{"id":571,"date":"2021-05-08T21:17:37","date_gmt":"2021-05-08T12:17:37","guid":{"rendered":"https:\/\/staka.jp\/wordpress\/?p=571"},"modified":"2021-05-08T21:17:37","modified_gmt":"2021-05-08T12:17:37","slug":"%e3%83%8b%e3%83%a5%e3%83%bc%e3%83%a9%e3%83%ab%e6%a9%9f%e6%a2%b0%e7%bf%bb%e8%a8%b3%e3%83%a2%e3%83%87%e3%83%ab%e3%81%a8%e5%af%be%e8%a8%b3%e3%83%87%e3%83%bc%e3%82%bf%e3%81%ae%e5%93%81%e8%b3%aa","status":"publish","type":"post","link":"https:\/\/staka.jp\/wordpress\/?p=571","title":{"rendered":"\u30cb\u30e5\u30fc\u30e9\u30eb\u6a5f\u68b0\u7ffb\u8a33\u30e2\u30c7\u30eb\u3068\u5bfe\u8a33\u30c7\u30fc\u30bf\u306e\u54c1\u8cea"},"content":{"rendered":"\n<p>\u30cb\u30e5\u30fc\u30e9\u30eb\u6a5f\u68b0\u7ffb\u8a33\u30e2\u30c7\u30eb\u3084DeepLearning\u306b\u9650\u3089\u305a\u3060\u304c\u30c7\u30fc\u30bf\u54c1\u8cea\u3068\u30c7\u30fc\u30bf\u91cf\u306e\u95a2\u4fc2\u306f\u826f\u304f\u8a71\u984c\u306b\u306a\u308b\u3002\u54c1\u8cea\u306e\u826f\u3044\u5c11\u91cf\u306e\u30c7\u30fc\u30bf\u3067\u30e2\u30c7\u30eb\u69cb\u7bc9\u3092\u884c\u3046\u3079\u304d\u306a\u306e\u304b\u3001\u591a\u5c11\u54c1\u8cea\u304c\u60aa\u304f\u3068\u3082\u5927\u91cf\u306e\u30c7\u30fc\u30bf\u3092\u7528\u3044\u308b\u3079\u304d\u306a\u306e\u304b\u3068\u3044\u3046\u8b70\u8ad6\u3067\u3042\u308b\u3002<\/p>\n\n\n\n<p>\u672c\u8a18\u4e8b\u3067\u306f\u30cb\u30e5\u30fc\u30e9\u30eb\u6a5f\u68b0\u7ffb\u8a33\u30e2\u30c7\u30eb\uff08Marin-NMT\u306b\u3088\u308bTransformer\uff09\u3092\u5bfe\u8c61\u3068\u3057\u3066\u30c7\u30fc\u30bf\u54c1\u8cea\u3092\u5411\u4e0a\u3055\u305b\u305f\u7d041000\u4e07\u5bfe\u8a33\u30da\u30a2\u306e\u30c7\u30fc\u30bf\u3068\u54c1\u8cea\u5411\u4e0a\u524d\uff08\u6700\u4f4e\u9650\u306e\u30af\u30ec\u30f3\u30b8\u30f3\u30b0\u3092\u5b9f\u65bd\uff09\u306e\u7d041400\u4e07\u5bfe\u8a33\u30da\u30a2\u306e\u30c7\u30fc\u30bf\u3067\u69cb\u7bc9\u3057\u305f\u30e2\u30c7\u30eb[1]\u3092\u6bd4\u8f03\u3057\u305f\u3002<\/p>\n\n\n\n<p>\u305d\u306e\u7d50\u679c\u3001\u54c1\u8cea\u5411\u4e0a\u5f8c\u30c7\u30fc\u30bf\u3067\u306fBLEU=28.6\u3001\u54c1\u8cea\u5411\u4e0a\u524d\u30c7\u30fc\u30bf\u3067\u306fBLEU=27.5\u3068BLEU\u3067\u7d041.1point\u306e\u54c1\u8cea\u5411\u4e0a\u306e\u52b9\u679c\u304c\u793a\u3055\u308c\u305f\u3002\u3053\u308c\u306f\u4e71\u6570\u3067\u51fa\u308b\u3088\u3046\u306a\u5f71\u97ff\u3067\u306f\u306a\u304f\u672c\u4ef6\u3067\u5b9f\u65bd\u3057\u305f\u54c1\u8cea\u5411\u4e0a\u51e6\u7406\u306b\u306f\u610f\u5473\u304c\u3042\u308b\u3082\u306e\u3068\u601d\u308f\u308c\u308b\u3002<\/p>\n\n\n\n<p>\u4ee5\u4e0b\u3001\u8a73\u7d30\u3092\u8a18\u8f09\u3059\u308b\u3002<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">\u57fa\u672c\u7684\u306a\u30af\u30ea\u30fc\u30cb\u30f3\u30b0\u30ed\u30b8\u30c3\u30af<\/h3>\n\n\n\n<p><a href=\"https:\/\/github.com\/s-taka\/fugumt\">FuguMT<\/a>\u3067\u306f\u69d8\u3005\u306a\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\uff08\u65e5\u82f1\u306e\u5bfe\u8a33\u30c7\u30fc\u30bf\uff09\u3092\u7528\u3044\u3066\u30e2\u30c7\u30eb\u69cb\u7bc9\u3092\u884c\u3063\u3066\u3044\u308b\u3002\u7528\u3044\u3066\u3044\u308b\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u306e\u54c1\u8cea\u306b\u306f\u5927\u304d\u306a\u5dee\u304c\u3042\u308b\u305f\u3081\u57fa\u672c\u7684\u306b\u4e0b\u8a18\u306e\u30c7\u30fc\u30bf\u30af\u30ea\u30fc\u30cb\u30f3\u30b0\u3092\u884c\u3063\u3066\u3044\u308b[1]\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li>\u65e5\u672c\u8a9e\u6587\u3001\u82f1\u8a9e\u6587\u305d\u308c\u305e\u308c\u3092Universal Sentence Encoder[1]\u3067\u5206\u6563\u8868\u73fe\u306b\u5909\u63db\u3001\u30b3\u30b5\u30a4\u30f3\u8ddd\u96e2\u304c\u4e00\u5b9a\u4ee5\u5185\u306e\u3082\u306e\u3092\u63a1\u7528<\/li><li>\u65e5\u672c\u8a9e\u6587\u3092MeCab[2]\u3067\u5f62\u614b\u7d20\u306b\u5206\u5272\u3001\u65e5\u672c\u8a9e\u6587\u306e\u5f62\u614b\u7d20\u306e\u6570\u3068\u82f1\u5358\u8a9e\u306e\u6570\u306e\u6bd4\u304c\u4e00\u5b9a\u7bc4\u56f2\u5185\u306e\u3082\u306e\u3092\u63a1\u7528<\/li><li>\u65e5\u672c\u8a9e\u6587\u304c\u65e5\u672c\u8a9e\u6587\u5b57\u30bb\u30c3\u30c8\u306e\u6587\u5b57\u3067\u69cb\u6210\u3055\u308c\u3066\u3044\u308b\u304b\u3092\u78ba\u8a8d\uff08Unicode\u306eHiragana letters\u3092\u542b\u3080\u304b\u306a\u3069\u4e00\u5b9a\u306e\u30eb\u30fc\u30eb\u3067\u5bfe\u5fdc\uff09\u3001\u82f1\u8a9e\u6587\u3082\u540c\u69d8\u306b\u69cb\u6210\u6587\u5b57\u304c\u59a5\u5f53\u304b\u3092\u78ba\u8a8d\u3002<\/li><li>\u5bfe\u8a33\u30da\u30a2\u304c\u6570\u5b57\u3092\u542b\u3080\u5834\u5408\u3001\u53cc\u65b9\u306b\u540c\u3058\u6570\u5b57\u304c\u3042\u308b\u304b\u3092\u78ba\u8a8d\u3002\u7121\u3044\u5834\u5408\u306f\u300c\u6570\u5b57\u3092\u8868\u3059\u5358\u8a9e\uff08\u4f8b\uff1aone, two, &#8230;\uff09\u300d\u300c\u6708\u3092\u8868\u3059\u5358\u8a9e\uff08\u4f8b\uff1aJan\uff5eDec\uff09\u306a\u3069\u6570\u5b57\u306b\u5909\u63db\u53ef\u80fd\u306a\u5358\u8a9e\u3092\u542b\u3093\u3067\u3044\u308b\u304b\u300d\u3068\u300c0\u306e\u6570\u3092\u5909\u5316\u3055\u305b\u308b\u8868\u73fe\uff08\u4f8b\uff1ahundred\u3084million\u306a\u3069\uff09\u300d\u304c\u3042\u308b\u304b\u3092\u78ba\u8a8d\u3001\u52a0\u3048\u3066\u3001\u305d\u308c\u304c\u6b63\u3057\u3044\u5bfe\u5fdc\u306b\u306a\u3063\u3066\u3044\u308b\u304b\u3092\u78ba\u8a8d[5]<\/li><\/ul>\n\n\n\n<p>\u4e0a\u8a18\u51e6\u7406\u306b\u3088\u3063\u3066\u591a\u304f\u306e\u5bfe\u8a33\u30da\u30a2\u304c\u30d5\u30a3\u30eb\u30bf\u30ea\u30f3\u30b0\u3055\u308c\u308b\u3002\u516c\u958b\u3055\u308c\u3066\u3044\u308b\u5bfe\u8a33\u30c7\u30fc\u30bf\u306b\u3088\u3063\u3066\u306f\u6709\u52b9\u306a\u5bfe\u8a33\u30da\u30a2\u304c\u534a\u5206\u7a0b\u5ea6\u306b\u306a\u308b\u3053\u3068\u3082\u3042\u308b\u3002<\/p>\n\n\n\n<p><a href=\"https:\/\/ai.facebook.com\/blog\/wikimatrix\/\">wikimatrix<\/a>\u306a\u3069\u6a5f\u68b0\u7684\u306b\u5bfe\u8a33\u30a2\u30e9\u30a4\u30e1\u30f3\u30c8\u3092\u884c\u3063\u305f\u30c7\u30fc\u30bf\u306b\u5bfe\u3057\u3066\u306f\u6700\u5f8c\u306e\u6570\u5b57\u5bfe\u5fdc\u306b\u3088\u308b\u30d5\u30a3\u30eb\u30bf\u30ea\u30f3\u30b0\u304c\u7279\u306b\u6709\u52b9\u3067\u3042\u308b\u3002\uff08\u9006\u306b\u8a00\u3046\u3068\u3053\u306e\u3088\u3046\u306a\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u306f\u6570\u5b57\u304c\u5bfe\u5fdc\u3057\u3066\u3044\u306a\u3044\u5bfe\u8a33\u30da\u30a2\u3092\u591a\u304f\u542b\u3093\u3067\u3044\u308b\u3002\uff09\u6a5f\u68b0\u7ffb\u8a33\u30e2\u30c7\u30eb\u3067\u3082\u6570\u5024\u5bfe\u5fdc\u306f\u8ab2\u984c\u306b\u306a\u308b\u3053\u3068\u304c\u591a\u3044\u3002\u6570\u5024\u3092\u542b\u3081\u3066\u306e\u81ea\u52d5\u5bfe\u8a33\u30a2\u30e9\u30a4\u30e1\u30f3\u30c8\u306f\u7c21\u5358\u3067\u306f\u306a\u3044\u306e\u3067\u3042\u308d\u3046\u3068\u601d\u3046\u3002\u4e00\u65b9\u3067\u6570\u5024\u3092\u5bfe\u8c61\u3068\u3057\u305f\u30eb\u30fc\u30eb\u30d9\u30fc\u30b9\u306b\u3088\u308b\u30d5\u30a3\u30eb\u30bf\u30ea\u30f3\u30b0\u306f\u305d\u308c\u307b\u3069\u56f0\u96e3\u3067\u306f\u306a\u3044\u3002\u3053\u306e\u3088\u3046\u306a\u30c7\u30fc\u30bf\u3092\u7528\u3044\u308b\u5834\u5408\u306f\u305c\u3072\u884c\u3063\u305f\u307b\u3046\u304c\u826f\u3044\u51e6\u7406\u3067\u3042\u308b\u3002<\/p>\n\n\n\n<p>\u73fe\u5728\u516c\u958b\u3055\u308c\u3066\u3044\u308b<a href=\"https:\/\/github.com\/s-taka\/fugumt\">FuguMT\u30e2\u30c7\u30eb<\/a>\u306b<a href=\"http:\/\/www.statmt.org\/cc-aligned\/\">CC Aligned\u30c7\u30fc\u30bf<\/a>\u3092\u52a0\u3048\u308b\u969b\u3082\u540c\u69d8\u306e\u51e6\u7406\u3092\u884c\u3063\u305f\u3002\u4eca\u56de\u3001\u3055\u3089\u306b\u53b3\u3057\u3044\u57fa\u6e96\u3092\u7528\u3044\u305f\u5834\u5408\u306b\u3069\u306e\u3088\u3046\u306a\u5909\u5316\u304c\u3042\u3063\u305f\u304b\u3092\u30e1\u30e2\u304c\u3066\u3089\u8a18\u8f09\u3059\u308b\u3002<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">CCAligned\u30c7\u30fc\u30bf\u8ffd\u52a0\u6642\u306b\u884c\u3063\u305f\u51e6\u7406<\/h3>\n\n\n\n<p>CCAligned\u30c7\u30fc\u30bf\u306e\u8ffd\u52a0\u306b\u3042\u305f\u308a\u3001\u57fa\u672c\u7684\u306a\u30af\u30ea\u30fc\u30cb\u30f3\u30b0\u30ed\u30b8\u30c3\u30af\u306b\u52a0\u3048\u30663\u3064\u306e\u51e6\u7406\u3092\u8ffd\u52a0\u3057\u305f\u3002\u3053\u308c\u306fCC Aligned\u306e\u30c7\u30fc\u30bf\u3092\u76ee\u691c\u3057\u305f\u3068\u3053\u308dSPAM\u7684\u306a\u30c7\u30fc\u30bf[6]\u304c\u591a\u304f\u542b\u307e\u308c\u3066\u304a\u308a\u54c1\u8cea\u304c\u3042\u307e\u308a\u826f\u304f\u306a\u3063\u305f\u305f\u3081\u3067\u3042\u308b\u3002\u6b8b\u5ff5\u306a\u304c\u3089\u524d\u8ff0\u306e\u30af\u30ea\u30fc\u30cb\u30f3\u30b0\u30ed\u30b8\u30c3\u30af\u3067\u306f\u54c1\u8cea\u304c\u4f4e\u3044\u6a5f\u68b0\u7ffb\u8a33\u3092\u30d5\u30a3\u30eb\u30bf\u30ea\u30f3\u30b0\u3059\u308b\u3053\u3068\u304c\u96e3\u3057\u304b\u3063\u305f\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li>LaBSE[7]\u3092\u7528\u3044\u3066\u30d9\u30af\u30c8\u30eb\u5316\u3057\u305f\u5bfe\u8a33\u6587\u30da\u30a2\u306e\u30b3\u30b5\u30a4\u30f3\u8ddd\u96e2\u3092\u8a08\u7b97\u3057\u3001\u4e00\u5b9a\u7bc4\u56f2\u306b\u5165\u3063\u3066\u3044\u308b\u304b<\/li><li>fasttext\u306eLanguage identification[8]\u306e\u7d50\u679c\u304c\u305d\u308c\u305e\u308c\u306e\u8a00\u8a9e\uff08\u65e5\u672c\u8a9e\u3001\u82f1\u8a9e\uff09\u3067\u5224\u5b9a\u3055\u308c\u3066\u3044\u308b\u304b<\/li><li>\u8a18\u8f09\u5185\u5bb9\uff0b\u53d6\u5f97\u5143URL\u306b\u5bfe\u3057\u3066SPAM\u7684\u306a\u30c7\u30fc\u30bf\u3067\u306f\u306a\u3044\u304b\u30922\u5024\u5206\u985e\u30e2\u30c7\u30eb\u3067\u5224\u65ad<\/li><\/ul>\n\n\n\n<p>\u6700\u5f8c\u306e\u51e6\u7406\u306f\u4e00\u5b9a\u6570\u306e\u30c7\u30fc\u30bf\u3092\u30cf\u30f3\u30c9\u30e9\u30d9\u30ea\u30f3\u30b0\u3057\u305f\u306e\u3061\u306bSPAM\u5224\u5b9a\u30e2\u30c7\u30eb\u3092\u69cb\u7bc9\u3057\u5b9f\u65bd\u3057\u305f[9]\u3002SPAM\u5224\u5b9a\u3067\u306fURL\u60c5\u5831\u304c\u975e\u5e38\u306b\u6709\u52b9\u3067\u3042\u3063\u305f\u3002\uff08\u5f53\u305f\u308a\u524d\u3067\u306f\u3042\u308b\u304c\uff09URL\u306b\u3088\u308d\u3057\u304f\u306a\u3044\u5358\u8a9e\u3092\u542b\u3093\u3067\u3044\u308b\u5834\u5408\u306f\u307b\u307cSPAM\u30b5\u30a4\u30c8\u3067\u3042\u3063\u305f\u3002<\/p>\n\n\n\n<p>\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u306bURL\u60c5\u5831\u304c\u306a\u3044\u306a\u3069\u9069\u7528\u4e0d\u53ef\u80fd\u306a\u5834\u5408\u3092\u9664\u304d\u3001\u65e2\u5b58\u30c7\u30fc\u30bf\u306b\u3082\u8ffd\u52a0\u30af\u30ea\u30fc\u30cb\u30f3\u30b0\u3092\u5b9f\u65bd\u3057\u305f\u3002<\/p>\n\n\n\n<p>\u54c1\u8cea\u5411\u4e0a\u524d\u306f\u5168\u90e8\u306714,023,805\u5bfe\u8a33\u30da\u30a2\u306e\u30c7\u30fc\u30bf\u6570\u3067\u3042\u3063\u305f\u304c\u3001\u30af\u30ea\u30fc\u30cb\u30f3\u30b0\u306e\u7d50\u679c10,903,035\u5bfe\u8a33\u30da\u30a2\u306b\u6e1b\u5c11\u3057\u305f\u3002<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">\u30e2\u30c7\u30eb\u69cb\u7bc9\u3068\u7cbe\u5ea6\u8a55\u4fa1\u7d50\u679c<\/h3>\n\n\n\n<p>\u4eca\u307e\u3067\u3068\u540c\u69d8\u3001<a href=\"https:\/\/marian-nmt.github.io\/\">Marian-NMT (transformer)&nbsp;<\/a>+&nbsp;<a href=\"https:\/\/github.com\/google\/sentencepiece\">sentencepiece&nbsp;<\/a>\u3092\u7528\u3044\u3066\u30cb\u30e5\u30fc\u30e9\u30eb\u6a5f\u68b0\u7ffb\u8a33\u30e2\u30c7\u30eb\u306e\u69cb\u7bc9\u3092\u884c\u3063\u305f\u3002\u4f7f\u7528\u3057\u305f\u6a5f\u6750\u306fAWS p3.2xlarge \u30a4\u30f3\u30b9\u30bf\u30f3\u30b9\u3067\u3042\u308b\u3002\u5b66\u7fd2\u306fearly stopping\u6709\u308a\u3067\u884c\u3044\u3001\u6700\u5584\u306e\u5b66\u7fd2\u3092\u884c\u3063\u305f\u7d50\u679c\u3092\u7528\u3044\u3066\u8a55\u4fa1\u3092\u884c\u3063\u305f\u3002\u5b66\u7fd2\u6642\u9593\u306f\u304a\u304a\u3080\u306d\u540c\u7b49\u3067\u3042\u308a\u30c7\u30fc\u30bf\u6570\u304c\u5c11\u306a\u304f\u306a\u3063\u305f\u5f71\u97ff\u306f\u898b\u3089\u308c\u306a\u304b\u3063\u305f\u3002<\/p>\n\n\n\n<p>\u6a5f\u68b0\u7ffb\u8a33\u30e2\u30c7\u30eb\u306e\u7cbe\u5ea6\u8a55\u4fa1\u306b\u306f\u5b66\u7fd2\u30c7\u30fc\u30bf\u306b\u542b\u307e\u308c\u306a\u30444,145\u5bfe\u8a33\u30da\u30a2[10]\u3092\u7528\u3044\u305f\u3002\u5bfe\u8a33\u30da\u30a2\u306e\u534a\u6570\u306f\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u306e\u5206\u5272\u306b\u3088\u308a\u3001\u6b8b\u308a\u306e\u534a\u6570\u306f\u30c9\u30e1\u30a4\u30f3\u5916\u306e\u30c7\u30fc\u30bf\u3092\u7528\u3044\u3066\u4f5c\u6210\u3057\u3066\u3044\u308b\u3002\u8a55\u4fa1\u306e\u30e1\u30c8\u30ea\u30af\u30b9\u306b\u306fBLEU\u3092\u7528\u3044\u305f\u3002<\/p>\n\n\n\n<p>\u7d50\u679c\u3001\u54c1\u8cea\u5411\u4e0a\u5f8c\u30c7\u30fc\u30bf\u3067\u306fBLEU=28.6\u3001\u54c1\u8cea\u5411\u4e0a\u524d\u30c7\u30fc\u30bf\u3067\u306fBLEU=27.5\u3068\u7d041.1\u30dd\u30a4\u30f3\u30c8BLEU\u304c\u5411\u4e0a\u3057\u305f\u3002<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">\u6240\u611f<\/h3>\n\n\n\n<p>BLEU\u3067+1.1 point\u306f<strong>\u7d4c\u9a13\u7684\u306b<\/strong>[11]\u4e71\u6570\u306e\u63fa\u308c\u3067\u51fa\u308b\u5024\u3067\u306f\u7121\u3044\u306e\u3067\u3001\u30c7\u30fc\u30bf\u54c1\u8cea\u5411\u4e0a\u306e\u52b9\u679c\u306f\u3042\u3063\u305f\u3082\u306e\u3068\u601d\u308f\u308c\u308b\u3002<\/p>\n\n\n\n<p>\u57fa\u672c\u7684\u306a\u30af\u30ea\u30fc\u30cb\u30f3\u30b0\u30ed\u30b8\u30c3\u30af\u3067\u3082\u76f8\u5fdc\u306e\u51e6\u7406\u3092\u884c\u3063\u3066\u3044\u308b\u3068\u601d\u3063\u3066\u3044\u3066\u3001\u305d\u306e\u52b9\u679c\u304c\u3042\u308b\u4e8b\u306f\u78ba\u8a8d\u3057\u3066\u3044\u308b\u3002\u305d\u306e\u305f\u3081\u8ffd\u52a0\u51e6\u7406\u3067300\u4e07\u5bfe\u8a33\u30da\u30a2\u3092\u524a\u3063\u3066\u3082BLEU\u304c\u4e0a\u304c\u308b\u306e\u306f\u6b63\u76f4\u9a5a\u304d\u3067\u3042\u3063\u305f\u3002\u524a\u3089\u308c\u305f\u30c7\u30fc\u30bf\u306b\u3082\u6709\u7528\u305d\u3046\u306a\u3082\u306e\u306f\u3042\u308b\u306e\u3067\u30c7\u30fc\u30bf\u54c1\u8cea\u3068\u30c7\u30fc\u30bf\u91cf\u306e\u30d0\u30e9\u30f3\u30b9\u3092\u826f\u3044\u611f\u3058\u306b\u3059\u308b\u53d6\u308a\u7d44\u307f\u306f\u4eca\u5f8c\u3082\u7d9a\u3051\u308b\u5fc5\u8981\u304c\u3042\u308a\u305d\u3046\u306a\u6c17\u304c\u3057\u3066\u3044\u308b\u3002<\/p>\n\n\n\n<p>\u4f5c\u6210\u3057\u305f\u30e2\u30c7\u30eb\u306e\u54c1\u8cea\u78ba\u8a8d\u306f\u73fe\u72b6\u3067\u3082\u7d9a\u3051\u3066\u3044\u308b\u3002\u73fe\u5b9f\u306e\u30c7\u30fc\u30bf\u3067\u3042\u308b\u7a0b\u5ea6\u6027\u80fd\u78ba\u8a8d\u304c\u3067\u304d\u305f\u3068\u3053\u308d\u3067\u30e2\u30c7\u30eb\u3092\u516c\u958b\u3057\u3088\u3046\u3068\u601d\u3063\u3066\u3044\u308b\u3002\u73fe\u72b6\u3067\u306f\u524d\u56de\u516c\u958b\u3057\u305f\u30e2\u30c7\u30eb\u3088\u308a\u3082\u82e5\u5e72\u6027\u80fd\u304c\u5411\u4e0a\u3057\u3066\u3044\u308b\u3088\u3046\u306b\u601d\u3046\u3002\u30c7\u30fc\u30bf\u91cf\u3092660\u4e07\u30da\u30a2\u304b\u30891000\u4e07\u5bfe\u8a33\u30da\u30a2\u306b\u62e1\u5f35\u3057\u305f\u52b9\u679c\u306f\u51fa\u3066\u3044\u308b\u3088\u3046\u3067\u3042\u308b\u3002<\/p>\n\n\n\n<h4 class=\"wp-block-heading\">\u811a\u6ce8<\/h4>\n\n\n\n<p>[1] \u30e2\u30c7\u30eb\u69cb\u7bc9\u624b\u6cd5\u306f<a href=\"https:\/\/staka.jp\/wordpress\/?p=276\">\u3053\u308c\u307e\u3067\u306e\u3082\u306e<\/a>\u3068\u540c\u3058\u3067Marian-NMT + sentencepiece\u3092\u7528\u3044\u3066\u3044\u308b\u3002\u6bd4\u8f03\u306f\u305d\u308c\u305e\u308c\u306e\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u306b\u542b\u307e\u308c\u306a\u30444,145\u6587\u3067\u884c\u3063\u305f\u3002<br \/>[2] \u554f\u984c\u306e\u3042\u308b\u5bfe\u8a33\u30da\u30a2\u304c\u5c11\u306a\u3044<a href=\"https:\/\/tatoeba.org\/jpn\">tatoeba<\/a>\u306e\u307f\u4f8b\u5916\u7684\u306b\u30af\u30ea\u30fc\u30cb\u30f3\u30b0\u306f\u884c\u3063\u3066\u3044\u306a\u3044\u3002\u305d\u308c\u4ee5\u5916\u306e\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u306b\u3064\u3044\u3066\u306f\u5bfe\u8a33\u30c7\u30fc\u30bf\u3068\u3055\u308c\u3066\u3044\u3066\u3082\u30af\u30ea\u30fc\u30cb\u30f3\u30b0\u306e\u5bfe\u8c61\u3068\u3057\u3066\u3044\u308b\u3002\uff08\u5b9f\u969b\u554f\u984c\u306e\u3042\u308b\u5bfe\u8a33\u30da\u30a2\u3092\u542b\u3093\u3067\u3044\u308b\uff09<br \/>[3] <a href=\"https:\/\/tfhub.dev\/google\/universal-sentence-encoder\/4\">https:\/\/tfhub.dev\/google\/universal-sentence-encoder\/4<\/a><br \/>[4] <a href=\"https:\/\/taku910.github.io\/mecab\/\">https:\/\/taku910.github.io\/mecab\/<\/a><br \/>[5] \u57fa\u672c\u7684\u306b\u30eb\u30fc\u30eb\u3067\u5bfe\u5fdc\u3057\u3066\u3044\u308b\u3002<br \/>[6] \u3053\u3053\u3067\u306f\u54c1\u8cea\u306e\u4f4e\u3044\u6a5f\u68b0\u7ffb\u8a33\u30e2\u30c7\u30eb\u306b\u3088\u3063\u3066\u6a5f\u68b0\u7684\u4f5c\u6210\u3055\u308c\u305f\u30b5\u30a4\u30c8\u306e\u3053\u3068<br \/>[7] <a href=\"https:\/\/tfhub.dev\/google\/LaBSE\/2\">https:\/\/tfhub.dev\/google\/LaBSE\/2<\/a><br \/>[8] <a href=\"https:\/\/fasttext.cc\/docs\/en\/language-identification.html\">https:\/\/fasttext.cc\/docs\/en\/language-identification.html<\/a><br \/>[9] \u30af\u30ea\u30fc\u30cb\u30f3\u30b0\u7528\u30e2\u30c7\u30eb\u3092\u4f5c\u3063\u3066\u30c7\u30fc\u30bf\u54c1\u8cea\u3092\u9ad8\u3081\u308b\u3053\u3068\u306f\u8fd1\u5e74\u306e\u5927\u898f\u6a21\u30e2\u30c7\u30eb\uff08T5\u3084GPT-3\u306a\u3069\uff09\u3067\u3082\u666e\u901a\u306b\u884c\u308f\u308c\u3066\u3044\u308b\u3002<br \/>[10] \u305d\u308c\u305e\u308c\u306e\u82f1\u8a9e\u6587\u306b\u5bfe\u3057\u3066\u3001\u4e00\u5f8b\u5c0f\u6587\u5b57\u306b\u5909\u63db\u3001\u8a18\u53f7\uff08\uff0b\u7a7a\u767d\uff09\u3092\u9664\u3044\u3066\u30ad\u30fc\u60c5\u5831\u3092\u4f5c\u6210\u3001\u5b66\u7fd2\u30c7\u30fc\u30bf\u3068\u30c6\u30b9\u30c8\u30c7\u30fc\u30bf\u3067\u30ad\u30fc\u60c5\u5831\u304c\u4e00\u81f4\u3059\u308b\u3082\u306e\u304c\u7121\u3044\u4e8b\u3092\u78ba\u8a8d\u3057\u3066\u3044\u308b\u3002<br \/>[11] \u672c\u6765\u306f\u6b63\u3057\u304f\u8a55\u4fa1\u3057\u305f\u3044\u3068\u3053\u308d\u3060\u304cAWS\u8cbb\u7528\u304c\u9ad8\u984d\u306a\u306e\u3067\u5341\u5206\u306a\u691c\u8a3c\u304c\u96e3\u3057\u3044\u3002\u3067\u304d\u308b\u3053\u3068\u306a\u3089ABCI\u3068\u304b\u4f7f\u3044\u305f\u3044\u3002<br \/><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u30cb\u30e5\u30fc\u30e9\u30eb\u6a5f\u68b0\u7ffb\u8a33\u30e2\u30c7\u30eb\u3084DeepLearning\u306b\u9650\u3089\u305a\u3060\u304c\u30c7\u30fc\u30bf\u54c1\u8cea\u3068\u30c7\u30fc\u30bf\u91cf\u306e\u95a2\u4fc2\u306f\u826f\u304f\u8a71\u984c\u306b\u306a\u308b\u3002\u54c1\u8cea\u306e\u826f\u3044\u5c11\u91cf\u306e\u30c7\u30fc\u30bf\u3067\u30e2\u30c7\u30eb\u69cb\u7bc9\u3092\u884c\u3046\u3079\u304d\u306a\u306e\u304b\u3001\u591a\u5c11\u54c1\u8cea\u304c\u60aa\u304f\u3068\u3082\u5927\u91cf\u306e\u30c7\u30fc\u30bf\u3092\u7528\u3044\u308b\u3079\u304d\u306a\u306e\u304b\u3068\u3044\u3046\u8b70\u8ad6\u3067\u3042 &hellip; <a href=\"https:\/\/staka.jp\/wordpress\/?p=571\" class=\"more-link\"><span class=\"screen-reader-text\">&#8220;\u30cb\u30e5\u30fc\u30e9\u30eb\u6a5f\u68b0\u7ffb\u8a33\u30e2\u30c7\u30eb\u3068\u5bfe\u8a33\u30c7\u30fc\u30bf\u306e\u54c1\u8cea&#8221; \u306e<\/span>\u7d9a\u304d\u3092\u8aad\u3080<\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2,11,13],"tags":[27,50,77],"class_list":["post-571","post","type-post","status-publish","format-standard","hentry","category-ai","category-11","category-13","tag-deeplearning","tag-neural-machine-translation","tag-77"],"_links":{"self":[{"href":"https:\/\/staka.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/571","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/staka.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/staka.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/staka.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/staka.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=571"}],"version-history":[{"count":0,"href":"https:\/\/staka.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/571\/revisions"}],"wp:attachment":[{"href":"https:\/\/staka.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=571"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/staka.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=571"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/staka.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=571"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}