r9y9/Text processing (ja) for DNN TTS.ipynb

## Text processing (ja) for DNN TTS.ipynb
{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Populating the interactive namespace from numpy and matplotlib\n"
     ]
    }
   ],
   "source": [
    "%pylab inline\n",
    "import pyopenjtalk\n",
    "from nnmnkwii.io import hts\n",
    "from nnmnkwii.frontend import merlin as fe\n",
    "from os.path import join, expanduser"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### OpenJtalkが内包するmecabの出力\n",
    "\n",
    "OpenJTalkの言語処理フロントエンドを使いやすいように、昔ライブラリを作りました。\n",
    "https://github.com/r9y9/pyopenjtalk"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['こんにちは,感動詞,*,*,*,*,*,こんにちは,コンニチハ,コンニチワ,0/5,-1,-1']"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pyopenjtalk.run_frontend(\"こんにちは\")[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### OpenJTalk の言語処理フロントエンドの出力（フルコンテキストラベル w/o time info)\n",
    "\n",
    "注意：OpenJtalkのコマンドラインツールでは音素継続長も出力されるが、言語処理部分ではなくHTS_engine が出力している。つまり\n",
    "言語処理フロントエンドとは関係がない"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['xx^xx-sil+k=o/A:xx+xx+xx/B:xx-xx_xx/C:xx_xx+xx/D:xx+xx_xx/E:xx_xx!xx_xx-xx/F:xx_xx#xx_xx@xx_xx|xx_xx/G:5_5%0_xx_xx/H:xx_xx/I:xx-xx@xx+xx&xx-xx|xx+xx/J:1_5/K:1+1-5',\n",
       " 'xx^sil-k+o=N/A:-4+1+5/B:xx-xx_xx/C:09_xx+xx/D:xx+xx_xx/E:xx_xx!xx_xx-xx/F:5_5#0_xx@1_1|1_5/G:xx_xx%xx_xx_xx/H:xx_xx/I:1-5@1+1&1-1|1+5/J:xx_xx/K:1+1-5',\n",
       " 'sil^k-o+N=n/A:-4+1+5/B:xx-xx_xx/C:09_xx+xx/D:xx+xx_xx/E:xx_xx!xx_xx-xx/F:5_5#0_xx@1_1|1_5/G:xx_xx%xx_xx_xx/H:xx_xx/I:1-5@1+1&1-1|1+5/J:xx_xx/K:1+1-5',\n",
       " 'k^o-N+n=i/A:-3+2+4/B:xx-xx_xx/C:09_xx+xx/D:xx+xx_xx/E:xx_xx!xx_xx-xx/F:5_5#0_xx@1_1|1_5/G:xx_xx%xx_xx_xx/H:xx_xx/I:1-5@1+1&1-1|1+5/J:xx_xx/K:1+1-5',\n",
       " 'o^N-n+i=ch/A:-2+3+3/B:xx-xx_xx/C:09_xx+xx/D:xx+xx_xx/E:xx_xx!xx_xx-xx/F:5_5#0_xx@1_1|1_5/G:xx_xx%xx_xx_xx/H:xx_xx/I:1-5@1+1&1-1|1+5/J:xx_xx/K:1+1-5',\n",
       " 'N^n-i+ch=i/A:-2+3+3/B:xx-xx_xx/C:09_xx+xx/D:xx+xx_xx/E:xx_xx!xx_xx-xx/F:5_5#0_xx@1_1|1_5/G:xx_xx%xx_xx_xx/H:xx_xx/I:1-5@1+1&1-1|1+5/J:xx_xx/K:1+1-5',\n",
       " 'n^i-ch+i=w/A:-1+4+2/B:xx-xx_xx/C:09_xx+xx/D:xx+xx_xx/E:xx_xx!xx_xx-xx/F:5_5#0_xx@1_1|1_5/G:xx_xx%xx_xx_xx/H:xx_xx/I:1-5@1+1&1-1|1+5/J:xx_xx/K:1+1-5',\n",
       " 'i^ch-i+w=a/A:-1+4+2/B:xx-xx_xx/C:09_xx+xx/D:xx+xx_xx/E:xx_xx!xx_xx-xx/F:5_5#0_xx@1_1|1_5/G:xx_xx%xx_xx_xx/H:xx_xx/I:1-5@1+1&1-1|1+5/J:xx_xx/K:1+1-5',\n",
       " 'ch^i-w+a=sil/A:0+5+1/B:xx-xx_xx/C:09_xx+xx/D:xx+xx_xx/E:xx_xx!xx_xx-xx/F:5_5#0_xx@1_1|1_5/G:xx_xx%xx_xx_xx/H:xx_xx/I:1-5@1+1&1-1|1+5/J:xx_xx/K:1+1-5',\n",
       " 'i^w-a+sil=xx/A:0+5+1/B:xx-xx_xx/C:09_xx+xx/D:xx+xx_xx/E:xx_xx!xx_xx-xx/F:5_5#0_xx@1_1|1_5/G:xx_xx%xx_xx_xx/H:xx_xx/I:1-5@1+1&1-1|1+5/J:xx_xx/K:1+1-5',\n",
       " 'w^a-sil+xx=xx/A:xx+xx+xx/B:xx-xx_xx/C:xx_xx+xx/D:xx+xx_xx/E:5_5!0_xx-xx/F:xx_xx#xx_xx@xx_xx|xx_xx/G:xx_xx%xx_xx_xx/H:1_5/I:xx-xx@xx+xx&xx-xx|xx+xx/J:xx_xx/K:1+1-5']"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pyopenjtalk.run_frontend(\"こんにちは\")[1]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 音素単位の言語特徴量の抽出\n",
    "\n",
    "継続帳モデルを学習するために必要です。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 言語特徴量ベクトルを抽出するための questionファイルを読む\n",
    "question = join(expanduser(\"~\"), \"sp/nnmnkwii_gallery/data/questions_jp.hed\")\n",
    "binary_dict, continuous_dict = hts.load_question_set(question)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# nnmnkwii の HTS に便宜上変換（nnmnkwiiのmerlin frontendを使うため）\n",
    "labels = hts.load(lines=pyopenjtalk.run_frontend(\"音声合成は難しいですね〜。大変です\")[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# *: 音響モデルを学習するための言語特徴量を抽出する場合は、add_frame_features=True にして、フレーム単位の特徴量を抽出する必要がある\n",
    "# https://r9y9.github.io/nnmnkwii/latest/references/generated/nnmnkwii.frontend.merlin.linguistic_features.html\n",
    "# ただし、その場合は HTS labelに時間アライメントの情報が必要（Juliusを使って付与するのが簡単です）\n",
    "features = fe.linguistic_features(labels, binary_dict, continuous_dict)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(41, 531)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "assert features.shape[0] == len(labels)\n",
    "features.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "figure(figsize=(15, 4))\n",
    "imshow(features.T, aspect=\"auto\")\n",
    "xlabel(\"phone-index\")\n",
    "ylabel(\"feature index\");"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## フレーム単位の言語特徴量の抽出"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0 3125000 xx^xx-sil+m=i/A:xx+xx+xx/B:xx-xx_xx/C:xx_xx+xx/D:02+xx_xx/E:xx_xx!xx_xx-xx/F:xx_xx#xx_xx@xx_xx|xx_xx/G:3_3%0_xx_xx/H:xx_xx/I:xx-xx@xx+xx&xx-xx|xx+xx/J:5_23/K:1+5-23\n",
       "3125000 3525000 xx^sil-m+i=z/A:-2+1+3/B:xx-xx_xx/C:02_xx+xx/D:13+xx_xx/E:xx_xx!xx_xx-xx/F:3_3#0_xx@1_5|1_23/G:7_2%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "3525000 4325000 sil^m-i+z=u/A:-2+1+3/B:xx-xx_xx/C:02_xx+xx/D:13+xx_xx/E:xx_xx!xx_xx-xx/F:3_3#0_xx@1_5|1_23/G:7_2%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "4325000 5225000 m^i-z+u=o/A:-1+2+2/B:xx-xx_xx/C:02_xx+xx/D:13+xx_xx/E:xx_xx!xx_xx-xx/F:3_3#0_xx@1_5|1_23/G:7_2%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "5225000 5525000 i^z-u+o=m/A:-1+2+2/B:xx-xx_xx/C:02_xx+xx/D:13+xx_xx/E:xx_xx!xx_xx-xx/F:3_3#0_xx@1_5|1_23/G:7_2%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "5525000 6525000 z^u-o+m=a/A:0+3+1/B:02-xx_xx/C:13_xx+xx/D:18+xx_xx/E:xx_xx!xx_xx-xx/F:3_3#0_xx@1_5|1_23/G:7_2%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "6525000 7524999 u^o-m+a=r/A:-1+1+7/B:13-xx_xx/C:18_xx+xx/D:13+xx_xx/E:3_3!0_xx-1/F:7_2#0_xx@2_4|4_20/G:6_6%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "7524999 8225000 o^m-a+r=e/A:-1+1+7/B:13-xx_xx/C:18_xx+xx/D:13+xx_xx/E:3_3!0_xx-1/F:7_2#0_xx@2_4|4_20/G:6_6%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "8225000 8725000 m^a-r+e=e/A:0+2+6/B:13-xx_xx/C:18_xx+xx/D:13+xx_xx/E:3_3!0_xx-1/F:7_2#0_xx@2_4|4_20/G:6_6%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "8725000 9125000 a^r-e+e=sh/A:0+2+6/B:13-xx_xx/C:18_xx+xx/D:13+xx_xx/E:3_3!0_xx-1/F:7_2#0_xx@2_4|4_20/G:6_6%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "9125000 9725000 r^e-e+sh=i/A:1+3+5/B:13-xx_xx/C:18_xx+xx/D:13+xx_xx/E:3_3!0_xx-1/F:7_2#0_xx@2_4|4_20/G:6_6%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "9725000 10925000 e^e-sh+i=a/A:2+4+4/B:13-xx_xx/C:18_xx+xx/D:13+xx_xx/E:3_3!0_xx-1/F:7_2#0_xx@2_4|4_20/G:6_6%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "10925000 11225000 e^sh-i+a=k/A:2+4+4/B:13-xx_xx/C:18_xx+xx/D:13+xx_xx/E:3_3!0_xx-1/F:7_2#0_xx@2_4|4_20/G:6_6%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "11225000 12325000 sh^i-a+k=a/A:3+5+3/B:13-xx_xx/C:18_xx+xx/D:13+xx_xx/E:3_3!0_xx-1/F:7_2#0_xx@2_4|4_20/G:6_6%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "12325000 12925000 i^a-k+a=r/A:4+6+2/B:18-xx_xx/C:13_xx+xx/D:20+1_0/E:3_3!0_xx-1/F:7_2#0_xx@2_4|4_20/G:6_6%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "12925000 13325000 a^k-a+r=a/A:4+6+2/B:18-xx_xx/C:13_xx+xx/D:20+1_0/E:3_3!0_xx-1/F:7_2#0_xx@2_4|4_20/G:6_6%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "13325000 13825000 k^a-r+a=k/A:5+7+1/B:18-xx_xx/C:13_xx+xx/D:20+1_0/E:3_3!0_xx-1/F:7_2#0_xx@2_4|4_20/G:6_6%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "13825000 14325000 a^r-a+k=a/A:5+7+1/B:18-xx_xx/C:13_xx+xx/D:20+1_0/E:3_3!0_xx-1/F:7_2#0_xx@2_4|4_20/G:6_6%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "14325000 15325000 r^a-k+a=w/A:-5+1+6/B:13-xx_xx/C:20_1+0/D:10+7_1/E:7_2!0_xx-1/F:6_6#0_xx@3_3|11_13/G:4_2%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "15325000 15925000 a^k-a+w=a/A:-5+1+6/B:13-xx_xx/C:20_1+0/D:10+7_1/E:7_2!0_xx-1/F:6_6#0_xx@3_3|11_13/G:4_2%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "15925000 16825000 k^a-w+a=n/A:-4+2+5/B:13-xx_xx/C:20_1+0/D:10+7_1/E:7_2!0_xx-1/F:6_6#0_xx@3_3|11_13/G:4_2%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "16825000 17225000 a^w-a+n=a/A:-4+2+5/B:13-xx_xx/C:20_1+0/D:10+7_1/E:7_2!0_xx-1/F:6_6#0_xx@3_3|11_13/G:4_2%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "17225000 17725000 w^a-n+a=k/A:-3+3+4/B:20-1_0/C:10_7+1/D:12+xx_xx/E:7_2!0_xx-1/F:6_6#0_xx@3_3|11_13/G:4_2%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "17725000 18425000 a^n-a+k=U/A:-3+3+4/B:20-1_0/C:10_7+1/D:12+xx_xx/E:7_2!0_xx-1/F:6_6#0_xx@3_3|11_13/G:4_2%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "18425000 18925000 n^a-k+U=t/A:-2+4+3/B:20-1_0/C:10_7+1/D:12+xx_xx/E:7_2!0_xx-1/F:6_6#0_xx@3_3|11_13/G:4_2%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "18925000 19225000 a^k-U+t=e/A:-2+4+3/B:20-1_0/C:10_7+1/D:12+xx_xx/E:7_2!0_xx-1/F:6_6#0_xx@3_3|11_13/G:4_2%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "19225000 19725000 k^U-t+e=w/A:-1+5+2/B:10-7_1/C:12_xx+xx/D:24+xx_xx/E:7_2!0_xx-1/F:6_6#0_xx@3_3|11_13/G:4_2%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "19725000 20025000 U^t-e+w=a/A:-1+5+2/B:10-7_1/C:12_xx+xx/D:24+xx_xx/E:7_2!0_xx-1/F:6_6#0_xx@3_3|11_13/G:4_2%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "20025000 20724999 t^e-w+a=n/A:0+6+1/B:12-xx_xx/C:24_xx+xx/D:17+1_0/E:7_2!0_xx-1/F:6_6#0_xx@3_3|11_13/G:4_2%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "20724999 21125000 e^w-a+n=a/A:0+6+1/B:12-xx_xx/C:24_xx+xx/D:17+1_0/E:7_2!0_xx-1/F:6_6#0_xx@3_3|11_13/G:4_2%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "21125000 21825000 w^a-n+a=r/A:-1+1+4/B:24-xx_xx/C:17_1+0/D:10+7_2/E:6_6!0_xx-1/F:4_2#0_xx@4_2|17_7/G:3_2%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "21825000 22525000 a^n-a+r=a/A:-1+1+4/B:24-xx_xx/C:17_1+0/D:10+7_2/E:6_6!0_xx-1/F:4_2#0_xx@4_2|17_7/G:3_2%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "22525000 23025000 n^a-r+a=n/A:0+2+3/B:24-xx_xx/C:17_1+0/D:10+7_2/E:6_6!0_xx-1/F:4_2#0_xx@4_2|17_7/G:3_2%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "23025000 23424999 a^r-a+n=a/A:0+2+3/B:24-xx_xx/C:17_1+0/D:10+7_2/E:6_6!0_xx-1/F:4_2#0_xx@4_2|17_7/G:3_2%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "23424999 24225000 r^a-n+a=i/A:1+3+2/B:17-1_0/C:10_7+2/D:22+xx_xx/E:6_6!0_xx-1/F:4_2#0_xx@4_2|17_7/G:3_2%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "24225000 24725000 a^n-a+i=n/A:1+3+2/B:17-1_0/C:10_7+2/D:22+xx_xx/E:6_6!0_xx-1/F:4_2#0_xx@4_2|17_7/G:3_2%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "24725000 25025000 n^a-i+n=o/A:2+4+1/B:17-1_0/C:10_7+2/D:22+xx_xx/E:6_6!0_xx-1/F:4_2#0_xx@4_2|17_7/G:3_2%0_xx_1/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "25025000 25724999 a^i-n+o=d/A:-1+1+3/B:10-7_2/C:22_xx+xx/D:10+7_2/E:4_2!0_xx-1/F:3_2#0_xx@5_1|21_3/G:xx_xx%xx_xx_xx/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "25724999 26125000 i^n-o+d=e/A:-1+1+3/B:10-7_2/C:22_xx+xx/D:10+7_2/E:4_2!0_xx-1/F:3_2#0_xx@5_1|21_3/G:xx_xx%xx_xx_xx/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "26125000 26525000 n^o-d+e=s/A:0+2+2/B:22-xx_xx/C:10_7+2/D:xx+xx_xx/E:4_2!0_xx-1/F:3_2#0_xx@5_1|21_3/G:xx_xx%xx_xx_xx/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "26525000 27325000 o^d-e+s=U/A:0+2+2/B:22-xx_xx/C:10_7+2/D:xx+xx_xx/E:4_2!0_xx-1/F:3_2#0_xx@5_1|21_3/G:xx_xx%xx_xx_xx/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "27325000 29625000 d^e-s+U=sil/A:1+3+1/B:22-xx_xx/C:10_7+2/D:xx+xx_xx/E:4_2!0_xx-1/F:3_2#0_xx@5_1|21_3/G:xx_xx%xx_xx_xx/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "29625000 30025000 e^s-U+sil=xx/A:1+3+1/B:22-xx_xx/C:10_7+2/D:xx+xx_xx/E:4_2!0_xx-1/F:3_2#0_xx@5_1|21_3/G:xx_xx%xx_xx_xx/H:xx_xx/I:5-23@1+1&1-5|1+23/J:xx_xx/K:1+5-23\n",
       "30025000 31825000 s^U-sil+xx=xx/A:xx+xx+xx/B:10-7_2/C:xx_xx+xx/D:xx+xx_xx/E:3_2!0_xx-xx/F:xx_xx#xx_xx@xx_xx|xx_xx/G:xx_xx%xx_xx_xx/H:5_23/I:xx-xx@xx+xx&xx-xx|xx+xx/J:xx_xx/K:1+5-23"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Juliusで付与したアライメント情報ありのフルコンテキストラベル\n",
    "# フレームシフトは5msを仮定（デフォルト）\n",
    "labels = hts.load(join(expanduser(\"~\"), \"data/jsut_ver1.1/basic5000/lab/BASIC5000_0001.lab\"));\n",
    "labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "features = fe.linguistic_features(labels, binary_dict, continuous_dict,\n",
    "                                  subphone_features=\"coarse_coding\",\n",
    "                                  add_frame_features=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(632, 535)"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "features.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "figure(figsize=(15, 4))\n",
    "imshow(features.T, aspect=\"auto\")\n",
    "xlabel(\"frame index\")\n",
    "ylabel(\"feature index\");"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 参考\n",
    "\n",
    "gantts/segmentation-kitに、JSUTで日本語音声合成を作るためのコードがあります。日本語音声合成に必要なフルコンテキストラベルを用意するコードがすべて含まれています（Juliusで音素アライメントを取る部分も）。\n",
    "ただし、特にsegmentation-kitは自分用のコードで、整理された読みやすいコードではないので、ご了承ください\n",
    "\n",
    "- https://github.com/r9y9/segmentation-kit/tree/jsut2\n",
    "- https://github.com/r9y9/gantts/tree/ja\n",
    "- https://github.com/r9y9/nnmnkwii_gallery/tree/master/data 日本語用のquestion fileがあります\n",
    "- https://github.com/r9y9/nnmnkwii フルコンテキストラベル -> 数値ベクトル化のツール（Merlinのコードを拝借しました）があります"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}