td>\n",
       "      <td>0.1098</td>\n",
       "      <td>0.9581</td>\n",
       "      <td>0.9833</td>\n",
       "      <td>0.5141</td>\n",
       "      <td>0.7428</td>\n",
       "      <td>0.6610</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>2020-07-08 03:36:37</td>\n",
       "      <td>6</td>\n",
       "      <td>3</td>\n",
       "      <td>0.0947</td>\n",
       "      <td>0.1081</td>\n",
       "      <td>0.9581</td>\n",
       "      <td>0.9825</td>\n",
       "      <td>0.4710</td>\n",
       "      <td>0.7377</td>\n",
       "      <td>0.6547</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>2020-07-08 03:46:38</td>\n",
       "      <td>7</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0669</td>\n",
       "      <td>0.1119</td>\n",
       "      <td>0.9582</td>\n",
       "      <td>0.9835</td>\n",
       "      <td>0.5151</td>\n",
       "      <td>0.7416</td>\n",
       "      <td>0.6640</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>2020-07-08 03:56:38</td>\n",
       "      <td>7</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0769</td>\n",
       "      <td>0.1122</td>\n",
       "      <td>0.9585</td>\n",
       "      <td>0.9833</td>\n",
       "      <td>0.4825</td>\n",
       "      <td>0.7438</td>\n",
       "      <td>0.6666</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>2020-07-08 04:06:37</td>\n",
       "      <td>7</td>\n",
       "      <td>3</td>\n",
       "      <td>0.0872</td>\n",
       "      <td>0.1127</td>\n",
       "      <td>0.9574</td>\n",
       "      <td>0.9835</td>\n",
       "      <td>0.5180</td>\n",
       "      <td>0.7337</td>\n",
       "      <td>0.6473</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>2020-07-08 04:16:36</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0658</td>\n",
       "      <td>0.1181</td>\n",
       "      <td>0.9565</td>\n",
       "      <td>0.9831</td>\n",
       "      <td>0.5284</td>\n",
       "      <td>0.7300</td>\n",
       "      <td>0.6401</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>2020-07-08 04:26:35</td>\n",
       "      <td>8</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0589</td>\n",
       "      <td>0.1176</td>\n",
       "      <td>0.9577</td>\n",
       "      <td>0.9831</td>\n",
       "      <td>0.5060</td>\n",
       "      <td>0.7377</td>\n",
       "      <td>0.6620</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>2020-07-08 04:36:34</td>\n",
       "      <td>8</td>\n",
       "      <td>3</td>\n",
       "      <td>0.0734</td>\n",
       "      <td>0.1164</td>\n",
       "      <td>0.9573</td>\n",
       "      <td>0.9826</td>\n",
       "      <td>0.4942</td>\n",
       "      <td>0.7350</td>\n",
       "      <td>0.6651</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>2020-07-08 04:46:34</td>\n",
       "      <td>9</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0663</td>\n",
       "      <td>0.1264</td>\n",
       "      <td>0.9555</td>\n",
       "      <td>0.9828</td>\n",
       "      <td>0.5319</td>\n",
       "      <td>0.7279</td>\n",
       "      <td>0.6572</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>2020-07-08 04:56:34</td>\n",
       "      <td>9</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0543</td>\n",
       "      <td>0.1242</td>\n",
       "      <td>0.9577</td>\n",
       "      <td>0.9829</td>\n",
       "      <td>0.4841</td>\n",
       "      <td>0.7367</td>\n",
       "      <td>0.6637</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>2020-07-08 05:06:34</td>\n",
       "      <td>9</td>\n",
       "      <td>3</td>\n",
       "      <td>0.0639</td>\n",
       "      <td>0.1240</td>\n",
       "      <td>0.9567</td>\n",
       "      <td>0.9830</td>\n",
       "      <td>0.5183</td>\n",
       "      <td>0.7304</td>\n",
       "      <td>0.6575</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>2020-07-08 05:16:35</td>\n",
       "      <td>10</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0668</td>\n",
       "      <td>0.1280</td>\n",
       "      <td>0.9567</td>\n",
       "      <td>0.9824</td>\n",
       "      <td>0.4868</td>\n",
       "      <td>0.7342</td>\n",
       "      <td>0.6632</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>2020-07-08 05:26:36</td>\n",
       "      <td>10</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0579</td>\n",
       "      <td>0.1312</td>\n",
       "      <td>0.9565</td>\n",
       "      <td>0.9831</td>\n",
       "      <td>0.5222</td>\n",
       "      <td>0.7380</td>\n",
       "      <td>0.6570</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>2020-07-08 05:36:35</td>\n",
       "      <td>10</td>\n",
       "      <td>3</td>\n",
       "      <td>0.0499</td>\n",
       "      <td>0.1296</td>\n",
       "      <td>0.9562</td>\n",
       "      <td>0.9828</td>\n",
       "      <td>0.5179</td>\n",
       "      <td>0.7294</td>\n",
       "      <td>0.6619</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>2020-07-08 05:46:35</td>\n",
       "      <td>11</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0649</td>\n",
       "      <td>0.1380</td>\n",
       "      <td>0.9559</td>\n",
       "      <td>0.9828</td>\n",
       "      <td>0.5191</td>\n",
       "      <td>0.7309</td>\n",
       "      <td>0.6645</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>2020-07-08 05:56:36</td>\n",
       "      <td>11</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0498</td>\n",
       "      <td>0.1391</td>\n",
       "      <td>0.9563</td>\n",
       "      <td>0.9827</td>\n",
       "      <td>0.5126</td>\n",
       "      <td>0.7341</td>\n",
       "      <td>0.6417</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32</th>\n",
       "      <td>2020-07-08 06:06:36</td>\n",
       "      <td>11</td>\n",
       "      <td>3</td>\n",
       "      <td>0.0402</td>\n",
       "      <td>0.1353</td>\n",
       "      <td>0.9560</td>\n",
       "      <td>0.9822</td>\n",
       "      <td>0.4995</td>\n",
       "      <td>0.7188</td>\n",
       "      <td>0.6643</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33</th>\n",
       "      <td>2020-07-08 06:16:35</td>\n",
       "      <td>12</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0435</td>\n",
       "      <td>0.1437</td>\n",
       "      <td>0.9559</td>\n",
       "      <td>0.9826</td>\n",
       "      <td>0.5208</td>\n",
       "      <td>0.7319</td>\n",
       "      <td>0.6617</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>2020-07-08 06:26:36</td>\n",
       "      <td>12</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0503</td>\n",
       "      <td>0.1507</td>\n",
       "      <td>0.9558</td>\n",
       "      <td>0.9831</td>\n",
       "      <td>0.5186</td>\n",
       "      <td>0.7206</td>\n",
       "      <td>0.6578</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>2020-07-08 06:36:36</td>\n",
       "      <td>12</td>\n",
       "      <td>3</td>\n",
       "      <td>0.0362</td>\n",
       "      <td>0.1464</td>\n",
       "      <td>0.9564</td>\n",
       "      <td>0.9830</td>\n",
       "      <td>0.5101</td>\n",
       "      <td>0.7229</td>\n",
       "      <td>0.6562</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>36</th>\n",
       "      <td>2020-07-08 06:46:36</td>\n",
       "      <td>13</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0475</td>\n",
       "      <td>0.1516</td>\n",
       "      <td>0.9564</td>\n",
       "      <td>0.9824</td>\n",
       "      <td>0.4928</td>\n",
       "      <td>0.7249</td>\n",
       "      <td>0.6568</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37</th>\n",
       "      <td>2020-07-08 06:56:37</td>\n",
       "      <td>13</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0403</td>\n",
       "      <td>0.1496</td>\n",
       "      <td>0.9561</td>\n",
       "      <td>0.9823</td>\n",
       "      <td>0.5025</td>\n",
       "      <td>0.7253</td>\n",
       "      <td>0.6549</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38</th>\n",
       "      <td>2020-07-08 07:06:37</td>\n",
       "      <td>13</td>\n",
       "      <td>3</td>\n",
       "      <td>0.0440</td>\n",
       "      <td>0.1469</td>\n",
       "      <td>0.9564</td>\n",
       "      <td>0.9826</td>\n",
       "      <td>0.4861</td>\n",
       "      <td>0.7269</td>\n",
       "      <td>0.6581</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39</th>\n",
       "      <td>2020-07-08 07:16:38</td>\n",
       "      <td>14</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0360</td>\n",
       "      <td>0.1575</td>\n",
       "      <td>0.9560</td>\n",
       "      <td>0.9822</td>\n",
       "      <td>0.4830</td>\n",
       "      <td>0.7222</td>\n",
       "      <td>0.6540</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>40</th>\n",
       "      <td>2020-07-08 07:26:39</td>\n",
       "      <td>14</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0374</td>\n",
       "      <td>0.1551</td>\n",
       "      <td>0.9556</td>\n",
       "      <td>0.9819</td>\n",
       "      <td>0.4801</td>\n",
       "      <td>0.7241</td>\n",
       "      <td>0.6250</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>41</th>\n",
       "      <td>2020-07-08 07:36:39</td>\n",
       "      <td>14</td>\n",
       "      <td>3</td>\n",
       "      <td>0.0327</td>\n",
       "      <td>0.1572</td>\n",
       "      <td>0.9555</td>\n",
       "      <td>0.9827</td>\n",
       "      <td>0.5040</td>\n",
       "      <td>0.7172</td>\n",
       "      <td>0.6429</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>42</th>\n",
       "      <td>2020-07-08 07:46:40</td>\n",
       "      <td>15</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0380</td>\n",
       "      <td>0.1645</td>\n",
       "      <td>0.9551</td>\n",
       "      <td>0.9824</td>\n",
       "      <td>0.5024</td>\n",
       "      <td>0.7128</td>\n",
       "      <td>0.6597</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>43</th>\n",
       "      <td>2020-07-08 07:56:40</td>\n",
       "      <td>15</td>\n",
       "      <td>2</td>\n",
       "      <td>0.0353</td>\n",
       "      <td>0.1662</td>\n",
       "      <td>0.9556</td>\n",
       "      <td>0.9823</td>\n",
       "      <td>0.4983</td>\n",
       "      <td>0.7270</td>\n",
       "      <td>0.6598</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>44</th>\n",
       "      <td>2020-07-08 08:06:39</td>\n",
       "      <td>15</td>\n",
       "      <td>3</td>\n",
       "      <td>0.0414</td>\n",
       "      <td>0.1679</td>\n",
       "      <td>0.9553</td>\n",
       "      <td>0.9825</td>\n",
       "      <td>0.5091</td>\n",
       "      <td>0.7234</td>\n",
       "      <td>0.6542</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                   time  epoch  iteration  training loss    loss  accuracy  \\\n",
       "0   2020-07-08 00:44:43      1          1         0.2112  0.2015    0.9337   \n",
       "1   2020-07-08 00:54:59      1          2         0.1448  0.1333    0.9495   \n",
       "2   2020-07-08 01:05:14      1          3         0.0999  0.1159    0.9541   \n",
       "3   2020-07-08 01:15:30      2          1         0.1238  0.1115    0.9557   \n",
       "4   2020-07-08 01:25:47      2          2         0.1087  0.1091    0.9573   \n",
       "5   2020-07-08 01:36:02      2          3         0.1212  0.1073    0.9572   \n",
       "6   2020-07-08 01:46:19      3          1         0.0958  0.1049    0.9582   \n",
       "7   2020-07-08 01:56:36      3          2         0.0931  0.1044    0.9582   \n",
       "8   2020-07-08 02:06:36      3          3         0.1031  0.1057    0.9582   \n",
       "9   2020-07-08 02:16:36      4          1         0.0945  0.1049    0.9581   \n",
       "10  2020-07-08 02:26:37      4          2         0.0943  0.1062    0.9590   \n",
       "11  2020-07-08 02:36:35      4          3         0.0936  0.1061    0.9588   \n",
       "12  2020-07-08 02:46:36      5          1         0.0945  0.1077    0.9584   \n",
       "13  2020-07-08 02:56:37      5          2         0.0943  0.1119    0.9562   \n",
       "14  2020-07-08 03:06:36      5          3         0.0866  0.1051    0.9588   \n",
       "15  2020-07-08 03:16:36      6          1         0.0777  0.1091    0.9587   \n",
       "16  2020-07-08 03:26:37      6          2         0.0762  0.1098    0.9581   \n",
       "17  2020-07-08 03:36:37      6          3         0.0947  0.1081    0.9581   \n",
       "18  2020-07-08 03:46:38      7          1         0.0669  0.1119    0.9582   \n",
       "19  2020-07-08 03:56:38      7          2         0.0769  0.1122    0.9585   \n",
       "20  2020-07-08 04:06:37      7          3         0.0872  0.1127    0.9574   \n",
       "21  2020-07-08 04:16:36      8          1         0.0658  0.1181    0.9565   \n",
       "22  2020-07-08 04:26:35      8          2         0.0589  0.1176    0.9577   \n",
       "23  2020-07-08 04:36:34      8          3         0.0734  0.1164    0.9573   \n",
       "24  2020-07-08 04:46:34      9          1         0.0663  0.1264    0.9555   \n",
       "25  2020-07-08 04:56:34      9          2         0.0543  0.1242    0.9577   \n",
       "26  2020-07-08 05:06:34      9          3         0.0639  0.1240    0.9567   \n",
       "27  2020-07-08 05:16:35     10          1         0.0668  0.1280    0.9567   \n",
       "28  2020-07-08 05:26:36     10          2         0.0579  0.1312    0.9565   \n",
       "29  2020-07-08 05:36:35     10          3         0.0499  0.1296    0.9562   \n",
       "30  2020-07-08 05:46:35     11          1         0.0649  0.1380    0.9559   \n",
       "31  2020-07-08 05:56:36     11          2         0.0498  0.1391    0.9563   \n",
       "32  2020-07-08 06:06:36     11          3         0.0402  0.1353    0.9560   \n",
       "33  2020-07-08 06:16:35     12          1         0.0435  0.1437    0.9559   \n",
       "34  2020-07-08 06:26:36     12          2         0.0503  0.1507    0.9558   \n",
       "35  2020-07-08 06:36:36     12          3         0.0362  0.1464    0.9564   \n",
       "36  2020-07-08 06:46:36     13          1         0.0475  0.1516    0.9564   \n",
       "37  2020-07-08 06:56:37     13          2         0.0403  0.1496    0.9561   \n",
       "38  2020-07-08 07:06:37     13          3         0.0440  0.1469    0.9564   \n",
       "39  2020-07-08 07:16:38     14          1         0.0360  0.1575    0.9560   \n",
       "40  2020-07-08 07:26:39     14          2         0.0374  0.1551    0.9556   \n",
       "41  2020-07-08 07:36:39     14          3         0.0327  0.1572    0.9555   \n",
       "42  2020-07-08 07:46:40     15          1         0.0380  0.1645    0.9551   \n",
       "43  2020-07-08 07:56:40     15          2         0.0353  0.1662    0.9556   \n",
       "44  2020-07-08 08:06:39     15          3         0.0414  0.1679    0.9553   \n",
       "\n",
       "      f1_O    f1_，    f1_。    f1_？  \n",
       "0   0.9694  0.0226  0.4289  0.0000  \n",
       "1   0.9806  0.3479  0.6815  0.1283  \n",
       "2   0.9827  0.4692  0.7050  0.4678  \n",
       "3   0.9832  0.5162  0.7234  0.4970  \n",
       "4   0.9833  0.4628  0.7312  0.4928  \n",
       "5   0.9837  0.5239  0.7358  0.4963  \n",
       "6   0.9835  0.4736  0.7400  0.5728  \n",
       "7   0.9839  0.5193  0.7397  0.5964  \n",
       "8   0.9839  0.5317  0.7432  0.6276  \n",
       "9   0.9837  0.5321  0.7424  0.6503  \n",
       "10  0.9838  0.4998  0.7443  0.6426  \n",
       "11  0.9836  0.4814  0.7422  0.6387  \n",
       "12  0.9838  0.5221  0.7402  0.6496  \n",
       "13  0.9829  0.5414  0.7451  0.6618  \n",
       "14  0.9838  0.5150  0.7457  0.6622  \n",
       "15  0.9837  0.5133  0.7424  0.6633  \n",
       "16  0.9833  0.5141  0.7428  0.6610  \n",
       "17  0.9825  0.4710  0.7377  0.6547  \n",
       "18  0.9835  0.5151  0.7416  0.6640  \n",
       "19  0.9833  0.4825  0.7438  0.6666  \n",
       "20  0.9835  0.5180  0.7337  0.6473  \n",
       "21  0.9831  0.5284  0.7300  0.6401  \n",
       "22  0.9831  0.5060  0.7377  0.6620  \n",
       "23  0.9826  0.4942  0.7350  0.6651  \n",
       "24  0.9828  0.5319  0.7279  0.6572  \n",
       "25  0.9829  0.4841  0.7367  0.6637  \n",
       "26  0.9830  0.5183  0.7304  0.6575  \n",
       "27  0.9824  0.4868  0.7342  0.6632  \n",
       "28  0.9831  0.5222  0.7380  0.6570  \n",
       "29  0.9828  0.5179  0.7294  0.6619  \n",
       "30  0.9828  0.5191  0.7309  0.6645  \n",
       "31  0.9827  0.5126  0.7341  0.6417  \n",
       "32  0.9822  0.4995  0.7188  0.6643  \n",
       "33  0.9826  0.5208  0.7319  0.6617  \n",
       "34  0.9831  0.5186  0.7206  0.6578  \n",
       "35  0.9830  0.5101  0.7229  0.6562  \n",
       "36  0.9824  0.4928  0.7249  0.6568  \n",
       "37  0.9823  0.5025  0.7253  0.6549  \n",
       "38  0.9826  0.4861  0.7269  0.6581  \n",
       "39  0.9822  0.4830  0.7222  0.6540  \n",
       "40  0.9819  0.4801  0.7241  0.6250  \n",
       "41  0.9827  0.5040  0.7172  0.6429  \n",
       "42  0.9824  0.5024  0.7128  0.6597  \n",
       "43  0.9823  0.4983  0.7270  0.6598  \n",
       "44  0.9825  0.5091  0.7234  0.6542  "
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "progress = pd.read_csv(os.path.join(path, 'progress.csv'), delimiter=';')\n",
    "progress"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "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": [
    "progress[['training loss', 'loss']].plot();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/zz/anaconda3/envs/torch_gpu_1.5/lib/python3.7/site-packages/matplotlib/backends/backend_agg.py:214: RuntimeWarning: Glyph 65292 missing from current font.\n",
      "  font.set_text(s, 0.0, flags=flags)\n",
      "/home/zz/anaconda3/envs/torch_gpu_1.5/lib/python3.7/site-packages/matplotlib/backends/backend_agg.py:214: RuntimeWarning: Glyph 12290 missing from current font.\n",
      "  font.set_text(s, 0.0, flags=flags)\n",
      "/home/zz/anaconda3/envs/torch_gpu_1.5/lib/python3.7/site-packages/matplotlib/backends/backend_agg.py:214: RuntimeWarning: Glyph 65311 missing from current font.\n",
      "  font.set_text(s, 0.0, flags=flags)\n",
      "/home/zz/anaconda3/envs/torch_gpu_1.5/lib/python3.7/site-packages/matplotlib/backends/backend_agg.py:183: RuntimeWarning: Glyph 65292 missing from current font.\n",
      "  font.set_text(s, 0, flags=flags)\n",
      "/home/zz/anaconda3/envs/torch_gpu_1.5/lib/python3.7/site-packages/matplotlib/backends/backend_agg.py:183: RuntimeWarning: Glyph 12290 missing from current font.\n",
      "  font.set_text(s, 0, flags=flags)\n",
      "/home/zz/anaconda3/envs/torch_gpu_1.5/lib/python3.7/site-packages/matplotlib/backends/backend_agg.py:183: RuntimeWarning: Glyph 65311 missing from current font.\n",
      "  font.set_text(s, 0, flags=flags)\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": [
    "progress[['accuracy', 'f1_O', 'f1_，', 'f1_。', 'f1_？']].plot();\n",
    "# progress[['accuracy', 'f1_O', 'f1_PERIOD']].plot();"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Model evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "bert_punc.load_state_dict(torch.load(os.path.join(path, 'model')))\n",
    "bert_punc.eval();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "batch_size = 200\n",
    "data_loader_test = create_data_loader(X_test, y_test, False, batch_size)\n",
    "data_loader_test_asr = create_data_loader(X_test_asr, y_test_asr, False, batch_size)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "def predictions(data_loader):\n",
    "    y_pred = []\n",
    "    y_true = []\n",
    "    for inputs, labels in tqdm(data_loader, total=len(data_loader)):\n",
    "        with torch.no_grad():\n",
    "            inputs, labels = inputs.cuda(), labels.cuda()\n",
    "            output = bert_punc(inputs)\n",
    "            y_pred += list(output.argmax(dim=1).cpu().data.numpy().flatten())\n",
    "            y_true += list(labels.cpu().data.numpy().flatten())\n",
    "    return y_pred, y_true"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "def evaluation(y_pred, y_test):\n",
    "    precision, recall, f1, _ = metrics.precision_recall_fscore_support(\n",
    "        y_test, y_pred, average=None, labels=[1, 2, 3])\n",
    "    overall = metrics.precision_recall_fscore_support(\n",
    "        y_test, y_pred, average='macro', labels=[1, 2, 3])\n",
    "    result = pd.DataFrame(\n",
    "        np.array([precision, recall, f1]), \n",
    "        columns=list(['O', 'COMMA', 'PERIOD', 'QUESTION'])[1:], \n",
    "        index=['Precision', 'Recall', 'F1']\n",
    "    )\n",
    "    result['OVERALL'] = overall[:3]\n",
    "    return result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "# def evaluation(y_pred, y_test):\n",
    "#     precision, recall, f1, _ = metrics.precision_recall_fscore_support(\n",
    "#         y_test, y_pred, average=None, labels=[1])\n",
    "#     overall = metrics.precision_recall_fscore_support(\n",
    "#         y_test, y_pred, average='macro', labels=[1])\n",
    "#     result = pd.DataFrame(\n",
    "#         np.array([precision, recall, f1]), \n",
    "#         columns=list(punctuation_enc.keys())[1:], \n",
    "#         index=['Precision', 'Recall', 'F1']\n",
    "#     )\n",
    "#     result['OVERALL'] = overall[:3]\n",
    "#     return result"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4/4 [00:06<00:00,  1.59s/it]\n"
     ]
    },
    {
     "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>COMMA</th>\n",
       "      <th>PERIOD</th>\n",
       "      <th>QUESTION</th>\n",
       "      <th>OVERALL</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Precision</th>\n",
       "      <td>0.528430</td>\n",
       "      <td>0.686711</td>\n",
       "      <td>0.797468</td>\n",
       "      <td>0.670870</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Recall</th>\n",
       "      <td>0.553400</td>\n",
       "      <td>0.815846</td>\n",
       "      <td>0.612840</td>\n",
       "      <td>0.660695</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>F1</th>\n",
       "      <td>0.540627</td>\n",
       "      <td>0.745730</td>\n",
       "      <td>0.693069</td>\n",
       "      <td>0.659809</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              COMMA    PERIOD  QUESTION   OVERALL\n",
       "Precision  0.528430  0.686711  0.797468  0.670870\n",
       "Recall     0.553400  0.815846  0.612840  0.660695\n",
       "F1         0.540627  0.745730  0.693069  0.659809"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_pred_test, y_true_test = predictions(data_loader_test)\n",
    "eval_test = evaluation(y_pred_test, y_true_test)\n",
    "eval_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "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>COMMA</th>\n",
       "      <th>PERIOD</th>\n",
       "      <th>QUESTION</th>\n",
       "      <th>OVERALL</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Precision</th>\n",
       "      <td>0.655</td>\n",
       "      <td>0.733</td>\n",
       "      <td>0.707</td>\n",
       "      <td>0.700</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Recall</th>\n",
       "      <td>0.471</td>\n",
       "      <td>0.725</td>\n",
       "      <td>0.630</td>\n",
       "      <td>0.597</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>F1</th>\n",
       "      <td>0.548</td>\n",
       "      <td>0.729</td>\n",
       "      <td>0.667</td>\n",
       "      <td>0.644</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           COMMA  PERIOD  QUESTION  OVERALL\n",
       "Precision  0.655   0.733     0.707    0.700\n",
       "Recall     0.471   0.725     0.630    0.597\n",
       "F1         0.548   0.729     0.667    0.644"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# https://www.isca-speech.org/archive/Interspeech_2016/pdfs/1517.PDF\n",
    "ref_test = pd.DataFrame({\n",
    "    'COMMA':       [0.655, 0.471, 0.548],\n",
    "    'PERIOD':       [0.733, 0.725, 0.729],\n",
    "    'QUESTION':       [0.707, 0.630, 0.667],\n",
    "    'OVERALL':  [0.700, 0.597, 0.644]\n",
    "}, index=['Precision', 'Recall', 'F1'])\n",
    "ref_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAsIAAAErCAYAAADUseUOAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAemElEQVR4nO3df5RdZX3v8fenw48ACiJJ7/ISQlKLCiYhhBFIqbZVosEfobVFQb0CC01rofUWVysoBQXLshXtvfbGWi5SqKJUxFtTjYbywx/XnwSMUJIAMUSZXLsMGMAAARK/9485SY/DJHMmOZOTzH6/1pqV/Tz7OXt/z5Az+fDMs/dOVSFJkiQ1za/0ugBJkiSpFwzCkiRJaiSDsCRJkhrJICxJkqRGMghLkiSpkQzCkiRJaiSDsCRJkhrJICxJYyTJm5IsTbIhyU+SfDnJb7b2HZVkUZJHkvw8ya1JfqPttVOTVJLvDznmxCRPJVnT1rem1TdxyNjvt44xdUj/+1r9x4/B25akPYZBWJLGQJLzgP8BXAb8F2AK8DHglCTPB74J3AVMA/4r8H+AG5PMGXKo/ZNMb2u/Cbh/mFPeD5zedv4ZwP7D1BXgrcDPWn9KUmPFJ8tJUnclOQhYC5xVVdcPs/+TwCFV9eoh/X8PvLiqXtaaxb0f+EvgoKr689aYpQyG5rdX1dRW3xrgSuCUqnpJq+9yYD3wAWBaVa1p9b8MWAK8Dfgo8LyqeqqLb1+S9hjOCEtS980BJjAYWIczF3hGQAY+C5yYZL+2vk8BpyXpS3IU8Czgu8O89jvAgUmOTNIHnNZ67VBnAP/aOhfA60Z6M5I0XhmEJan7DgEerKpN29g/EfjJMP0/YfDn8nPb+gaAe4CTGFzK8MntnPeTrTFzgRUMzkpvlWR/4FTg01X1NPA5XB4hqcH26nUBkjQOPQRMTLLXNsLwg8Dzhul/HvALBpc0/Gpb/z8BZwK/AbwUeME2zvtJ4OsMrjv+p2H2/x6wCVjcal8L3JRkUlWt294bkqTxyBlhSeq+bwNPAr+7jf03MTgzO9QbgG9X1eND+m8AXgOsrqofb+ukVfUjBtcVvxr4/DBDzmBwacWPk/wHg8sz9mbwAjxJahxnhCWpy6rqkSQXAQuTbAJuBJ5mcHnD7wDvB25L8lfAh1v7zmRwmcIrhzneY0lezuBM8UjOBg5uvWbrz/gkhwKvAE4G7mwb/99b5/2fo32fkrSnMwhL0hioqg+3Zl0vZHAJws+B24G/qqr7WvcT/iCwhsHfzi0FXlVV39zG8ZZ2eN4fbmPXfwOWVdWN7Z1JPgq8K8n0qvr3Ts4hSeOFt0+TJElSI7lGWJIkSY1kEJYkSVIjGYQlSZLUSAZhSZIkNVJHQTjJvCT3JFmV5Pxh9k9JcmuS7ye5M8mru1+qJEmS1D0j3jWi9cz6exl8ZOcAcBtwelUtbxtzBfD9qvr7JEcBi6tq6vaOO3HixJo6dbtDJEmSpJ12++23P1hVk4b2d3If4eOAVVW1GiDJdcApwPK2MQUc2No+CPh/Ix106tSpLF3a0W0xJUmSpB2W5EfD9XeyNOJQ4IG29kCrr937gLckGWDwGfZ/so0iFiRZmmTpunU+1l6SJEm9062L5U4Hrq6qyQw+4/6TSZ5x7Kq6oqr6q6p/0qRnzE5LkiRJu0wnQXgtcFhbe3Krr93ZwGcBqurbwARgYjcKlCRJksZCJ2uEbwOOSDKNwQB8GvCmIWN+DLwCuDrJkQwG4VGvfXj66acZGBhg48aNo32ptmHChAlMnjyZvffeu9elSJIk7VZGDMJVtSnJucASoA+4qqruTnIJsLSqFgHvAv53kj9j8MK5M2uk21EMY2BggGc/+9lMnTqVJKN9uYaoKh566CEGBgaYNm1ar8uRJEnarXQyI0xVLWbwIrj2vovatpcDJ+5sMRs3bjQEd1ESDjnkELwwUZIk6Zl2uyfLGYK7y++nJEnS8Ha7INxrfX19zJo1i+nTp/O6172Ohx9+eLvj161bx/HHH88xxxzDN77xjV1UpSRJknZWR0sjemXq+V/q6vHWfPA1I47Zb7/9WLZsGQBnnHEGCxcu5L3vfe82x998883MmDGDK6+8suM6Nm/eTF9fX8fjJWmr9x3U6wrG1vse6XUFkhrEGeHtmDNnDmvXDt4p7oc//CHz5s3j2GOP5aUvfSkrV65k2bJl/MVf/AVf+MIXmDVrFk888QQ33ngjc+bMYfbs2Zx66qls2LABGHyS3rvf/W5mz57N9ddfv91xF198MbNnz2bGjBmsXLkSgA0bNnDWWWcxY8YMZs6cyQ033ACwzeNIkiRp+wzC27B582Zuvvlm5s+fD8CCBQv4u7/7O26//XYuv/xy/viP/5hZs2ZxySWX8MY3vpFly5bx2GOP8YEPfICbbrqJO+64g/7+fj7ykY9sPeYhhxzCHXfcwUknnbTdcRMnTuSOO+7gHe94B5dffjkAl156KQcddBB33XUXd955Jy9/+ct58MEHt3scSZIkbdtuvTSiF5544glmzZrF2rVrOfLII5k7dy4bNmzgW9/6FqeeeurWcU8++eQzXvud73yH5cuXc+KJgzfQeOqpp5gzZ87W/W984xs7Gvf6178egGOPPZbPf/7zANx0001cd911W8ccfPDBfPGLX9zucSRJkrRtBuEhtqwRfvzxx3nVq17FwoULOfPMM3nOc56zde3wtlQVc+fO5TOf+cyw+w844ICOxu27777A4IV7mzZt2uHzSZKkLnF9/rjk0oht2H///fnoRz/Khz/8Yfbff3+mTZvG9ddfDwwG0B/84AfPeM0JJ5zAN7/5TVatWgXAY489xr333rvD49rNnTuXhQsXbm2vX79+h44jSZKkQQbh7TjmmGOYOXMmn/nMZ7j22mv5xCc+wdFHH82LX/xivvCFLzxj/KRJk7j66qs5/fTTmTlzJnPmzNl6sduOjGt34YUXsn79eqZPn87RRx/NrbfeukPHkSRJ0qDswJOQu6K/v7+WLl36S30rVqzgyCOP7Ek945nfV2kc8dezUm/42dujJbm9qvqH9rtGWDtmPP9AGOc/DCRJ0iCXRkiSJKmRDMKSJElqJIOwJEmSGskgLEmSpEYyCEuSJKmRDMJD9PX1MWvWLI4++mhmz57Nt771rVEf47LLLhv2mNOnT+fUU0/l8ccf71a5kiRJ2kG79+3Tun2Lrg5ui7XlEcsAS5Ys4YILLuBrX/taR4evKqqKyy67jPe85z3DHvPNb34zH//4xznvvPN24A1IkiSpWzqaEU4yL8k9SVYlOX+Y/X+bZFnr694kD3e/1F3v0Ucf5eCDD97a/tCHPsRLXvISZs6cycUXXwzAmjVreOELX8hb3/pWpk+fztlnn80TTzzBrFmzePOb3/yMY770pS9l1apVfPWrX+W1r33t1v5zzz2Xq6++GoCpU6dy8cUXM3v2bGbMmLH1aXEbNmzgrLPOYsaMGcycOZMbbrhhDN+9JEnS+DbijHCSPmAhMBcYAG5Lsqiqlm8ZU1V/1jb+T4BjxqDWXWJLiN24cSM/+clPuOWWWwC48cYbue+++/je975HVTF//ny+/vWvM2XKFO677z6uueYaTjjhBACuv/76rTPA7TZt2sSXv/xl5s2bN2IdEydO5I477uBjH/sYl19+OVdeeSWXXnopBx10EHfddRcA69ev7+I7lyRJapZOZoSPA1ZV1eqqegq4DjhlO+NPBz7TjeJ6YcsyhpUrV/KVr3yFt771rVQVN954IzfeeCPHHHMMs2fPZuXKldx3330AHH744VtD8HC2hOv+/n6mTJnC2WefPWIdr3/96wE49thjWbNmDQA33XQT55xzztYx7bPVkiRJGp1O1ggfCjzQ1h4Ajh9uYJLDgWnALdvYvwBYADBlypRRFdoLc+bM4cEHH2TdunVUFRdccAF/+Id/+Etj1qxZwwEHHLDd47SvEd5ir7324he/+MXW9saNG39p/7777gsMXmi3adOmnXkbkiRJGka37xpxGvC5qto83M6quqKq+quqf9KkSV0+dfetXLmSzZs3c8ghh/CqV72Kq666ig0bNgCwdu1afvrTnw77ur333punn356u8c+/PDDWb58OU8++SQPP/wwN99884j1zJ07l4ULF25tuzRCkiRpx3UyI7wWOKytPbnVN5zTgHO2sW+PsGUZAwzeBeKaa66hr6+PV77ylaxYsYI5c+YA8KxnPYtPfepT9PX1PeMYCxYsYObMmcyePZtrr7122PMcdthhvOENb2D69OlMmzaNY44ZeVn1hRdeyDnnnMP06dPp6+vj4osv3rqEQpIkSaOTqtr+gGQv4F7gFQwG4NuAN1XV3UPGvQj4CjCtRjoo0N/fX0uXLv2lvhUrVnDkkUeO6g1oZGPyfe32re12Jx3cZk/qmfH82QM/f9p9+dnboyW5var6h/aPuDSiqjYB5wJLgBXAZ6vq7iSXJJnfNvQ04LpOQrAkSZLUax09UKOqFgOLh/RdNKT9vu6VJUmSJI0tH7EsSZKkRtrtgrArK7rL76ckSdLwdqsgPGHCBB566CHDW5dUFQ899BATJkzodSmSJEm7nY7WCO8qkydPZmBggHXr1vW6lHFjwoQJTJ48uddlSJIk7XZ2qyC89957M23atF6XIUmSpAbYrZZGSJIkSbvKbjUjPJ5MPf9LvS5hTK1x2bEkSdrDOSMsSZKkRjIIS5IkqZEMwpIkSWokg7AkSZIaySAsSZKkRjIIS5IkqZEMwpIkSWokg7AkSZIaySAsSZKkRjIIS5IkqZE6CsJJ5iW5J8mqJOdvY8wbkixPcneST3e3TEmSJKm79hppQJI+YCEwFxgAbkuyqKqWt405ArgAOLGq1if51bEqWJIkSeqGTmaEjwNWVdXqqnoKuA44ZciYtwMLq2o9QFX9tLtlSpIkSd3VSRA+FHigrT3Q6mv3AuAFSb6Z5DtJ5nWrQEmSJGksjLg0YhTHOQL4bWAy8PUkM6rq4fZBSRYACwCmTJnSpVNLkiRJo9fJjPBa4LC29uRWX7sBYFFVPV1V9wP3MhiMf0lVXVFV/VXVP2nSpB2tWZIkSdppnQTh24AjkkxLsg9wGrBoyJh/YXA2mCQTGVwqsbqLdUqSJEldNWIQrqpNwLnAEmAF8NmqujvJJUnmt4YtAR5Kshy4FfjzqnporIqWJEmSdlZHa4SrajGweEjfRW3bBZzX+pIkSZJ2ez5ZTpIkSY1kEJYkSVIjGYQlSZLUSAZhSZIkNZJBWJIkSY1kEJYkSVIjGYQlSZLUSAZhSZIkNZJBWJIkSY3U0ZPlJEmStmfq+V/qdQljas2EXlegseCMsCRJkhrJICxJkqRGMghLkiSpkQzCkiRJaiSDsCRJkhrJICxJkqRGMghLkiSpkQzCkiRJaqSOgnCSeUnuSbIqyfnD7D8zyboky1pfb+t+qZIkSVL3jPhkuSR9wEJgLjAA3JZkUVUtHzL0n6vq3DGoUZIkSeq6TmaEjwNWVdXqqnoKuA44ZWzLkiRJksZWJ0H4UOCBtvZAq2+o309yZ5LPJTmsK9VJkiRJY6RbF8v9KzC1qmYC/wZcM9ygJAuSLE2ydN26dV06tSRJkjR6nQThtUD7DO/kVt9WVfVQVT3Zal4JHDvcgarqiqrqr6r+SZMm7Ui9kiRJUld0EoRvA45IMi3JPsBpwKL2AUme19acD6zoXomSJElS941414iq2pTkXGAJ0AdcVVV3J7kEWFpVi4A/TTIf2AT8DDhzDGuWJEmSdtqIQRigqhYDi4f0XdS2fQFwQXdLkyRJksaOT5aTJElSIxmEJUmS1EgdLY2QpD3F1PO/1OsSxtSaCb2uQJLGD2eEJUmS1EgGYUmSJDWSQViSJEmNZBCWJElSIxmEJUmS1EgGYUmSJDWSQViSJEmNZBCWJElSIxmEJUmS1EgGYUmSJDWSQViSJEmNZBCWJElSIxmEJUmS1EgGYUmSJDWSQViSJEmN1FEQTjIvyT1JViU5fzvjfj9JJenvXomSJElS940YhJP0AQuBk4GjgNOTHDXMuGcD7wS+2+0iJUmSpG7rZEb4OGBVVa2uqqeA64BThhl3KfDXwMYu1idJkiSNiU6C8KHAA23tgVbfVklmA4dV1Ze6WJskSZI0Znb6YrkkvwJ8BHhXB2MXJFmaZOm6det29tSSJEnSDuskCK8FDmtrT271bfFsYDrw1SRrgBOARcNdMFdVV1RVf1X1T5o0acerliRJknZSJ0H4NuCIJNOS7AOcBizasrOqHqmqiVU1taqmAt8B5lfV0jGpWJIkSeqCEYNwVW0CzgWWACuAz1bV3UkuSTJ/rAuUJEmSxsJenQyqqsXA4iF9F21j7G/vfFmSJEnS2PLJcpIkSWokg7AkSZIaySAsSZKkRjIIS5IkqZEMwpIkSWokg7AkSZIaySAsSZKkRjIIS5IkqZEMwpIkSWokg7AkSZIaySAsSZKkRjIIS5IkqZEMwpIkSWokg7AkSZIaySAsSZKkRjIIS5IkqZEMwpIkSWokg7AkSZIaqaMgnGReknuSrEpy/jD7/yjJXUmWJfm/SY7qfqmSJElS94wYhJP0AQuBk4GjgNOHCbqfrqoZVTUL+BvgI12vVJIkSeqiTmaEjwNWVdXqqnoKuA44pX1AVT3a1jwAqO6VKEmSJHXfXh2MORR4oK09ABw/dFCSc4DzgH2Alw93oCQLgAUAU6ZMGW2tkiRJUtd07WK5qlpYVc8H3g1cuI0xV1RVf1X1T5o0qVunliRJkkatkyC8FjisrT251bct1wG/uzNFSZIkSWOtkyB8G3BEkmlJ9gFOAxa1D0hyRFvzNcB93StRkiRJ6r4R1whX1aYk5wJLgD7gqqq6O8klwNKqWgScm+Qk4GlgPXDGWBYtSZIk7axOLpajqhYDi4f0XdS2/c4u1yVJkiSNKZ8sJ0mSpEYyCEuSJKmRDMKSJElqJIOwJEmSGskgLEmSpEYyCEuSJKmRDMKSJElqJIOwJEmSGskgLEmSpEYyCEuSJKmRDMKSJElqJIOwJEmSGskgLEmSpEYyCEuSJKmRDMKSJElqJIOwJEmSGskgLEmSpEYyCEuSJKmROgrCSeYluSfJqiTnD7P/vCTLk9yZ5OYkh3e/VEmSJKl7RgzCSfqAhcDJwFHA6UmOGjLs+0B/Vc0EPgf8TbcLlSRJkrqpkxnh44BVVbW6qp4CrgNOaR9QVbdW1eOt5neAyd0tU5IkSequToLwocADbe2BVt+2nA18ebgdSRYkWZpk6bp16zqvUpIkSeqyrl4sl+QtQD/woeH2V9UVVdVfVf2TJk3q5qklSZKkUdmrgzFrgcPa2pNbfb8kyUnAe4Hfqqonu1OeJEmSNDY6mRG+DTgiybQk+wCnAYvaByQ5BvgHYH5V/bT7ZUqSJEndNWIQrqpNwLnAEmAF8NmqujvJJUnmt4Z9CHgWcH2SZUkWbeNwkiRJ0m6hk6URVNViYPGQvovatk/qcl2SJEnSmPLJcpIkSWokg7AkSZIaySAsSZKkRjIIS5IkqZEMwpIkSWokg7AkSZIaySAsSZKkRjIIS5IkqZEMwpIkSWokg7AkSZIaySAsSZKkRjIIS5IkqZEMwpIkSWokg7AkSZIaySAsSZKkRjIIS5IkqZEMwpIkSWokg7AkSZIaqaMgnGReknuSrEpy/jD7X5bkjiSbkvxB98uUJEmSumvEIJykD1gInAwcBZye5Kghw34MnAl8utsFSpIkSWNhrw7GHAesqqrVAEmuA04Blm8ZUFVrWvt+MQY1SpIkSV3XydKIQ4EH2toDrb5RS7IgydIkS9etW7cjh5AkSZK6YpdeLFdVV1RVf1X1T5o0aVeeWpIkSfolnQThtcBhbe3JrT5JkiRpj9VJEL4NOCLJtCT7AKcBi8a2LEmSJGlsjRiEq2oTcC6wBFgBfLaq7k5ySZL5AElekmQAOBX4hyR3j2XRkiRJ0s7q5K4RVNViYPGQvovatm9jcMmEJEmStEfwyXKSJElqJIOwJEmSGskgLEmSpEYyCEuSJKmRDMKSJElqJIOwJEmSGskgLEmSpEYyCEuSJKmRDMKSJElqJIOwJEmSGskgLEmSpEYyCEuSJKmRDMKSJElqJIOwJEmSGskgLEmSpEYyCEuSJKmRDMKSJElqJIOwJEmSGqmjIJxkXpJ7kqxKcv4w+/dN8s+t/d9NMrXbhUqSJEndNGIQTtIHLAROBo4CTk9y1JBhZwPrq+rXgb8F/rrbhUqSJEnd1MmM8HHAqqpaXVVPAdcBpwwZcwpwTWv7c8ArkqR7ZUqSJEndtVcHYw4FHmhrDwDHb2tMVW1K8ghwCPBg+6AkC4AFreaGJPfsSNHqvcBEhvz3HTfe7//Dafc1rj974OdPuy0/e3u8w4fr7CQId01VXQFcsSvPqbGRZGlV9fe6Dqlp/OxJveFnb3zqZGnEWuCwtvbkVt+wY5LsBRwEPNSNAiVJkqSx0EkQvg04Ism0JPsApwGLhoxZBJzR2v4D4Jaqqu6VKUmSJHXXiEsjWmt+zwWWAH3AVVV1d5JLgKVVtQj4BPDJJKuAnzEYljW+ucRF6g0/e1Jv+Nkbh+LErSRJkprIJ8tJkiSpkQzCkiRJaiSDsCRJkhrJICxJkqRGMgirY0lOTPJvSe5NsjrJ/UlW97ouSZJ2lSQv6nUN6h7vGqGOJVkJ/BlwO7B5S39V+fAUaQwk+Tkw3A/pAFVVB+7ikqTGS/LjqprS6zrUHbv0Ecva4z1SVV/udRFSU1TVs3tdg9REST66rV3Ac3ZlLRpbzgirY0k+yOBDVT4PPLmlv6ru6FlR0jiW5Lnb219VP9tVtUhN0vptzLto+7euzYerauIuLkljxCCsjiW5dZjuqqqX7/JipAZIcj+DSyMyzO6qql/bxSVJjZDkFuDCqvrWMPvur6ppPShLY8AgLEmS1Kb125iNVfV4r2vR2HKNsDqW5CDgYuBlra6vAZdU1SO9q0pqhiQHA0cAE7b0VdXXe1eRNK49y6VHzeDt0zQaVwE/B97Q+noU+MeeViQ1QJK3AV8HlgDvb/35vl7WJI1z/7JlI8kNvSxEY8sgrNF4flVdXFWrW1/vB1yjKI29dwIvAX5UVb8DHAM83NuSpHGtfV2+/86NYwZhjcYTSX5zSyPJicATPaxHaoqNVbURIMm+VbUSeGGPa5LGs9rGtsYZ1whrNN4BXNNaKxzgZ8CZPa1IaoaBJM9h8Ne1/5ZkPfCjHtckjWdHJ3mUwX/r9mttgw+zGXe8a4RGLcmBAFX16EhjJXVXkt8CDgK+UlVP9boeSdqTGYQ1oiRvqapPJTlvuP1V9ZFdXZPUJElOAO6uqp+32gcCR1bVd3tbmSTt2VwjrE4c0Prz2dv4kjS2/h7Y0Nbe0OqTJO0EZ4QlaTeXZFlVzRrSd2dVzexVTZI0HjgjrI4l+ZskBybZO8nNSdYleUuv65IaYHWSP2199vZO8k5gda+LkqQ9nUFYo/HK1gVyrwXWAL8O/HlPK5Ka4Y+A3wDWAgPA8cCCnlYkSeOAt0/TaGz5+/Ia4PqqeiTJ9sZL6oKq+ilwWq/rkKTxxhlhjcYXk6wEjgVuTjIJ2NjjmqRxL8kLWsuR/r3Vnpnkwl7XJUl7Oi+W06gkeS7wSFVtTrI/cGBV/Uev65LGsyRfY3AZ0j9U1TGtvn+vqum9rUyS9mwujdCIkry8qm5J8vq2vvYhn9/1VUmNsn9VfW/I525Tr4qRpPHCIKxO/BZwC/C6YfYVBmFprD2Y5PkMft5I8gfAT3pbkiTt+VwaIUm7uSS/BlzB4J0j1gP3A2+uqh/1tDBJ2sN5sZw6luSyJM9pax+c5AO9rElqgqpaXVUnAZOAFzH4W5rf7G1VkrTnMwhrNE6uqoe3NKpqPfDqHtYjjWutB9hckOR/JZkLPA6cAawC3tDb6iRpz+caYY1GX5J9q+pJgCT7Afv2uCZpPPskg0shvg28HXgvEOD3qmpZLwuTpPHAIKzRuJbB+wf/Y6t9FnBND+uRxrtfq6oZAEmuZPACuSlV5f27JakLDMLqWFX9dZIfACe1ui6tqiW9rEka557estG6d/eAIViSuse7RmhUkhwOHFFVN7UeqNFXVT/vdV3SeJRkM/DYliawH4PrhANUVR3Yq9okaTxwRlgdS/J2YAHwXOD5wKHAx4FX9LIuabyqqr5e1yBJ45l3jdBonAOcCDwKUFX3Ab/a04okSZJ2kEFYo/FkVT21pZFkL1pPupIkSdrTGIQ1Gl9L8h5gv9Y9Ta8H/rXHNUmSJO0QL5ZTx5IEeBvwSgYv1lkCXFn+JZIkSXsgg7A6kqQPuLuqXtTrWiRJkrrBpRHqSFVtBu5JMqXXtUiSJHWDt0/TaBwM3J3ke/znvU2pqvm9K0mSJGnHGIQ1Gn/Z6wIkSZK6xSCsESWZAPwR8OvAXcAnqmpTb6uSJEnaOV4spxEl+WfgaeAbwMnAj6rqnb2tSpIkaecYhDWiJHdV1YzW9l7A96pqdo/LkiRJ2ineNUKdeHrLhksiJEnSeOGMsEaUZDP/eZeIAPsBj7e2q6oO7FVtkiRJO8ogLEmSpEZyaYQkSZIaySAsSZKkRjIIS5IkqZEMwpIkSWokg7AkSZIa6f8Dv8KG6+LFJ2kAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 864x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "for col in ref_test.columns:\n",
    "    pd.DataFrame({'Reference': ref_test[col], 'BertPunc': eval_test[col]}).plot.bar(\n",
    "        title=col, figsize=(12, 4))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Test ASR"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 14/14 [00:09<00:00,  1.64it/s]\n"
     ]
    },
    {
     "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>COMMA</th>\n",
       "      <th>PERIOD</th>\n",
       "      <th>QUESTION</th>\n",
       "      <th>OVERALL</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Precision</th>\n",
       "      <td>0.485539</td>\n",
       "      <td>0.771995</td>\n",
       "      <td>0.469388</td>\n",
       "      <td>0.575641</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Recall</th>\n",
       "      <td>0.694236</td>\n",
       "      <td>0.770087</td>\n",
       "      <td>0.657143</td>\n",
       "      <td>0.707155</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>F1</th>\n",
       "      <td>0.571429</td>\n",
       "      <td>0.771040</td>\n",
       "      <td>0.547619</td>\n",
       "      <td>0.630029</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              COMMA    PERIOD  QUESTION   OVERALL\n",
       "Precision  0.485539  0.771995  0.469388  0.575641\n",
       "Recall     0.694236  0.770087  0.657143  0.707155\n",
       "F1         0.571429  0.771040  0.547619  0.630029"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_pred_test_asr, y_true_test_asr = predictions(data_loader_test_asr)\n",
    "eval_test_asr = evaluation(y_pred_test_asr, y_true_test_asr)\n",
    "eval_test_asr"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "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>COMMA</th>\n",
       "      <th>PERIOD</th>\n",
       "      <th>QUESTION</th>\n",
       "      <th>OVERALL</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Precision</th>\n",
       "      <td>0.596</td>\n",
       "      <td>0.707</td>\n",
       "      <td>0.607</td>\n",
       "      <td>0.660</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Recall</th>\n",
       "      <td>0.429</td>\n",
       "      <td>0.720</td>\n",
       "      <td>0.486</td>\n",
       "      <td>0.573</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>F1</th>\n",
       "      <td>0.499</td>\n",
       "      <td>0.714</td>\n",
       "      <td>0.540</td>\n",
       "      <td>0.614</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           COMMA  PERIOD  QUESTION  OVERALL\n",
       "Precision  0.596   0.707     0.607    0.660\n",
       "Recall     0.429   0.720     0.486    0.573\n",
       "F1         0.499   0.714     0.540    0.614"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# https://www.isca-speech.org/archive/Interspeech_2016/pdfs/1517.PDF\n",
    "ref_test_asr = pd.DataFrame({\n",
    "    'COMMA':    [0.596, 0.429, 0.499],\n",
    "    'PERIOD':   [0.707, 0.720, 0.714],\n",
    "    'QUESTION': [0.607, 0.486, 0.540],\n",
    "    'OVERALL':  [0.660, 0.573, 0.614]\n",
    "}, index=['Precision', 'Recall', 'F1'])\n",
    "ref_test_asr"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAsYAAAErCAYAAADdWkV0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvOIA7rQAAHctJREFUeJzt3Xuc1XW97/HXpwHE+wWxh4rIVGhQjAKjSZ3KbYr4yHDbLrceT9rORB9JZhcfSp3wcsou2vZsDTNOWaQWae5jbMMkLzvbOy8MiVuREKIpB+w4olkot8HP+WOtoeWwhlkDa83Cmdfz8ZiH6/f7fddvvWd0DW9/fNf3F5mJJEmSNNC9od4BJEmSpJ2BxViSJEnCYixJkiQBFmNJkiQJsBhLkiRJgMVYkiRJAizGkiRJEmAxlqSqiYiPRsQTEfFKRPwpIr4VEftExBkR0RoR0WX8oIh4LiJOjohjI+LViFjb5WtScey/R8T64r7nI+JfI+LAbjJkRJzWZf+xEdHWTe7vR8SXqvmzkKTXI4uxJFVBRHwW+BpwMbA3cAxwKPAL4GfAPsB7uzxtCpDAz4vbqzNzjy5fD5WMn56ZewBvAfYArikT5WzgheI/JUm9YDGWpB0UEXsBVwCfzMyfZ+amzGwFTqNQjj8E3Aac1eWpZwG3ZmZHb14vM/8M3Akc2SXHoRTK9zTgxIh443Z8O5I0YFmMJWnHvRMYCvxr6c7MXAvcDZwAzAE+FBG7AkTE3sAHgB/09sUiYhjwQWBFl0NnAS2ZeQewFDizt+eWpIHMYixJO25/4Plurvw+C+yfmf8J/D/g1OL+04CnM3NxydiDIuLPXb52Lzl+XUS8BDxffM1Pdnmts4AfFh//EKdTSFKvWIwlacc9D+wfEYPKHDuweBwKV4c7p1N8hMJV5FKrM3OfLl8vlxy/MDP3BpqAfYERnQci4l1AIzC3uOuHwLiIeM10C0lS9yzGkrTjHgI2UJjesEXxau9JwH3FXT8A3ldcaeIY/nZ1t1cy8wngS8CskpUuzgYCWBwRfwIeKe7vOq9ZktQNi7Ek7aDMfInCh++uj4gpETE4IkYBtwNtwM3FcX8A/gP4EfCLzPzTDrzsHOAAYGpEDKUwNWMahQ/kdX59Ejiz9Ep2RAzt8tVZrBu67B+yA9kk6XXJYixJVZCZXwc+T2EJtb9QuGL7DPC+zNxQMnQOhZUqyn3o7qAy6xj/QzevtxG4Dvgi8PfAOuAHmfmnzi/gu0ADhWXhAA4ujiv9enPx2KVd9t+/HT8GSXpdi8ysdwZJkiSp7rxiLEmSJGExliRJkgCLsSRJkgRYjCVJkiQAyi1G3yf233//HDVqVL1eXpIkSQPEokWLns/M4T2Nq1sxHjVqFC0tLfV6eUmSJA0QEfGHSsY5lUKSJEnCYixJkiQBFmNJkiQJqOMcY0mSpIFq06ZNtLW1sX79+npH6VeGDh3KiBEjGDx48HY932IsSZLUx9ra2thzzz0ZNWoUEVHvOP1CZrJmzRra2tpobGzcrnM4lUKSJKmPrV+/nmHDhlmKqygiGDZs2A5dhbcYS5Ik1YGluPp29GdqMZYkSZKocI5xREwB/gVoAL6TmV/tcvxa4O+Km7sBB2TmPtUMKkmS1F+NuvRnVT1f61ff3+OYhoYGxo0bR0dHB42Njdx8883ss8+269t1113Ht771LSZMmMCtt95arbg7jR6LcUQ0ALOAE4A2YGFEzMvMpzrHZOanS8Z/Ehhfg6ySpMv3rneC2rr8pXonkAaMXXfdlcWLFwNw9tlnM2vWLL7whS9s8zk33HADd999d8Ufbuvo6GDQoNfPWg+VTKU4GliRmSszcyMwFzhlG+PPAH5UjXCSJEmqvUmTJrFq1aot21dffTVHHXUUTU1NXHbZZQCcf/75rFy5kqlTp3Lttdfy8ssv87GPfYyjjjqK8ePH89Of/hSA73//+3z4wx/mAx/4AJMnT+72fK2trYwZM4Zzzz2Xt73tbUyePJl169YBsGLFCo4//niOOOIIJkyYwO9+97tuz1NNlRTjg4FnSrbbivu2EhGHAo3A/d0cnxYRLRHR0t7e3tuskiRJqrLNmzdz3333MXXqVAAWLFjA8uXLefTRR1m8eDGLFi3iwQcf5MYbb+Sggw7igQce4NOf/jRf/vKXOe6441i4cCEPPPAAF198MS+//DIADz30EHPmzOH+++/v9nwAy5cv54ILLmDJkiXss88+3HHHHQCceeaZXHDBBTz++OP8+te/5sADD9zmeaqlkmvb5T7el92MPR34SWZuLncwM2cDswGam5u7O4ckSZJqbN26dRx55JG0trYyceJETjjhBKBQjBcsWMD48YWZsWvXrmX58uW85z3vec3zFyxYwLx587jmmmuAwhJ0f/zjHwE44YQT2G+//bZ5vpEjR9LY2MiRRx4JwMSJE2ltbeWvf/0rq1at4tRTTwUKN+3oTa4dUUkxbgMOKdkeAazuZuzpwAU7GkqSJEm11TnH+KWXXuLkk09m1qxZXHjhhWQmM2bM4Lzzztvm8zOTO+64g8MPP/w1+x955BF2333314wrd77W1lZ22WWXLdsNDQ2sW7eOzPLXTivNtSMqmUqxEBgdEY0RMYRC+Z3XdVBEHA7sCzxU3YiSJEmqlb333pvrrruOa665hk2bNnHiiSdy0003sXbtWgBWrVrFc889t9XzTjzxRK6//votRfaxxx4re/5Kz9dpr732YsSIEdx5550AbNiwgVdeeaXX59kePV4xzsyOiJgO3ENhubabMnNJRFwJtGRmZ0k+A5ib3dV8SZIklVXJ8mq1NH78eI444gjmzp3LRz7yEZYuXcqkSZMA2GOPPbjllls44IADXvOcL37xi1x00UU0NTWRmYwaNYq77rprq3NPnjy57PkaGhq6zXPzzTdz3nnnMXPmTAYPHsztt9/e7Xm65toRUa8e29zcnC0tLXV5bUl63XK5NqlfWLp0KWPGjKl3jH6p3M82IhZlZnNPz/XOd5IkSRIWY0mSJAmwGEuSJEmAxViSJEkCLMaSJEkSUNkNPrQdRl36s3pHqKl6LysjSZJUbRZjSZKkeqv2UowVLH3Y0NDAuHHjyEwaGhr45je/yTvf+c5evcxVV13F5z//+a3O2dHRwZgxY5gzZw677bZbr+PXi1MpJEmSBqDOW0I//vjjfOUrX2HGjBkVPzczefXVV7nqqqvKnvPJJ59kyJAh3HjjjdWOXVMWY0mSpAHuL3/5C/vuu++W7auvvpqjjjqKpqYmLrvsMgBaW1sZM2YMn/jEJ5gwYQLnnHMO69at48gjj+TMM8/c6pzvfve7WbFiBa2trbz97W/fsv+aa67h8ssvB+DYY4/lkksu4eijj+awww7jV7/6FQCbN2/mc5/7HOPGjaOpqYnrr7++ht/93ziVQpIkaQDqLLXr16/n2Wef5f777wdgwYIFLF++nEcffZTMZOrUqTz44IOMHDmSZcuW8b3vfY8bbrgBgNtvv53Fixdvde6Ojg7uvvtupkyZ0mOOjo4OHn30UebPn88VV1zBvffey+zZs/n973/PY489xqBBg3jhhReq+813w2IsSZI0AHVOewB46KGHOOuss3jyySdZsGABCxYsYPz48QCsXbuW5cuXM3LkSA499FCOOeaYbs/ZWbahcMX4nHPOYfXq1dvM8cEPfhCAiRMn0traCsC9997L+eefz6BBhaq633777dD3WimLsSRJ0gA3adIknn/+edrb28lMZsyYwXnnnfeaMa2trey+++7bPE9p2e40aNAgXn311S3b69evf83xXXbZBSh8cK+jowMozGGOiO3+fraXc4wlSZIGuN/+9rds3ryZYcOGceKJJ3LTTTexdu1aAFatWsVzzz1X9nmDBw9m06ZN2zz3G9/4Rp577jnWrFnDhg0buOuuu3rMM3nyZG688cYtRdmpFJIkSQNFBcurVVvptIfMZM6cOTQ0NDB58mSWLl3KpEmTANhjjz245ZZbaGho2Ooc06ZNo6mpiQkTJnDrrbeWfZ3Bgwczc+ZM3vGOd9DY2Mhb3/rWHrN9/OMf5+mnn6apqYnBgwdz7rnnMn369B34bisTmVnzFymnubk5W1pa6vLafcEbfEiqiWqvdbqzqUM5kCpS5ffe0hNvY8yhB1T1nDvkoPH1TlA1S5cuZcyYMa/ZFxGLMrO5p+c6lUKSJEnCYixJkiQBFmNJkqQ6SOo1nbU/29GfqcVYkiSpjw19aSVrXu6wHFdRZrJmzRqGDh263edwVQpJkrTD+v2Hzre/a5U14jdfo41LaN/7TUDfr9e7lZeW1jtBVQwdOpQRI0Zs9/MrKsYRMQX4F6AB+E5mfrXMmNOAy4EEHs/M/77dqSRJkvqxwRv/TOPDM+od429cEQaooBhHRAMwCzgBaAMWRsS8zHyqZMxoYAbwrsx8MSJ2ovVHJEmSpJ5VMsf4aGBFZq7MzI3AXOCULmPOBWZl5osAmVn+9iiSJEnSTqqSYnww8EzJdltxX6nDgMMi4j8j4uHi1IutRMS0iGiJiJb29vbtSyxJkiTVQCXFuNyM8K4foRwEjAaOBc4AvhMR+2z1pMzZmdmcmc3Dhw/vbVZJkiSpZiopxm3AISXbI4DVZcb8NDM3ZebvgWUUirIkSZL0ulBJMV4IjI6IxogYApwOzOsy5k7g7wAiYn8KUytWVjOoJEmSVEs9FuPM7ACmA/cAS4HbMnNJRFwZEVOLw+4B1kTEU8ADwMWZuaZWoSVJkqRqq2gd48ycD8zvsm9myeMEPlP8kiRJkl53vCW0JEmShLeE1va6fO96J6gd7/4jSdKAZDGW1K+MuvRn9Y5QU61D651Akvovp1JIkiRJWIwlSZIkwGIsSZIkARZjSZIkCbAYS5IkSYDFWJIkSQIsxpIkSRJgMZYkSZIAi7EkSZIEWIwlSZIkwGIsSZIkARZjSZIkCbAYS5IkSYDFWJIkSQIsxpIkSRJgMZYkSZIAi7EkSZIEVFiMI2JKRCyLiBURcWmZ4x+NiPaIWFz8+nj1o0qSJEm1M6inARHRAMwCTgDagIURMS8zn+oy9MeZOb0GGSVJkqSaq+SK8dHAisxcmZkbgbnAKbWNJUmSJPWtSorxwcAzJdttxX1d/UNE/FdE/CQiDil3ooiYFhEtEdHS3t6+HXElSZKk2qikGEeZfdll+9+AUZnZBNwLzCl3osycnZnNmdk8fPjw3iWVJEmSaqiSYtwGlF4BHgGsLh2QmWsyc0Nx8/8AE6sTT5IkSeoblRTjhcDoiGiMiCHA6cC80gERcWDJ5lRgafUiSpIkSbXX46oUmdkREdOBe4AG4KbMXBIRVwItmTkPuDAipgIdwAvAR2uYWZIkSaq6HosxQGbOB+Z32Tez5PEMYEZ1o0mSJEl9xzvfSZIkSViMJUmSJMBiLEmSJAEWY0mSJAmwGEuSJEmAxViSJEkCLMaSJEkSYDGWJEmSAIuxJEmSBFiMJUmSJMBiLEmSJAEWY0mSJAmwGEuSJEmAxViSJEkCLMaSJEkSYDGWJEmSAIuxJEmSBFiMJUmSJMBiLEmSJAEWY0mSJAmwGEuSJElAhcU4IqZExLKIWBERl25j3IciIiOiuXoRJUmSpNrrsRhHRAMwCzgJGAucERFjy4zbE7gQeKTaISVJkqRaq+SK8dHAisxcmZkbgbnAKWXG/S/g68D6KuaTJEmS+kQlxfhg4JmS7bbivi0iYjxwSGbeta0TRcS0iGiJiJb29vZeh5UkSZJqpZJiHGX25ZaDEW8ArgU+29OJMnN2ZjZnZvPw4cMrTylJkiTVWCXFuA04pGR7BLC6ZHtP4O3Av0dEK3AMMM8P4EmSJOn1pJJivBAYHRGNETEEOB2Y13kwM1/KzP0zc1RmjgIeBqZmZktNEkuSJEk10GMxzswOYDpwD7AUuC0zl0TElRExtdYBJUmSpL4wqJJBmTkfmN9l38xuxh6747EkSZKkvuWd7yRJkiQsxpIkSRJgMZYkSZIAi7EkSZIEWIwlSZIkwGIsSZIkARZjSZIkCbAYS5IkSYDFWJIkSQIsxpIkSRJgMZYkSZIAi7EkSZIEWIwlSZIkwGIsSZIkARZjSZIkCbAYS5IkSYDFWJIkSQIsxpIkSRJgMZYkSZIAi7EkSZIEVFiMI2JKRCyLiBURcWmZ4+dHxBMRsTgi/iMixlY/qiRJklQ7PRbjiGgAZgEnAWOBM8oU3x9m5rjMPBL4OvDPVU8qSZIk1VAlV4yPBlZk5srM3AjMBU4pHZCZfynZ3B3I6kWUJEmSam9QBWMOBp4p2W4D3tF1UERcAHwGGAIcV+5EETENmAYwcuTI3maVJEmSaqaSK8ZRZt9WV4Qzc1Zmvhm4BPif5U6UmbMzszkzm4cPH967pJIkSVINVVKM24BDSrZHAKu3MX4u8Pc7EkqSJEnqa5UU44XA6IhojIghwOnAvNIBETG6ZPP9wPLqRZQkSZJqr8c5xpnZERHTgXuABuCmzFwSEVcCLZk5D5geEccDm4AXgbNrGVqSJEmqtko+fEdmzgfmd9k3s+Txp6qcS5IkSepT3vlOkiRJwmIsSZIkARZjSZIkCbAYS5IkSYDFWJIkSQIsxpIkSRJgMZYkSZIAi7EkSZIEWIwlSZIkwGIsSZIkARZjSZIkCbAYS5IkSYDFWJIkSQIsxpIkSRJgMZYkSZIAi7EkSZIEWIwlSZIkwGIsSZIkARZjSZIkCbAYS5IkSUCFxTgipkTEsohYERGXljn+mYh4KiL+KyLui4hDqx9VkiRJqp0ei3FENACzgJOAscAZETG2y7DHgObMbAJ+Any92kElSZKkWqrkivHRwIrMXJmZG4G5wCmlAzLzgcx8pbj5MDCiujElSZKk2qqkGB8MPFOy3Vbc151zgLvLHYiIaRHREhEt7e3tlaeUJEmSaqySYhxl9mXZgRH/A2gGri53PDNnZ2ZzZjYPHz688pSSJElSjQ2qYEwbcEjJ9ghgdddBEXE88AXgvZm5oTrxJEmSpL5RyRXjhcDoiGiMiCHA6cC80gERMR74NjA1M5+rfkxJkiSptnosxpnZAUwH7gGWArdl5pKIuDIiphaHXQ3sAdweEYsjYl43p5MkSZJ2SpVMpSAz5wPzu+ybWfL4+CrnkiRJkvqUd76TJEmSsBhLkiRJgMVYkiRJAizGkiRJEmAxliRJkgCLsSRJkgRYjCVJkiTAYixJkiQBFmNJkiQJsBhLkiRJgMVYkiRJAizGkiRJEmAxliRJkgCLsSRJkgRYjCVJkiTAYixJkiQBFmNJkiQJsBhLkiRJgMVYkiRJAizGkiRJElBhMY6IKRGxLCJWRMSlZY6/JyJ+ExEdEfGh6seUJEmSaqvHYhwRDcAs4CRgLHBGRIztMuyPwEeBH1Y7oCRJktQXBlUw5mhgRWauBIiIucApwFOdAzKztXjs1RpklCRJkmqukqkUBwPPlGy3Fff1WkRMi4iWiGhpb2/fnlNIkiRJNVFJMY4y+3J7XiwzZ2dmc2Y2Dx8+fHtOIUmSJNVEJcW4DTikZHsEsLo2cSRJkqT6qKQYLwRGR0RjRAwBTgfm1TaWJEmS1Ld6LMaZ2QFMB+4BlgK3ZeaSiLgyIqYCRMRREdEGfBj4dkQsqWVoSZIkqdoqWZWCzJwPzO+yb2bJ44UUplhIkiRJr0ve+U6SJEnCYixJkiQBFmNJkiQJsBhLkiRJgMVYkiRJAizGkiRJEmAxliRJkgCLsSRJkgRYjCVJkiTAYixJkiQBFmNJkiQJsBhLkiRJgMVYkiRJAizGkiRJEmAxliRJkgCLsSRJkgRYjCVJkiTAYixJkiQBFmNJkiQJsBhLkiRJgMVYkiRJAiosxhExJSKWRcSKiLi0zPFdIuLHxeOPRMSoageVJEmSaqnHYhwRDcAs4CRgLHBGRIztMuwc4MXMfAtwLfC1ageVJEmSaqmSK8ZHAysyc2VmbgTmAqd0GXMKMKf4+CfA+yIiqhdTkiRJqq1BFYw5GHimZLsNeEd3YzKzIyJeAoYBz5cOiohpwLTi5tqIWLY9oVV/AfvT5d9vv3GF/0+nnVe/fu+B7z/ttHzvve4dWsmgSopxuZ9UbscYMnM2MLuC19ROLiJaMrO53jmkgcb3nlQfvvcGhkqmUrQBh5RsjwBWdzcmIgYBewMvVCOgJEmS1BcqKcYLgdER0RgRQ4DTgXldxswDzi4+/hBwf2ZudcVYkiRJ2ln1OJWiOGd4OnAP0ADclJlLIuJKoCUz5wHfBW6OiBUUrhSfXsvQ2ik4JUaqD997Un343hsAwgu7kiRJkne+kyRJkgCLsSRJkgRYjCVJkiTAYixJkiQBld3gQ9oiIg6mcPeYLf/tZOaD9Usk9V8R8ZltHc/Mf+6rLJIKIuKtmfnbeudQbViMVbGI+Brwj8BTwObi7gQsxlJt7FnvAJK2sgAYWe8Qqg2Xa1PFImIZ0JSZG+qdRZKkWomI67o7BJydmXv1ZR71Ha8YqzdWAoMBi7HUB7bxhzMAmXlhX2WRBph/Aj5L+T/vzujjLOpDFmP1xivA4oi4j5JfFv7hLNXMonoHkAaohcCTmfnrrgci4vK+j6O+4lQKVSwizi63PzPn9HUWSZJqJSL2A9Zn5iv1zqK+ZTFWr0TEEOCw4uayzNxUzzzSQBARw4FLgLHA0M79mXlc3UJJ/VhEjMzMP9Y7h/qe6xirYhFxLLAcmAXcADwdEe+payhpYLgVWAo0AlcArRT+qldSbdzZ+SAi7qhnEPUt5xirN74BTM7MZQARcRjwI2BiXVNJ/d+wzPxuRHwqM38J/DIiflnvUFI/FiWP31S3FOpzFmP1xuDOUgyQmU9HxOB6BpIGiM4pS89GxPuB1cCIOuaR+rvs5rH6OecYq2IRcROFXxA3F3edCQzKzH+qXyqp/4uIk4FfAYcA1wN7AVdk5ry6BpP6qYjYDLxM4crxrhRWZaK4na5j3H9ZjFWxiNgFuAD4bxR+OTwI3OANPyRJUn9gMZaknVxEzAE+lZl/Lm7vC3wjMz9W32SS1L84x1g9iojbMvO0iHiCMnOtMrOpDrGkgaSpsxQDZOaLETG+noEkqT+yGKsSnyr+8+S6ppAGrjdExL6Z+SJsufmAv78lqcr8xaoeZeazxYfPA+sy89XiUm1vBe6uXzJpwPgG8OuI+AmFv7U5DfhyfSNJUv/jHGNVLCIWAe8G9gUeBlqAVzLzzLoGkwaAiBgLHEfhg6/3ZeZTdY4kSf2Od75Tb0TxvvEfBK7PzFMp3KJWUu3tB7ycmdcD7RHRWO9AktTfWIzVGxERkyisX/yz4j6n40g1FhGXAZcAM4q7BgO31C+RJPVPFmP1xkUU/mD+v5m5JCLeBDxQ50zSQHAqMJXCDQfIzNXAnnVNJEn9kFf7VLHM/CXwy5LtlcCF9UskDRgbMzMjIgEiYvd6B5Kk/shirB5FxP/OzIsi4t8ov47x1DrEkgaS2yLi28A+EXEu8DHgO3XOJEn9jqtSqEcRMTEzF0XEe8sdL15JllRDEXECMJnCqhT3ZOYv6hxJkvodi7EqVvzr23WZ+WpxuwHYpbhShaQ+UnzvnZ6Zt9Y7iyT1J374Tr1xH7BbyfauwL11yiL1exGxV0TMiIhvRsTkKJgOrKRwkw9JUhU5x1i9MTQz13ZuZObaiNhtW0+QtENuBl4EHgI+DlwMDAFOyczF9QwmSf2RxVi98XJETMjM30Bh7jGwrs6ZpP7sTZk5DiAivkPhtuwjM/Ov9Y0lSf2TxVi9cRFwe0SsLm4fCPxjHfNI/d2mzgeZuTkifm8plqTa8cN36pWIGAwcTuGT8b/NzE09PEXSdoqIzRRv6kHhPbcr8ErxcWbmXvXKJkn9kcVYFSvOJ/4McGhmnhsRo4HDM/OuOkeTJEnaYa5Kod74HrARmFTcbgO+VL84kiRJ1WMxVm+8OTO/TnHeY2auo/BXupIkSa97FmP1xsaI2JXibaEj4s3AhvpGkiRJqg5XpVBvXAb8HDgkIm4F3gV8tK6JJEmSqsQP36kiERHACAqfiD+GwhSKhzPz+boGkyRJqhKLsSoWEYsyc2K9c0iSJNWCc4zVGw9HxFH1DiFJklQLXjFWxSLiKQo392ilcNOBzpsMNNUzlyRJUjVYjFWxiDi03P7M/ENfZ5EkSao2V6VQjyJiKHA+8BbgCeC7mdlR31SSJEnV5RVj9Sgifkzhph6/Ak4C/pCZn6pvKkmSpOqyGKtHEfFEZo4rPh4EPJqZE+ocS5IkqapclUKV2NT5wCkUkiSpv/KKsXoUEZsprEIBhZUodqVwo4/OVSn2qlc2SZKkarEYS5IkSTiVQpIkSQIsxpIkSRJgMZYkSZIAi7EkSZIEwP8HlJeBIl6OMDIAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 864x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "for col in ref_test_asr.columns:\n",
    "    pd.DataFrame({'Reference': ref_test_asr[col], 'BertPunc': eval_test_asr[col]}).plot.bar(\n",
    "        title=col, figsize=(12, 4))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "torch_gpu_1.5",
   "language": "python",
   "name": "torch_gpu_1.5"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}