update example

example_usage.ipynb

@@ -2,28 +2,30 @@
  "cells": [
   {
    "cell_type": "code",
-   "execution_count": 2,
+   "execution_count": 80,
    "id": "5920c653-448e-43b3-93eb-12d7073ad352",
    "metadata": {
     "tags": []
    },
-   "outputs": [
-    {
-     "name": "stderr",
-     "output_type": "stream",
-     "text": [
-      "/opt/espnet/tools/anaconda/envs/espnet/lib/python3.9/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
-      "  from .autonotebook import tqdm as notebook_tqdm\n",
-      "[nltk_data] Downloading package averaged_perceptron_tagger to\n",
-      "[nltk_data]     /root/nltk_data...\n",
-      "[nltk_data]   Unzipping taggers/averaged_perceptron_tagger.zip.\n",
-      "[nltk_data] Downloading package cmudict to /root/nltk_data...\n",
-      "[nltk_data]   Unzipping corpora/cmudict.zip.\n"
-     ]
-    }
-   ],
+   "outputs": [],
    "source": [
-    "from espnet2.bin.asr_inference import Speech2Text"
+    "from espnet2.bin.asr_inference import Speech2Text\n",
+    "from espnet2.bin.asr_align import CTCSegmentation\n",
+    "import soundfile\n",
+    "import pandas as pd"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 81,
+   "id": "83058587-1a8a-4b01-92ff-e9125fbe55a3",
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [],
+   "source": [
+    "import torch\n",
+    "torch.set_num_threads(1)"
    ]
   },
   {
@@ -36,7 +38,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 3,
+   "execution_count": 82,
    "id": "5e4670d6-0949-48cf-b6b1-d9cc4cf3ad65",
    "metadata": {
     "tags": []
@@ -58,11 +60,45 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 9,
+   "execution_count": 92,
    "id": "e8120e8e-3718-4a1a-ab7a-46ef98a6bc11",
    "metadata": {
     "tags": []
    },
+   "outputs": [],
+   "source": [
+    "speech, rate = soundfile.read(\"example_audio/emt16k.wav\")\n",
+    "assert rate == 16000"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 93,
+   "id": "eec8d4b2-c27a-4780-aeed-8aa7538f70e5",
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "CPU times: user 3.67 s, sys: 191 ms, total: 3.87 s\n",
+      "Wall time: 3.86 s\n"
+     ]
+    }
+   ],
+   "source": [
+    "%time text, *_ = speech2text(speech)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 94,
+   "id": "39f41a8b-94c3-42d6-a989-6c7183a6f94d",
+   "metadata": {
+    "tags": []
+   },
    "outputs": [
     {
      "name": "stdout",
@@ -73,432 +109,434 @@
     }
    ],
    "source": [
-    "import soundfile\n",
-    "speech, rate = soundfile.read(\"example_audio/emt16k.wav\")\n",
-    "assert rate == 16000\n",
-    "text, *_ = speech2text(speech)\n",
     "print(text[0])"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 10,
-   "id": "7cb79f15-3091-446f-9e22-e0eae96e278e",
+   "execution_count": 116,
+   "id": "812060a6-90de-4134-8d1f-9f3d98853bc2",
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "\n",
+      "Input File     : 'example_audio/emt16k.wav'\n",
+      "Channels       : 1\n",
+      "Sample Rate    : 16000\n",
+      "Precision      : 16-bit\n",
+      "Duration       : 00:00:12.74 = 203815 samples ~ 955.383 CDDA sectors\n",
+      "File Size      : 408k\n",
+      "Bit Rate       : 256k\n",
+      "Sample Encoding: 16-bit Signed Integer PCM\n",
+      "\n"
+     ]
+    }
+   ],
+   "source": [
+    "!soxi example_audio/emt16k.wav"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "7d07e8a4-1dbf-4a79-bdf0-aeaeb160ba19",
+   "metadata": {},
+   "source": [
+    "## Example token level alignment"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 95,
+   "id": "e6b7331c-52f1-4162-b564-2e6a08b325b0",
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "WARNING:root:No RNN model detected; memory consumption may be high.\n"
+     ]
+    }
+   ],
+   "source": [
+    "aligner = CTCSegmentation(\"exp/config.yaml\", \"exp/valid.acc.ave_10best.pth\" , kaldi_style_text=False, blank_transition_cost_zero=True)\n",
+    "segments = aligner(speech, text[0].split())"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 96,
+   "id": "e6d18b5f-3d2a-4fcf-bf4e-00480e58094a",
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "utt_0000 utt 0.36 0.78 -0.0001 mina\n",
+      "utt_0001 utt 0.78 1.19 -0.0003 tahaksin\n",
+      "utt_0002 utt 1.19 1.59 -0.0017 homme\n",
+      "utt_0003 utt 1.67 2.19 -0.0001 täna\n",
+      "utt_0004 utt 3.24 3.76 -0.0037 ja\n",
+      "utt_0005 utt 3.76 4.28 -0.0000 homme\n",
+      "utt_0006 utt 5.61 6.13 -0.0001 kui\n",
+      "utt_0007 utt 6.17 6.69 -0.0009 saanud\n",
+      "utt_0008 utt 6.81 7.33 -0.0018 on\n",
+      "utt_0009 utt 7.98 8.50 -0.0862 kui\n",
+      "utt_0010 utt 8.50 9.34 -0.1062 krampsumas\n",
+      "utt_0011 utt 9.34 9.54 -0.1183 ise\n",
+      "utt_0012 utt 9.54 10.07 -0.2033 veiki\n",
+      "utt_0013 utt 10.07 10.31 -0.1041 panna\n",
+      "\n"
+     ]
+    }
+   ],
+   "source": [
+    "print(segments)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "77f82a7d-08dc-40cb-88e5-48ef8c36af7d",
+   "metadata": {
+    "tags": []
+   },
+   "source": [
+    "## Get timestamps with some correction"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 97,
+   "id": "ae9f7e3f-b75d-4bcb-98d1-ae2f037fb4af",
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [],
+   "source": [
+    "def get_timestamps(aligner, speech, text, time_correction=0.2):\n",
+    "    tokens=text.split()\n",
+    "    segments = aligner(speech, tokens)\n",
+    "    df=pd.DataFrame(segments.segments)\n",
+    "    df.columns=['start', 'end', 'confidence']\n",
+    "    df['start']=df.start+time_correction\n",
+    "    df['end']=df.end+time_correction\n",
+    "    df['words']=tokens\n",
+    "    return df"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 128,
+   "id": "93aa6281-3b73-47b7-93ca-e90fedd8d398",
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [],
+   "source": [
+    "torch.set_num_threads(5)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 131,
+   "id": "0215d312-1896-43f1-9782-c92aced787b7",
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "CPU times: user 14.4 s, sys: 89.6 ms, total: 14.5 s\n",
+      "Wall time: 2.9 s\n"
+     ]
+    }
+   ],
+   "source": [
+    "speech, rate = soundfile.read(\"example_audio/oden_kypsis16k_subset2.wav\")\n",
+    "assert rate == 16000\n",
+    "\n",
+    "%time text, *_ = speech2text(speech)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 132,
+   "id": "d31d6840-3a80-411a-969c-05f4a5e3e9a1",
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "\n",
+      "Input File     : 'example_audio/oden_kypsis16k_subset2.wav'\n",
+      "Channels       : 1\n",
+      "Sample Rate    : 16000\n",
+      "Precision      : 16-bit\n",
+      "Duration       : 00:00:09.19 = 146983 samples ~ 688.983 CDDA sectors\n",
+      "File Size      : 294k\n",
+      "Bit Rate       : 256k\n",
+      "Sample Encoding: 16-bit Signed Integer PCM\n",
+      "\n"
+     ]
+    }
+   ],
+   "source": [
+    "!soxi example_audio/oden_kypsis16k_subset2.wav"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 108,
+   "id": "53f3b63f-9b40-432b-b58c-f5b7223252ed",
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "CPU times: user 474 ms, sys: 30.2 ms, total: 504 ms\n",
+      "Wall time: 501 ms\n"
+     ]
+    }
+   ],
+   "source": [
+    "%time df_times=get_timestamps(aligner, speech, text[0])"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 109,
+   "id": "1b4dd747-4be2-4ace-a301-6207f7dd9a71",
    "metadata": {
     "tags": []
    },
    "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>start</th>\n",
+       "      <th>end</th>\n",
+       "      <th>confidence</th>\n",
+       "      <th>words</th>\n",
+       "    </tr>\n",
+       "  </thead>\n",
+       "  <tbody>\n",
+       "    <tr>\n",
+       "      <th>0</th>\n",
+       "      <td>0.260154</td>\n",
+       "      <td>0.661184</td>\n",
+       "      <td>-0.107317</td>\n",
+       "      <td>klikid</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1</th>\n",
+       "      <td>0.661184</td>\n",
+       "      <td>0.821596</td>\n",
+       "      <td>-0.001331</td>\n",
+       "      <td>neid</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>2</th>\n",
+       "      <td>0.822883</td>\n",
+       "      <td>1.784067</td>\n",
+       "      <td>-0.002136</td>\n",
+       "      <td>allserva</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>3</th>\n",
+       "      <td>1.784067</td>\n",
+       "      <td>1.984582</td>\n",
+       "      <td>-0.041078</td>\n",
+       "      <td>tekivad</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>4</th>\n",
+       "      <td>2.547310</td>\n",
+       "      <td>3.067362</td>\n",
+       "      <td>-0.008251</td>\n",
+       "      <td>need</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>5</th>\n",
+       "      <td>3.067362</td>\n",
+       "      <td>4.029833</td>\n",
+       "      <td>-0.007814</td>\n",
+       "      <td>lubaküpsiseid</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>6</th>\n",
+       "      <td>4.752973</td>\n",
+       "      <td>5.273025</td>\n",
+       "      <td>-0.000333</td>\n",
+       "      <td>mis</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>7</th>\n",
+       "      <td>5.273025</td>\n",
+       "      <td>5.413385</td>\n",
+       "      <td>-0.063720</td>\n",
+       "      <td>on</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>8</th>\n",
+       "      <td>5.413385</td>\n",
+       "      <td>5.553745</td>\n",
+       "      <td>-0.000231</td>\n",
+       "      <td>nagu</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>9</th>\n",
+       "      <td>5.553745</td>\n",
+       "      <td>5.834466</td>\n",
+       "      <td>-0.000573</td>\n",
+       "      <td>ilusti</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>10</th>\n",
+       "      <td>5.834466</td>\n",
+       "      <td>6.115187</td>\n",
+       "      <td>-0.001930</td>\n",
+       "      <td>kohati</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>11</th>\n",
+       "      <td>6.236783</td>\n",
+       "      <td>7.037555</td>\n",
+       "      <td>-0.004937</td>\n",
+       "      <td>tõlgitud</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>12</th>\n",
+       "      <td>7.037555</td>\n",
+       "      <td>7.238070</td>\n",
+       "      <td>-0.001092</td>\n",
+       "      <td>eesti</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>13</th>\n",
+       "      <td>7.238070</td>\n",
+       "      <td>7.679202</td>\n",
+       "      <td>-0.003088</td>\n",
+       "      <td>keelde</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>14</th>\n",
+       "      <td>7.800798</td>\n",
+       "      <td>8.320850</td>\n",
+       "      <td>-0.001126</td>\n",
+       "      <td>see</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>15</th>\n",
+       "      <td>8.320850</td>\n",
+       "      <td>8.601571</td>\n",
+       "      <td>-0.033408</td>\n",
+       "      <td>idee</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>16</th>\n",
+       "      <td>8.601571</td>\n",
+       "      <td>9.363527</td>\n",
+       "      <td>-0.032846</td>\n",
+       "      <td>arusaadavamaks</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>17</th>\n",
+       "      <td>9.363527</td>\n",
+       "      <td>9.584093</td>\n",
+       "      <td>-0.390966</td>\n",
+       "      <td>ma</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>18</th>\n",
+       "      <td>9.584093</td>\n",
+       "      <td>9.764557</td>\n",
+       "      <td>-0.053868</td>\n",
+       "      <td>tean</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>19</th>\n",
+       "      <td>9.764557</td>\n",
+       "      <td>9.924968</td>\n",
+       "      <td>-0.000163</td>\n",
+       "      <td>et</td>\n",
+       "    </tr>\n",
+       "  </tbody>\n",
+       "</table>\n",
+       "</div>"
+      ],
       "text/plain": [
-       "('mina tahaksin homme täna ja homme kui saanud on kui krampsumas ise veiki panna',\n",
-       " ['▁mina',\n",
-       "  '▁tahaksin',\n",
-       "  '▁homme',\n",
-       "  '▁täna',\n",
-       "  '▁ja',\n",
-       "  '▁homme',\n",
-       "  '▁kui',\n",
-       "  '▁saanud',\n",
-       "  '▁on',\n",
-       "  '▁kui',\n",
-       "  '▁kra',\n",
-       "  'm',\n",
-       "  'psu',\n",
-       "  'mas',\n",
-       "  '▁ise',\n",
-       "  '▁ve',\n",
-       "  'i',\n",
-       "  'ki',\n",
-       "  '▁panna'],\n",
-       " [160,\n",
-       "  2003,\n",
-       "  1403,\n",
-       "  101,\n",
-       "  4,\n",
-       "  1403,\n",
-       "  13,\n",
-       "  303,\n",
-       "  3,\n",
-       "  13,\n",
-       "  2652,\n",
-       "  60,\n",
-       "  2940,\n",
-       "  410,\n",
-       "  143,\n",
-       "  376,\n",
-       "  15,\n",
-       "  88,\n",
-       "  467],\n",
-       " Hypothesis(yseq=tensor([4999,  160, 2003, 1403,  101,    4, 1403,   13,  303,    3,   13, 2652,\n",
-       "           60, 2940,  410,  143,  376,   15,   88,  467, 4999]), score=tensor(-11.9828), scores={'decoder': tensor(-14.2027), 'ctc': tensor(-9.7630)}, states={'decoder': [tensor([[ -5.1263,   3.1550,  -2.0594,  ...,  24.0211,   0.3968,  -0.0377],\n",
-       "         [ -6.2288,   3.8375,   1.5658,  ...,   1.4794, -19.9161,  -9.2051],\n",
-       "         [ -0.8955, -12.1962,  11.2881,  ...,  14.5357,   2.5597,   5.0430],\n",
-       "         ...,\n",
-       "         [ -1.5868,   0.9150,   7.6028,  ...,   6.9893,  -9.0237,  -6.0684],\n",
-       "         [ -0.2270,  -5.4793,  -4.3761,  ...,   8.3707,   3.3640, -10.0500],\n",
-       "         [ -3.5901,   5.2413,  -6.3700,  ...,   1.8956,  -1.3764,   2.8161]]), tensor([[ -3.6827,   2.9746,  -2.4586,  ...,  21.4053,   0.7093,   0.0590],\n",
-       "         [ -3.6746,   2.3007,   6.9940,  ...,   0.4042, -26.4674,  -7.6587],\n",
-       "         [ -4.1463, -12.4967,  23.4652,  ...,  20.2923,   5.0455,   2.7187],\n",
-       "         ...,\n",
-       "         [ -3.0836,   8.8793,  15.3461,  ...,   6.7483, -10.2614, -12.3279],\n",
-       "         [ -3.0388,  -4.8860,   1.2961,  ...,   7.1166,   7.1762, -11.6877],\n",
-       "         [ -6.3154,  10.7874,  -8.3653,  ...,  -2.7268,   6.2414,   2.0028]]), tensor([[ -4.2097,   3.3128,  -4.3004,  ...,  19.7619,   0.0523,   0.5325],\n",
-       "         [ -2.0206,   5.0642,   3.8430,  ...,   4.3120, -31.3507,  -4.5523],\n",
-       "         [ -4.9023, -11.4819,  30.5385,  ...,  23.1088,   7.1218,   6.8997],\n",
-       "         ...,\n",
-       "         [ -7.4921,  14.9905,  18.8936,  ...,  14.4608, -14.1381, -20.3670],\n",
-       "         [ -7.2482,  -3.5175,  -1.7658,  ...,  13.9724,   3.8404, -19.6644],\n",
-       "         [ -4.5970,  14.7940,   1.5388,  ...,  -3.3481,   2.3485,   1.7171]]), tensor([[ -2.6693,   5.9532,  -6.0862,  ...,  18.1037,   1.8700,   1.8734],\n",
-       "         [ -3.4684,   1.5924,  -2.2865,  ...,  17.7742, -34.5136,  -5.1747],\n",
-       "         [-14.9132, -16.1804,  39.0452,  ...,  19.7372,  15.8912,  11.7538],\n",
-       "         ...,\n",
-       "         [-13.3347,  13.1394,  22.9955,  ...,  15.1552, -26.7795, -35.4029],\n",
-       "         [-20.9897, -12.5458,   2.4344,  ...,  25.7768,   4.4565, -20.4715],\n",
-       "         [-11.9333,  21.3926,   0.9612,  ...,  -7.3184,   1.1050,  -8.7355]]), tensor([[ -8.3364,  -8.5930,  12.5300,  ...,  14.0010,   9.2563,  15.4071],\n",
-       "         [-11.7589,  -8.5604, -14.7583,  ...,  27.1039, -22.6838,   8.2407],\n",
-       "         [-26.8291, -28.9856,  66.4735,  ...,  10.2667,  51.7921,  13.9935],\n",
-       "         ...,\n",
-       "         [-24.8075,   8.7724,  35.3504,  ..., -15.1498, -51.6167, -44.5583],\n",
-       "         [-25.6442,  -8.3698,  14.0967,  ...,  21.2507,  16.2526, -28.0444],\n",
-       "         [ -9.2648,  33.3486,  10.6099,  ...,  -8.6877,   0.9520, -13.3015]]), tensor([[-30.5616,  -7.4708,  56.1848,  ...,   1.5934,  20.7831,  44.0383],\n",
-       "         [-51.2013,  29.8913, -73.0093,  ...,  -2.1801, -11.0278, 151.5184],\n",
-       "         [-74.1030, -51.1745, 158.5696,  ...,  40.8046,  92.3456,  42.8557],\n",
-       "         ...,\n",
-       "         [-56.5551,   4.2527, 102.6477,  ...,   4.8044, -99.0702, -65.3296],\n",
-       "         [-22.1416,   8.2998,  48.0735,  ..., -36.2092,  40.0850,   6.7160],\n",
-       "         [-21.1855,  35.3772, -11.6307,  ...,  17.0477,   6.7013, -32.2759]])], 'ctc': (tensor([[-1.0000e+10, -1.0000e+10],\n",
-       "         [-1.0000e+10, -1.0000e+10],\n",
-       "         [-1.0000e+10, -1.0000e+10],\n",
-       "         [-1.0000e+10, -1.0000e+10],\n",
-       "         [-1.0000e+10, -1.0000e+10],\n",
-       "         [-1.0000e+10, -1.0000e+10],\n",
-       "         [-1.0000e+10, -1.0000e+10],\n",
-       "         [-1.0000e+10, -1.0000e+10],\n",
-       "         [-1.0000e+10, -1.0000e+10],\n",
-       "         [-1.0000e+10, -1.0000e+10],\n",
-       "         [-1.0000e+10, -1.0000e+10],\n",
-       "         [-1.0000e+10, -1.0000e+10],\n",
-       "         [-1.0000e+10, -1.0000e+10],\n",
-       "         [-1.0000e+10, -1.0000e+10],\n",
-       "         [-1.0000e+10, -1.0000e+10],\n",
-       "         [-1.0000e+10, -1.0000e+10],\n",
-       "         [-1.0000e+10, -1.0000e+10],\n",
-       "         [-1.0000e+10, -1.0000e+10],\n",
-       "         [-1.0000e+10, -1.0000e+10],\n",
-       "         [-3.9543e+02, -1.0000e+10],\n",
-       "         [-3.9364e+02, -3.9543e+02],\n",
-       "         [-3.9464e+02, -3.9349e+02],\n",
-       "         [-3.9263e+02, -3.9321e+02],\n",
-       "         [-3.8944e+02, -3.9219e+02],\n",
-       "         [-3.8704e+02, -3.8938e+02],\n",
-       "         [-3.8350e+02, -3.8695e+02],\n",
-       "         [-3.8215e+02, -3.8347e+02],\n",
-       "         [-3.8234e+02, -3.8191e+02],\n",
-       "         [-3.8383e+02, -3.8141e+02],\n",
-       "         [-3.8014e+02, -3.9431e+02],\n",
-       "         [-3.8677e+02, -3.8014e+02],\n",
-       "         [-3.8578e+02, -3.8013e+02],\n",
-       "         [-3.8176e+02, -3.8013e+02],\n",
-       "         [-3.8125e+02, -3.7995e+02],\n",
-       "         [-3.8104e+02, -3.7971e+02],\n",
-       "         [-3.7952e+02, -3.7948e+02],\n",
-       "         [-3.7984e+02, -3.7881e+02],\n",
-       "         [-3.8364e+02, -3.7850e+02],\n",
-       "         [-3.8166e+02, -3.7851e+02],\n",
-       "         [-3.8023e+02, -3.8624e+02],\n",
-       "         [-3.8322e+02, -3.8023e+02],\n",
-       "         [-3.8385e+02, -3.8019e+02],\n",
-       "         [-3.8177e+02, -3.8016e+02],\n",
-       "         [-3.7862e+02, -3.7998e+02],\n",
-       "         [-3.7735e+02, -3.7839e+02],\n",
-       "         [-3.7684e+02, -3.7705e+02],\n",
-       "         [-3.7739e+02, -3.7624e+02],\n",
-       "         [-3.7643e+02, -3.7597e+02],\n",
-       "         [-3.7122e+02, -3.7548e+02],\n",
-       "         [-3.6641e+02, -3.7120e+02],\n",
-       "         [-3.6524e+02, -3.6640e+02],\n",
-       "         [-3.6296e+02, -3.6497e+02],\n",
-       "         [-3.6198e+02, -3.6283e+02],\n",
-       "         [-3.5907e+02, -3.6162e+02],\n",
-       "         [-3.4759e+02, -3.6746e+02],\n",
-       "         [-3.4178e+02, -3.4759e+02],\n",
-       "         [-3.4269e+02, -3.4178e+02],\n",
-       "         [-3.4194e+02, -3.4144e+02],\n",
-       "         [-3.3753e+02, -3.4097e+02],\n",
-       "         [-3.3310e+02, -3.3750e+02],\n",
-       "         [-3.2918e+02, -3.3309e+02],\n",
-       "         [-3.2564e+02, -3.2916e+02],\n",
-       "         [-3.2297e+02, -3.2561e+02],\n",
-       "         [-3.2136e+02, -3.2290e+02],\n",
-       "         [-3.1502e+02, -3.2117e+02],\n",
-       "         [-3.1264e+02, -3.1502e+02],\n",
-       "         [-3.1198e+02, -3.1255e+02],\n",
-       "         [-3.1081e+02, -3.1153e+02],\n",
-       "         [-3.1095e+02, -3.1041e+02],\n",
-       "         [-3.0918e+02, -3.0995e+02],\n",
-       "         [-3.0741e+02, -3.0880e+02],\n",
-       "         [-3.0711e+02, -3.0719e+02],\n",
-       "         [-3.0739e+02, -3.0645e+02],\n",
-       "         [-3.0782e+02, -3.0612e+02],\n",
-       "         [-3.0775e+02, -3.0595e+02],\n",
-       "         [-3.0780e+02, -3.0580e+02],\n",
-       "         [-3.0755e+02, -3.0567e+02],\n",
-       "         [-3.0734e+02, -3.0553e+02],\n",
-       "         [-3.0679e+02, -3.0538e+02],\n",
-       "         [-3.0643e+02, -3.0516e+02],\n",
-       "         [-3.0644e+02, -3.0491e+02],\n",
-       "         [-3.0617e+02, -3.0472e+02],\n",
-       "         [-3.0556e+02, -3.0451e+02],\n",
-       "         [-3.0395e+02, -3.0421e+02],\n",
-       "         [-3.0080e+02, -3.0338e+02],\n",
-       "         [-2.9807e+02, -3.0073e+02],\n",
-       "         [-2.9489e+02, -2.9800e+02],\n",
-       "         [-2.9230e+02, -2.9485e+02],\n",
-       "         [-2.9033e+02, -2.9223e+02],\n",
-       "         [-2.8882e+02, -2.9019e+02],\n",
-       "         [-2.9074e+02, -2.8859e+02],\n",
-       "         [-2.9428e+02, -2.8848e+02],\n",
-       "         [-2.8970e+02, -2.8848e+02],\n",
-       "         [-2.8296e+02, -2.9165e+02],\n",
-       "         [-2.9051e+02, -2.8296e+02],\n",
-       "         [-2.8984e+02, -2.8296e+02],\n",
-       "         [-2.8508e+02, -2.8296e+02],\n",
-       "         [-2.8530e+02, -2.8285e+02],\n",
-       "         [-2.8455e+02, -2.8277e+02],\n",
-       "         [-2.8364e+02, -2.8261e+02],\n",
-       "         [-2.8490e+02, -2.8231e+02],\n",
-       "         [-2.8613e+02, -2.8223e+02],\n",
-       "         [-2.8638e+02, -2.8221e+02],\n",
-       "         [-2.8644e+02, -2.8220e+02],\n",
-       "         [-2.8776e+02, -2.8218e+02],\n",
-       "         [-2.9038e+02, -2.8218e+02],\n",
-       "         [-2.8499e+02, -2.9242e+02],\n",
-       "         [-2.9040e+02, -2.8500e+02],\n",
-       "         [-2.9231e+02, -2.8500e+02],\n",
-       "         [-2.9122e+02, -2.8500e+02],\n",
-       "         [-2.8800e+02, -2.8500e+02],\n",
-       "         [-2.8509e+02, -2.8495e+02],\n",
-       "         [-2.8517e+02, -2.8432e+02],\n",
-       "         [-2.8559e+02, -2.8397e+02],\n",
-       "         [-2.8567e+02, -2.8379e+02],\n",
-       "         [-2.8590e+02, -2.8365e+02],\n",
-       "         [-2.8608e+02, -2.8355e+02],\n",
-       "         [-2.8644e+02, -2.8347e+02],\n",
-       "         [-2.8695e+02, -2.8342e+02],\n",
-       "         [-2.8715e+02, -2.8339e+02],\n",
-       "         [-2.8738e+02, -2.8337e+02],\n",
-       "         [-2.8759e+02, -2.8335e+02],\n",
-       "         [-2.8722e+02, -2.8334e+02],\n",
-       "         [-2.8677e+02, -2.8331e+02],\n",
-       "         [-2.8624e+02, -2.8328e+02],\n",
-       "         [-2.8576e+02, -2.8323e+02],\n",
-       "         [-2.8529e+02, -2.8316e+02],\n",
-       "         [-2.8409e+02, -2.8304e+02],\n",
-       "         [-2.8393e+02, -2.8274e+02],\n",
-       "         [-2.8336e+02, -2.8248e+02],\n",
-       "         [-2.8150e+02, -2.8213e+02],\n",
-       "         [-2.7884e+02, -2.8107e+02],\n",
-       "         [-2.7735e+02, -2.7874e+02],\n",
-       "         [-2.7646e+02, -2.7713e+02],\n",
-       "         [-2.7592e+02, -2.7605e+02],\n",
-       "         [-2.7555e+02, -2.7529e+02],\n",
-       "         [-2.7501e+02, -2.7472e+02],\n",
-       "         [-2.7359e+02, -2.7416e+02],\n",
-       "         [-2.7206e+02, -2.7314e+02],\n",
-       "         [-2.7034e+02, -2.7176e+02],\n",
-       "         [-2.6631e+02, -2.7013e+02],\n",
-       "         [-2.6184e+02, -2.6629e+02],\n",
-       "         [-2.5695e+02, -2.6183e+02],\n",
-       "         [-2.5373e+02, -2.5694e+02],\n",
-       "         [-2.5259e+02, -2.5369e+02],\n",
-       "         [-2.5437e+02, -2.5230e+02],\n",
-       "         [-2.5365e+02, -2.5218e+02],\n",
-       "         [-2.5339e+02, -2.5198e+02],\n",
-       "         [-2.5424e+02, -2.5176e+02],\n",
-       "         [-2.5564e+02, -2.5168e+02],\n",
-       "         [-2.5749e+02, -2.5166e+02],\n",
-       "         [-2.5639e+02, -2.5166e+02],\n",
-       "         [-2.4779e+02, -2.5773e+02],\n",
-       "         [-2.5443e+02, -2.4779e+02],\n",
-       "         [-2.5357e+02, -2.4779e+02],\n",
-       "         [-2.4874e+02, -2.4779e+02],\n",
-       "         [-2.4432e+02, -2.4747e+02],\n",
-       "         [-2.3843e+02, -2.4428e+02],\n",
-       "         [-2.3581e+02, -2.3843e+02],\n",
-       "         [-2.3313e+02, -2.3574e+02],\n",
-       "         [-2.3312e+02, -2.3306e+02],\n",
-       "         [-2.3310e+02, -2.3239e+02],\n",
-       "         [-2.2699e+02, -2.3199e+02],\n",
-       "         [-2.2357e+02, -2.2699e+02],\n",
-       "         [-2.1992e+02, -2.2354e+02],\n",
-       "         [-2.1943e+02, -2.1989e+02],\n",
-       "         [-2.1140e+02, -2.2310e+02],\n",
-       "         [-2.1110e+02, -2.1140e+02],\n",
-       "         [-2.1043e+02, -2.1055e+02],\n",
-       "         [-2.0790e+02, -2.0981e+02],\n",
-       "         [-2.0805e+02, -2.0777e+02],\n",
-       "         [-2.0844e+02, -2.0721e+02],\n",
-       "         [-2.0603e+02, -2.0695e+02],\n",
-       "         [-2.0277e+02, -2.0569e+02],\n",
-       "         [-1.9900e+02, -2.0271e+02],\n",
-       "         [-1.9703e+02, -1.9898e+02],\n",
-       "         [-1.9753e+02, -1.9690e+02],\n",
-       "         [-1.9758e+02, -1.9647e+02],\n",
-       "         [-1.9675e+02, -1.9619e+02],\n",
-       "         [-1.9541e+02, -1.9574e+02],\n",
-       "         [-1.9552e+02, -1.9487e+02],\n",
-       "         [-1.9666e+02, -1.9446e+02],\n",
-       "         [-1.9090e+02, -1.9553e+02],\n",
-       "         [-1.8694e+02, -1.9095e+02],\n",
-       "         [-1.9004e+02, -1.8693e+02],\n",
-       "         [-1.8831e+02, -1.8688e+02],\n",
-       "         [-1.8653e+02, -1.8667e+02],\n",
-       "         [-1.8658e+02, -1.8591e+02],\n",
-       "         [-1.8689e+02, -1.8549e+02],\n",
-       "         [-1.8719e+02, -1.8527e+02],\n",
-       "         [-1.8760e+02, -1.8514e+02],\n",
-       "         [-1.8826e+02, -1.8506e+02],\n",
-       "         [-1.8857e+02, -1.8502e+02],\n",
-       "         [-1.8863e+02, -1.8499e+02],\n",
-       "         [-1.8759e+02, -1.8496e+02],\n",
-       "         [-1.8709e+02, -1.8489e+02],\n",
-       "         [-1.8714e+02, -1.8479e+02],\n",
-       "         [-1.8638e+02, -1.8469e+02],\n",
-       "         [-1.8282e+02, -1.8452e+02],\n",
-       "         [-1.8101e+02, -1.8265e+02],\n",
-       "         [-1.8089e+02, -1.8084e+02],\n",
-       "         [-1.8437e+02, -1.8017e+02],\n",
-       "         [-1.8485e+02, -1.8016e+02],\n",
-       "         [-1.8380e+02, -1.8015e+02],\n",
-       "         [-1.8273e+02, -1.8012e+02],\n",
-       "         [-1.8204e+02, -1.8005e+02],\n",
-       "         [-1.8135e+02, -1.7992e+02],\n",
-       "         [-1.8096e+02, -1.7971e+02],\n",
-       "         [-1.8007e+02, -1.7947e+02],\n",
-       "         [-1.8093e+02, -1.7904e+02],\n",
-       "         [-1.8179e+02, -1.7891e+02],\n",
-       "         [-1.7581e+02, -1.8068e+02],\n",
-       "         [-1.7191e+02, -1.7584e+02],\n",
-       "         [-1.7005e+02, -1.7195e+02],\n",
-       "         [-1.6664e+02, -1.6997e+02],\n",
-       "         [-1.5918e+02, -1.6662e+02],\n",
-       "         [-1.5845e+02, -1.5919e+02],\n",
-       "         [-1.5920e+02, -1.5808e+02],\n",
-       "         [-1.5987e+02, -1.5787e+02],\n",
-       "         [-1.6164e+02, -1.5777e+02],\n",
-       "         [-1.6259e+02, -1.5776e+02],\n",
-       "         [-1.5753e+02, -1.6083e+02],\n",
-       "         [-1.5305e+02, -1.5760e+02],\n",
-       "         [-1.5195e+02, -1.5419e+02],\n",
-       "         [-1.5250e+02, -1.5235e+02],\n",
-       "         [-1.4619e+02, -1.5538e+02],\n",
-       "         [-1.4701e+02, -1.4619e+02],\n",
-       "         [-1.4396e+02, -1.4594e+02],\n",
-       "         [-1.4333e+02, -1.4386e+02],\n",
-       "         [-1.4368e+02, -1.4288e+02],\n",
-       "         [-1.3974e+02, -1.4251e+02],\n",
-       "         [-1.3701e+02, -1.3969e+02],\n",
-       "         [-1.3037e+02, -1.4083e+02],\n",
-       "         [-1.2646e+02, -1.3052e+02],\n",
-       "         [-1.2280e+02, -1.2654e+02],\n",
-       "         [-1.2127e+02, -1.2277e+02],\n",
-       "         [-1.1802e+02, -1.2107e+02],\n",
-       "         [-1.0934e+02, -1.2088e+02],\n",
-       "         [-1.0717e+02, -1.0934e+02],\n",
-       "         [-1.0696e+02, -1.0706e+02],\n",
-       "         [-1.0192e+02, -1.0632e+02],\n",
-       "         [-9.6160e+01, -1.0192e+02],\n",
-       "         [-9.1622e+01, -9.6166e+01],\n",
-       "         [-9.0682e+01, -9.1624e+01],\n",
-       "         [-9.0639e+01, -9.0360e+01],\n",
-       "         [-9.3666e+01, -8.9798e+01],\n",
-       "         [-8.5321e+01, -9.6196e+01],\n",
-       "         [-8.7987e+01, -8.5552e+01],\n",
-       "         [-8.8252e+01, -8.6939e+01],\n",
-       "         [-8.5141e+01, -8.7025e+01],\n",
-       "         [-6.2668e+01, -8.9043e+01],\n",
-       "         [-5.8723e+01, -6.2668e+01],\n",
-       "         [-5.4701e+01, -5.8703e+01],\n",
-       "         [-5.2488e+01, -5.4683e+01],\n",
-       "         [-4.8607e+01, -5.2382e+01],\n",
-       "         [-4.4409e+01, -4.8588e+01],\n",
-       "         [-3.6444e+01, -5.1356e+01],\n",
-       "         [-3.5016e+01, -3.6445e+01],\n",
-       "         [-3.5325e+01, -3.4801e+01],\n",
-       "         [-3.4360e+01, -3.4336e+01],\n",
-       "         [-3.2938e+01, -3.3655e+01],\n",
-       "         [-3.0829e+01, -3.2540e+01],\n",
-       "         [-2.9628e+01, -3.0663e+01],\n",
-       "         [-2.9922e+01, -2.9324e+01],\n",
-       "         [-3.0902e+01, -2.8886e+01],\n",
-       "         [-3.1163e+01, -2.8761e+01],\n",
-       "         [-2.7488e+01, -2.8674e+01],\n",
-       "         [-2.6982e+01, -2.7221e+01],\n",
-       "         [-3.0372e+01, -2.6402e+01],\n",
-       "         [-2.9038e+01, -2.6383e+01],\n",
-       "         [-2.8164e+01, -2.6316e+01],\n",
-       "         [-2.7254e+01, -2.6178e+01],\n",
-       "         [-2.4680e+01, -2.6031e+01],\n",
-       "         [-2.5801e+01, -2.4487e+01],\n",
-       "         [-2.6773e+01, -2.4260e+01],\n",
-       "         [-2.5797e+01, -2.4190e+01],\n",
-       "         [-2.6708e+01, -2.4018e+01],\n",
-       "         [-2.7858e+01, -2.3958e+01],\n",
-       "         [-2.2139e+01, -2.5121e+01],\n",
-       "         [-2.8328e+01, -2.2100e+01],\n",
-       "         [-2.7177e+01, -2.2113e+01],\n",
-       "         [-2.5953e+01, -2.2147e+01],\n",
-       "         [-2.5987e+01, -2.2175e+01],\n",
-       "         [-2.6544e+01, -2.2229e+01],\n",
-       "         [-2.7662e+01, -2.2247e+01],\n",
-       "         [-2.9118e+01, -2.2250e+01],\n",
-       "         [-2.5194e+01, -2.2554e+01],\n",
-       "         [-2.5209e+01, -2.2896e+01],\n",
-       "         [-3.0700e+01, -2.2829e+01],\n",
-       "         [-2.9220e+01, -2.2893e+01],\n",
-       "         [-2.9130e+01, -2.2960e+01],\n",
-       "         [-2.9412e+01, -2.2982e+01],\n",
-       "         [-2.9311e+01, -2.2987e+01],\n",
-       "         [-2.9602e+01, -2.2988e+01],\n",
-       "         [-3.0431e+01, -2.2988e+01],\n",
-       "         [-3.1378e+01, -2.2987e+01],\n",
-       "         [-3.1774e+01, -2.2987e+01],\n",
-       "         [-3.2103e+01, -2.2987e+01],\n",
-       "         [-3.2398e+01, -2.2987e+01],\n",
-       "         [-3.2698e+01, -2.2987e+01],\n",
-       "         [-3.2968e+01, -2.2987e+01],\n",
-       "         [-3.3121e+01, -2.2987e+01],\n",
-       "         [-3.3213e+01, -2.2987e+01],\n",
-       "         [-3.3279e+01, -2.2987e+01],\n",
-       "         [-3.3436e+01, -2.2987e+01],\n",
-       "         [-3.3520e+01, -2.2987e+01],\n",
-       "         [-3.3524e+01, -2.2987e+01],\n",
-       "         [-3.3570e+01, -2.2987e+01],\n",
-       "         [-3.3440e+01, -2.2987e+01],\n",
-       "         [-3.3221e+01, -2.2987e+01],\n",
-       "         [-3.0416e+01, -2.2987e+01],\n",
-       "         [-2.9359e+01, -2.2986e+01],\n",
-       "         [-2.9178e+01, -2.2985e+01],\n",
-       "         [-2.9116e+01, -2.2983e+01],\n",
-       "         [-2.9091e+01, -2.2981e+01],\n",
-       "         [-2.9035e+01, -2.2979e+01],\n",
-       "         [-2.8889e+01, -2.2977e+01]]), tensor([-9.7630, -9.7630, -9.7630,  ..., -9.7630, -9.7630, -9.7630]), 0, 0)}))"
+       "       start       end  confidence           words\n",
+       "0   0.260154  0.661184   -0.107317          klikid\n",
+       "1   0.661184  0.821596   -0.001331            neid\n",
+       "2   0.822883  1.784067   -0.002136        allserva\n",
+       "3   1.784067  1.984582   -0.041078         tekivad\n",
+       "4   2.547310  3.067362   -0.008251            need\n",
+       "5   3.067362  4.029833   -0.007814   lubaküpsiseid\n",
+       "6   4.752973  5.273025   -0.000333             mis\n",
+       "7   5.273025  5.413385   -0.063720              on\n",
+       "8   5.413385  5.553745   -0.000231            nagu\n",
+       "9   5.553745  5.834466   -0.000573          ilusti\n",
+       "10  5.834466  6.115187   -0.001930          kohati\n",
+       "11  6.236783  7.037555   -0.004937        tõlgitud\n",
+       "12  7.037555  7.238070   -0.001092           eesti\n",
+       "13  7.238070  7.679202   -0.003088          keelde\n",
+       "14  7.800798  8.320850   -0.001126             see\n",
+       "15  8.320850  8.601571   -0.033408            idee\n",
+       "16  8.601571  9.363527   -0.032846  arusaadavamaks\n",
+       "17  9.363527  9.584093   -0.390966              ma\n",
+       "18  9.584093  9.764557   -0.053868            tean\n",
+       "19  9.764557  9.924968   -0.000163              et"
       ]
      },
-     "execution_count": 10,
+     "execution_count": 109,
      "metadata": {},
      "output_type": "execute_result"
     }
    ],
    "source": [
-    "text"
+    "df_times.head(20)"
    ]
   },
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e03dcce0-e322-465f-b1d1-c266dec80b6b",
+   "id": "7a4be2b1-5e0f-4558-8097-b37be0b83785",
    "metadata": {},
    "outputs": [],
    "source": []