{"id":1716,"date":"2023-05-23T00:00:00","date_gmt":"2023-05-23T00:00:00","guid":{"rendered":"urn:uuid:469f39c6-823d-4432-8328-c4687341082d"},"modified":"2023-05-23T00:00:00","modified_gmt":"2023-05-23T00:00:00","slug":"google-translatenoaihatonoyounixun-lian-saretaka","status":"publish","type":"post","link":"https:\/\/www.sekaiken.com\/?p=1716","title":{"rendered":"Google Translate\u306eAI\u306f\u3069\u306e\u3088\u3046\u306b\u8a13\u7df4\u3055\u308c\u305f\u304b"},"content":{"rendered":"<p>Google Translate \u306e\u97f3\u58f0\u8a8d\u8b58\u306b\u95a2\u3059\u308b\u4eca\u5e74\u306e\u8ad6\u6587\u3092\u898b\u3064\u3051\u307e\u3057\u305f\u3002<br \/>\nhttps:\/\/arxiv.org\/abs\/2303.01037<br \/>\narxiv\uff08\u30a2\u30fc\u30ab\u30a4\u30f4\u3068\u767a\u97f3\u3059\u308b\uff09\u306b\u63d0\u51fa\u3055\u308c\u305f\u30d7\u30ec\u30d7\u30ea\u30f3\u30c8\u3067\u3059\u3002<br \/>\n\u4ee5\u4e0b\u306e\uff13\u7a2e\u306e\u30c7\u30fc\u30bf\u304b\u3089300\u8a00\u8a9e\u306b\u5bfe\u3059\u308b\u97f3\u58f0\u8a8d\u8b58\u306e\u6a5f\u68b0\u5b66\u7fd2\u3092\u884c\u3063\u305f\u3068\u306e\u3053\u3068\u3067\u3059\u3002<br \/>\n1) \u97f3\u58f0\u306e\u307f\u3000\u3000YT-NTL-U\u3000youtube\u306b\u3042\u308b1200\u4e07\u6642\u9593\u306e\u97f3\u58f0\u3002300\u8a00\u8a9e\u304c\u3042\u308b<br \/>\n\u3000\u3000\u3000\u3000\u3000\u3000\u3000\u3000\u3000\u3000Pub-U \u516c\u5171\u30c7\u30fc\u30bf\u30d9\u30fc\u30b9\u306b\u3042\u308b42.9\u4e07\u6642\u9593\u300151\u8a00\u8a9e\u306e\u6f14\u8aac\u3002<br \/>\n2) \u6587\u7ae0\u306e\u307f\u3000\u3000Web-NTL web\u306b\u843d\u3061\u3066\u3044\u308b1140\u8a00\u8a9e\u306b\u308f\u305f\u308b280\u5104\u306e\u6587<br \/>\n3) \u97f3\u58f0\u3068\u6587\u7ae0\u306e\u7d44\u3000\u3000YT-SUP+ 73\u8a00\u8a9e\u30019\u4e07\u6642\u9593\u306e\u3061\u3083\u3093\u3068\u3057\u305f\u30c7\u30fc\u30bf\u3068\u3001youtube \u304b\u3089&rdquo;noisy student training&rdquo;\u3067\u751f\u6210\u3055\u308c\u305f10\u4e07\u6642\u9593\u306e\u7c73\u8a9e\u306e\u30c7\u30fc\u30bf\u3002<br \/>\n                 Pub-S  1\u4e07\u6642\u9593\u306e\u7c73\u8a9e\u3068\u8a081\u4e07\u6642\u9593\u306e102\u8a00\u8a9e\u306e\u30c7\u30fc\u30bf<br \/>\n\u8a13\u7df4\u306f\u4e0b\u8a18\u306e\u3088\u3046\u306b\u3084\u3063\u305f\u305d\u3046\u3067\u3059\u3002<br \/>\n1) BEST-RQ(Bert-based speech pre-training with random projection quantizer)\u3068\u3044\u3046\u65b9\u5f0f\u3092\u7528\u3044\u3066YT-NTL-U\u3092\u4f7f\u3063\u3066\u8a13\u7df4<br \/>\n2) MOST(Multi-Objective supervised pre-training)\u3068\u3044\u3046\u65b9\u5f0f\u3092\u7528\u3044\u3066\u8a13\u7df4<br \/>\n3) 1)2)\u3092\u4f7f\u3063\u3066\u8a13\u7df4\u3057\u305f\u30cb\u30e5\u30fc\u30e9\u30eb\u30cd\u30c3\u30c8\u30ef\u30fc\u30af\u304c\u3001\u6587\u7ae0\u3068\u97f3\u58f0\u306e\u7d44\u3092\u4f7f\u3063\u3066\u5b66\u7fd2<\/p>\n<p>1)\u306f\u30cb\u30e5\u30fc\u30e9\u30eb\u30cd\u30c3\u30c8\u30ef\u30fc\u30af\u306b\u6587\u7ae0\u7121\u3057\u3067\u97f3\u58f0\u3092\u305f\u304f\u3055\u3093\u805e\u304b\u305b\u305f\u308f\u3051\u3067\u3059\u304c\u3001\u4f3c\u305f\u3082\u306e\u306e\u7d44\u3092\u3067\u304d\u308b\u3060\u3051\u591a\u304f\u4f5c\u308b\u3088\u3046\u306a\u6761\u4ef6\u4ed8\u3051\u3092\u3057\u3066\u8a13\u7df4\u3057\u305f\u306e\u3067\u3057\u3087\u3046\uff08\u8a73\u7d30\u306f\u5225\u8ad6\u6587\u306a\u306e\u3067\u3001\u60f3\u50cf\u3067\u3059\uff09\u3002\u3053\u308c\u306b\u3088\u308a\u3001\u3044\u308d\u3044\u308d\u306a\u58f0\u3084\u901f\u5ea6\u3067\u8a71\u3055\u308c\u305f\u5358\u8a9e\u3092\u540c\u3058\u3082\u306e\u3067\u3042\u308b\u3068\u8a8d\u8b58\u3055\u305b\u305f\u306e\u3067\u3057\u3087\u3046\u3002\u3053\u3053\u3067\u4eba\u9593\u306e\u97f3\u58f0\u306e\u7279\u5fb4\u304c\u5171\u901a\u306a\u3082\u306e\u3068\u3057\u3066\u8a8d\u8b58\u3055\u308c\u307e\u3059\u3002\u4eba\u9593\u3082\u3001\u58f0\u306e\u7279\u5fb4\u306f\u3042\u3063\u3066\u3082\u8a00\u8449\u304c\u901a\u3058\u308b\u306e\u3067\u3001\u540c\u3058\u3088\u3046\u306a\u62bd\u51fa\u3092\u3084\u3063\u3066\u3044\u308b\u3068\u601d\u3044\u307e\u3059\u3002<br \/>\n2)\u306b\u306f\u7ffb\u8a33\u524d\u3068\u7ffb\u8a33\u5f8c\u306e\u7d44\u304c\u3042\u3063\u305f\u3068\u601d\u3046\u306e\u3067\u3001\u30ed\u30bc\u30c3\u30bf\u30b9\u30c8\u30fc\u30f3\u306e\u3088\u3046\u306b\u4f7f\u3063\u3066\u7ffb\u8a33\u6a5f\u3092\u8a13\u7df4\u3057\u305f\u306e\u3067\u3057\u3087\u3046\u3002\u307e\u305f\u3001\u4e00\u3064\u306e\u8a00\u8a9e\u306e\u6587\u6cd5\u3082\u5927\u91cf\u306e\u6587\u7ae0\u304b\u3089\u62bd\u51fa\u3067\u304d\u308b\u3067\u3057\u3087\u3046\u3002<br \/>\n\u305d\u3057\u30663)\u3067\u6587\u7ae0\u3068\u7d44\u306b\u3059\u308b\u3053\u3068\u306b\u3088\u3063\u3066\u3001\u6b63\u78ba\u306b\u6587\u5b57\u306b\u3059\u308b\u3053\u3068\u304c\u3067\u304d\u308b\u306e\u3067\u3057\u3087\u3046\u3002<br \/>\n\u6587\u5b57\u3068\u97f3\u58f0\u306e\u7d44\u3092\u308f\u3056\u308f\u3056\u4f5c\u3089\u306a\u3044\u3067\u3082\u5b66\u7fd2\u3067\u304d\u308b\u306e\u3067\u3001\u5b9f\u306b\u8ce2\u3044\u3067\u3059\u306d\u3002\u81ea\u52d5\u30ed\u30dc\u30c3\u30c8\u3067\u3042\u308b&rdquo;google crawler&rdquo;\u304c\u5e38\u306bweb\u304b\u3089\u60c5\u5831\u3092\u629c\u304d\u53d6\u3063\u3066\u3044\u307e\u3059\u304c\u3001\u305d\u3053\u304b\u3089\u6587\u7ae0\u30c7\u30fc\u30bf\u3084\u97f3\u58f0\u3068\u6587\u7ae0\u306e\u7d44\u3092\u5165\u624b\u3057\u305f\u306e\u3067\u3057\u3087\u3046\u3002\u5927\u91cf\u306eyoutube\u97f3\u58f0\u3082\u5bb9\u6613\u306b\u4f7f\u3048\u308bGoogle \u306a\u3089\u3067\u306f\u306e\u3084\u308a\u65b9\u3060\u3068\u601d\u3044\u307e\u3059\u3002<br \/>\n\u8a13\u7df4\u306e\u305f\u3081\u4eba\u9593\u306e\u7ffb\u8a33\u5bb6\u3082\u52d5\u54e1\u3057\u3066\u3044\u305f\u3068\u3044\u3046\u60c5\u5831\u3082\u3042\u308a\u307e\u3059\u3002\u660e\u65e5\u3082\u3046\u5c11\u3057\u7d30\u90e8\u3092\u898b\u3066\u307f\u305f\u3044\u3067\u3059\u3002<\/p>\n<p>\u82f1\u8a9e\u306f\u4e0a\u8a18\u8ad6\u6587\u304b\u3089\u3002<br \/>\nmultilingual \u30de\u308b\u30c6\u30a3\u300c\u308a\u300d\u30f3\u30ac\u308b\u3000\u591a\u8a00\u8a9e\u306e<br \/>\narchive \u300c\u30a2\u300d\u30fc\u30ab\u30a4\u30f4\u3000\u53ce\u96c6\u3057\u3066\u4fdd\u7ba1\u3057\u305f\u3082\u306e<br \/>\n\u201dWith conventional supervised training approaches, audio data needs to be manually transcribed, which is lengthy and expensive, or collected from existing transcribed sources which are hard to find for tail languages.&rdquo;  \u5f93\u6765\u306e\u6559\u5e2b\u3042\u308a\u5b66\u7fd2\u6cd5\u3067\u306f\u3001\u97f3\u58f0\u30c7\u30fc\u30bf\u306f\u624b\u3067\uff08\u4eba\u529b\u3067\uff09\u6587\u5b57\u8d77\u3053\u3057\u3059\u308b\u5fc5\u8981\u304c\u3042\u308a\u3001\u6642\u9593\u304c\u304b\u304b\u308a\u9ad8\u4fa1\u3067\u3042\u3063\u305f\u3002\u307e\u305f\u306f\u3001\u3059\u3067\u306b\u5b58\u5728\u3059\u308b\u6587\u5b57\u8d77\u3053\u3057\u6e08\u307f\u306e\u97f3\u58f0\u6e90\u304b\u3089\u96c6\u3081\u308b\u5fc5\u8981\u304c\u3042\u308a\u3001\u3042\u307e\u308a\u8a71\u3055\u308c\u3066\u3044\u306a\u3044\u8a00\u8a9e\u3067\u306f\u898b\u3064\u3051\u308b\u306e\u304c\u56f0\u96e3\u3067\u3042\u308b\u3002<br \/>\ntranscribe \u6587\u5b57\u8d77\u3053\u3057\u3059\u308b<br \/>\ntail language \u5206\u5e03\u306e\u88fe\u306e\u307b\u3046\u306e\u8a00\u8a9e\u3001\u3042\u307e\u308a\u8a71\u3055\u308c\u3066\u3044\u306a\u3044\u8a00\u8a9e\u3000tail \u306f\u3057\u3063\u307d\u3067\u3059\u304c\u3001\u30c7\u30fc\u30bf\u306e\u5206\u91ce\u3067\u306f\u300c\u88fe\uff08\u3059\u305d\uff09\u300d\u306e\u610f\u5473\u3067\u3088\u304f\u4f7f\u3044\u307e\u3059\u3002<br \/>\ncorpus \u30b3\u30fc\u30d1\u30b9\u3000\u96c6\u7a4d\u3001\u5168\u96c6 level 10<br \/>\ngeneric \u4e00\u822c\u7684\u306a generic drug \u30b8\u30a7\u30cd\u30ea\u30c3\u30af\u533b\u85ac\u54c1<br \/>\nWe explore the possibility of &hellip;\u3000\u6211\u3005\u306f\uff5e\u306e\u53ef\u80fd\u6027\u3092\u63a2\u7d22\u3059\u308b\u3002<br \/>\nexplore\u3000\u63a2\u7d22\u3059\u308b\u3001\u63a2\u691c\u3059\u308b<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Google Translate \u306e\u97f3\u58f0\u8a8d\u8b58\u306b\u95a2\u3059\u308b\u4eca\u5e74\u306e\u8ad6\u6587\u3092\u898b\u3064\u3051\u307e\u3057\u305f\u3002 https:\/\/arxiv.org\/abs\/2303.01037 arxiv\uff08\u30a2\u30fc\u30ab\u30a4\u30f4\u3068\u767a\u97f3\u3059\u308b\uff09\u306b\u63d0\u51fa\u3055\u308c\u305f\u30d7\u30ec\u30d7\u30ea\u30f3\u30c8\u3067\u3059\u3002 \u4ee5\u4e0b\u306e\uff13\u7a2e\u306e\u30c7\u30fc\u30bf\u304b\u3089300\u8a00\u8a9e\u306b\u5bfe\u3059\u308b\u97f3\u58f0\u8a8d\u8b58\u306e\u6a5f\u68b0\u5b66\u7fd2\u3092\u884c\u3063\u305f\u3068\u306e\u3053\u3068\u3067\u3059\u3002 1) \u97f3\u58f0\u306e\u307f\u3000\u3000YT-NTL-U\u3000youtube\u306b\u3042\u308b1200\u4e07\u6642\u9593\u306e\u97f3\u58f0\u3002300\u8a00\u8a9e\u304c\u3042\u308b \u3000\u3000\u3000\u3000\u3000\u3000\u3000\u3000\u3000\u3000Pub-U \u516c\u5171\u30c7\u30fc\u30bf\u30d9\u30fc\u30b9\u306b\u3042\u308b42.9\u4e07\u6642\u9593\u300151\u8a00\u8a9e\u306e\u6f14\u8aac\u3002 2) \u6587\u7ae0\u306e\u307f\u3000\u3000Web-NTL web\u306b\u843d\u3061\u3066\u3044\u308b1140\u8a00\u8a9e\u306b\u308f\u305f\u308b280\u5104\u306e\u6587 3) \u97f3\u58f0\u3068\u6587\u7ae0\u306e\u7d44\u3000\u3000YT-SUP&hellip;<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"om_disable_all_campaigns":false,"_monsterinsights_skip_tracking":false,"_monsterinsights_sitenote_active":false,"_monsterinsights_sitenote_note":"","_monsterinsights_sitenote_category":0,"footnotes":""},"categories":[60,42],"tags":[23,5],"class_list":["post-1716","post","type-post","status-publish","format-standard","hentry","category-computer","category-tech","tag-computer","tag-tech"],"aioseo_notices":[],"_links":{"self":[{"href":"https:\/\/www.sekaiken.com\/index.php?rest_route=\/wp\/v2\/posts\/1716","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.sekaiken.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.sekaiken.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.sekaiken.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.sekaiken.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=1716"}],"version-history":[{"count":0,"href":"https:\/\/www.sekaiken.com\/index.php?rest_route=\/wp\/v2\/posts\/1716\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.sekaiken.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=1716"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.sekaiken.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=1716"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.sekaiken.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=1716"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}