{"id":2916,"date":"2022-06-22T15:41:00","date_gmt":"2022-06-22T06:41:00","guid":{"rendered":"https:\/\/kakaoenterprise.com\/?post_type=about&p=2916"},"modified":"2022-07-27T14:16:05","modified_gmt":"2022-07-27T05:16:05","slug":"%ec%b9%b4%ec%b9%b4%ec%98%a4%ec%97%94%ed%84%b0%ed%94%84%eb%9d%bc%ec%9d%b4%ec%a6%88-%ec%84%b8%ea%b3%84-%ec%b5%9c%eb%8c%80-%ec%9d%8c%ec%84%b1-%eb%b0%8f-%ec%8b%a0%ed%98%b8%ec%b2%98%eb%a6%ac-%ed%95%99","status":"publish","type":"about","link":"https:\/\/kakaoenterprise.com\/about\/%ec%b9%b4%ec%b9%b4%ec%98%a4%ec%97%94%ed%84%b0%ed%94%84%eb%9d%bc%ec%9d%b4%ec%a6%88-%ec%84%b8%ea%b3%84-%ec%b5%9c%eb%8c%80-%ec%9d%8c%ec%84%b1-%eb%b0%8f-%ec%8b%a0%ed%98%b8%ec%b2%98%eb%a6%ac-%ed%95%99\/","title":{"rendered":"\uc138\uacc4 \ucd5c\ub300 \uc74c\uc131 \ubc0f \uc2e0\ud638\ucc98\ub9ac \ud559\ud68c INTERSPEECH\uc11c \ub17c\ubb38 4\uac74 \ubc1c\ud45c"},"content":{"rendered":"\n

\uce74\uce74\uc624\uc5d4\ud130\ud504\ub77c\uc774\uc988\uac00 \uc138\uacc4 \ucd5c\ub300 \uc74c\uc131 \ubc0f \uc2e0\ud638\ucc98\ub9ac \ud559\ud68c INTERSPEECH\uc5d0\uc11c \ucd1d 4\uac74\uc758 \ub17c\ubb38\uc744 \ubc1c\ud45c\ud569\ub2c8\ub2e4. \ud2b9\ud788 \uc774\ubc88 INTERSPEECH 2022\ub294 9\uc6d4 18\uc77c\ubd80\ud130 22\uc77c\uae4c\uc9c0 \uc778\ucc9c \uc1a1\ub3c4\uc5d0\uc11c \uc624\ud504\ub77c\uc778 \uac1c\ucd5c\ub420 \uc608\uc815\uc73c\ub85c, \uad6d\ub0b4\uc678 \ud559\uacc4 \ubc0f AI \uac1c\ubc1c\uc790, IT \uae30\uc5c5\ub4e4\uc758 \ub9ce\uc740 \uad00\uc2ec\uacfc \uc8fc\ubaa9\uc744 \ubc1b\uace0 \uc788\uc2b5\ub2c8\ub2e4. \uce74\uce74\uc624\uc5d4\ud130\ud504\ub77c\uc774\uc988\ub294 \uc774\ubc88 INTERSPEECH 2022\uc5d0 ‘\ud50c\ub798\ud2f0\ub118’ \uc2a4\ud3f0\uc11c\ub85c \ucc38\uc5ec\ud558\uc5ec \uc74c\uc131 \ubc0f \uc790\uc5f0\uc5b4\ucc98\ub9ac \ubd84\uc57c \ucd5c\uc2e0 \uae30\uc220\uc744 \uc120\ubcf4\uc774\uace0 \uc5f0\uad6c \uc131\uacfc\ub97c \uacf5\uc720\ud558\ub294 \ub370 \ud798\uc4f8 \uc608\uc815\uc785\ub2c8\ub2e4.<\/p>\n\n\n\n

1) JETS: Jointly Training FastSpeech2 and HiFi-GAN for End to End Text to Speech<\/p>\n\n\n\n

\uae30\uc874 Neural TTS \uac1c\ubc1c \uacfc\uc815\uc5d0\uc11c\ub294 2\uac1c\uc758 \ubaa8\ub378(acoustic feature generator, neural vocoder)\uc744 \uac01\uac01 \ud559\uc2b5\ud55c \ub4a4 \ud30c\uc778\ud29c\ub2dd\uc744 \uac70\uccd0\uc57c \ud558\ub294 \ub4f1 \ubcf5\uc7a1\ud55c \uacfc\uc815\uc774 \uc218\ubc18\ub418\uc5c8\uc2b5\ub2c8\ub2e4. \ubcf8 \ub17c\ubb38\uc5d0\uc11c\ub294 \uc774\ub7ec\ud55c \ud559\uc2b5 \ud30c\uc774\ud504\ub77c\uc778\uc744 E2E-TTS \ubaa8\ub378\ub85c \ub2e8\uc21c\ud654\ud558\ub294 \ubc29\ubc95\ub860\uc744 \uc0c8\ub86d\uac8c \uc81c\uc548\ud558\uc600\uc2b5\ub2c8\ub2e4. \uc2e4\uc81c \uc774\ub97c \uc801\uc6a9\ud55c \uacb0\uacfc, \uae30\uc874 \ubc29\ubc95 \ub300\ube44 \ud5a5\uc0c1\ub41c \ud569\uc131 \ud488\uc9c8\uc744 \ud655\uc778\ud560 \uc218 \uc788\uc5c8\uc2b5\ub2c8\ub2e4.<\/p>\n\n\n\n

\uc790\uc138\ud788 \ubcf4\uae30<\/a><\/p>\n\n\n\n

2) Automatic Pronunciation Assessment using Self-Supervised Speech Representation Learning<\/p>\n\n\n\n

\ubcf8 \ub17c\ubb38\uc740 \ube44\uc6d0\uc5b4\ubbfc \uc601\uc5b4 \ud559\uc2b5\uc790\ub4e4\uc758 \uc601\uc5b4 \ubc1c\uc74c\uc744 \uc810\uc218\ud654\ud558\uc5ec \ud3c9\uac00\ud560 \uc218 \uc788\ub294 \ubc29\ubc95\uc744 \ub2e4\ub8e8\uace0 \uc788\uc2b5\ub2c8\ub2e4. wav2vec 2.0\uc774\ub098 HuBERT \uad6c\uc870 \uae30\ubc18\uc758 Self-Supervised Learning \uae30\ubc95\uc744 \uc0ac\uc6a9\ud558\uc5ec \uc601\uc5b4 \ubc1c\uc74c\uacfc \uad00\ub828\ub41c \ud6a8\uc728\uc801\uc778 \ud2b9\uc9d5\ub4e4\uc744 \ucd94\ucd9c\ud558\uace0, \uc810\uc218\ub97c \ud3c9\uac00\ud560 \uc218 \uc788\ub294 \uc54c\uace0\ub9ac\uc998\uc744 \uc0c8\ub86d\uac8c \uc81c\uc548\ud558\uc600\uc2b5\ub2c8\ub2e4.<\/p>\n\n\n\n

? \uc790\uc138\ud788 \ubcf4\uae30<\/a><\/p>\n\n\n\n

3) Generalizing RNN-Transducer to Out-Domain Audio via Sparse Self-Attention Layers<\/p>\n\n\n\n

\uc774 \ub17c\ubb38\uc5d0\uc11c\ub294 \uc2e4\uc81c \uc11c\ube44\uc2a4 \ud658\uacbd\uc5d0\uc11c \uc911\uc694\ud558\uac8c \ub2e4\ub904\uc9c0\ub294 \ub3c4\uba54\uc778 \ubd88\uc77c\uce58 \ud658\uacbd\uc5d0\uc11c\uc758 \uc131\ub2a5 \uac1c\uc120\uc744 \uc704\ud55c \ubc29\ubc95\ub860\uc744 \uc81c\uc548\ud558\uc600\uc2b5\ub2c8\ub2e4. \ucd5c\uc2e0 \uc778\ucf54\ub354 \uad6c\uc870\uc778 \ucee8\ud3ec\uba38(conformer)\ub294 \ub300\uccb4\ub85c \uc88b\uc740 \uc131\ub2a5\uc744 \ubcf4\uc774\uc9c0\ub9cc, \ub3c4\uba54\uc778\uc774 \ubd88\uc77c\uce58\ud558\ub294 \uae34 \ubc1c\ud654\uc758 \uacbd\uc6b0 \ub192\uc740 \ub51c\ub9ac\uc158(deletion) \uc5d0\ub7ec\ub85c \uc131\ub2a5\uc774 \uc800\ud558\ub418\ub294 \ubb38\uc81c\uac00 \uc788\uc2b5\ub2c8\ub2e4. \uc774\ub97c \uac1c\uc120\ud558\uae30 \uc704\ud574 sparse self-attention \ub808\uc774\uc5b4\ub97c \uc81c\uc548\ud558\uc600\uace0, \uadf8 \uacb0\uacfc \uae30\uc874 \uad6c\uc870 \ub300\ube44 \ud5a5\uc0c1\ub41c \uc778\uc2dd \uc131\ub2a5\uc744 \ud655\uc778\ud560 \uc218 \uc788\uc5c8\uc2b5\ub2c8\ub2e4.<\/p>\n\n\n\n

\uc790\uc138\ud788 \ubcf4\uae30<\/a><\/p>\n\n\n\n

4) The Emotion is Not One-hot Encoding: Learning with Grayscale Label for Emotion Recognition in Conversation<\/p>\n\n\n\n

\uc2a4\ubab0\ud1a1 \ucc57\ubd07 \uc5f0\uad6c \uacfc\uc815\uc5d0\uc11c \uc0ac\uc6a9\uc790 \ubc1c\ud654 \ub610\ub294 AI\uac00 \uc0dd\uc131\ud55c \ubc1c\ud654 \uc0c1\uc758 \uac10\uc815\uac12\uc744 \uc778\uc2dd\ud558\uac8c \ub418\uba74 \ubcf4\ub2e4 \uc0c1\ud669\uc5d0 \uc801\uc808\ud55c \ub2f5\ubcc0\uc744 \uc0dd\uc131\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc77c\ubc18\uc801\uc73c\ub85c \uac10\uc815\uc778\uc2dd\uc744 \uc704\ud574 \uc0ac\uc6a9\ub418\ub294 one-hot encoding \ubc29\uc2dd\uc740 \ubc1c\ud654\uc5d0 \ud3ec\ud568\ub41c \uc5ec\ub7ec \uac10\uc815\uac12\uc744 \uc778\uc2dd\ud558\uae30 \uc5b4\ub824\uc6cc, \uc2e4\uc81c \ubc1c\ud654 \uac10\uc815\uacfc\ub294 \ucc28\uc774\uac00 \ub0a0 \uc218 \ubc16\uc5d0 \uc5c6\ub294 \ud55c\uacc4\uac00 \uc788\uc2b5\ub2c8\ub2e4. \ubcf8 \ub17c\ubb38\uc5d0\uc11c\ub294 \uc880 \ub354 \ud604\uc2e4\uc5d0 \uac00\uae4c\uc6b4 \uac10\uc815\uac12\uc744 \ud559\uc2b5\ud558\uae30 \uc704\ud574, \uc790\ub3d9\uc73c\ub85c one-hot encoding\uc744 grayscale label(soft-label)\ub85c \ub9e4\ud551\ud558\uc5ec \ub2e4\uc591\ud55c \uac10\uc815 \ubd84\ud3ec\ub97c \ud559\uc2b5\ud560 \uc218 \uc788\ub294 \ubc29\ubc95\ub860\uc744 \uc81c\uc548\ud558\uc600\uc2b5\ub2c8\ub2e4.<\/p>\n\n\n\n

\uc790\uc138\ud788 \ubcf4\uae30<\/a><\/p>\n\n\n\n

\uce74\uce74\uc624\uc5d4\ud130\ud504\ub77c\uc774\uc988\ub294 \uace0\ub3c4\ud654\ub41c AI \uae30\uc220 \uc5f0\uad6c\uc0ac\ub840\ub97c \uc9c0\uc18d\uc801\uc73c\ub85c \uacf5\uc720\ud558\uace0, \ub354\uc6b1 \ubc1c\uc804\ub41c \uc11c\ube44\uc2a4\uc758 \ubaa8\uc2b5\uc73c\ub85c \uc120\uc21c\ud658\uc2dc\ud0ac \uc608\uc815\uc785\ub2c8\ub2e4. \ub9ce\uc740 \uad00\uc2ec\uacfc \uc131\uc6d0 \ubd80\ud0c1\ub4dc\ub9bd\ub2c8\ub2e4.<\/p>\n","protected":false},"excerpt":{"rendered":"

\uce74\uce74\uc624\uc5d4\ud130\ud504\ub77c\uc774\uc988\uac00 \uc138\uacc4 \ucd5c\ub300 \uc74c\uc131 \ubc0f \uc2e0\ud638\ucc98\ub9ac \ud559\ud68c INTERSPEECH\uc5d0\uc11c \ucd1d 4\uac74\uc758 \ub17c\ubb38\uc744 \ubc1c\ud45c\ud569\ub2c8\ub2e4. \ud2b9\ud788 \uc774\ubc88 INTERSPEECH 2022\ub294 9\uc6d4 18\uc77c\ubd80\ud130 22\uc77c\uae4c\uc9c0 \uc778\ucc9c \uc1a1\ub3c4\uc5d0\uc11c \uc624\ud504\ub77c\uc778 \uac1c\ucd5c\ub420 \uc608\uc815\uc73c\ub85c, \uad6d\ub0b4\uc678 \ud559\uacc4 \ubc0f AI \uac1c\ubc1c\uc790, IT \uae30\uc5c5\ub4e4\uc758 \ub9ce\uc740 \uad00\uc2ec\uacfc \uc8fc\ubaa9\uc744 \ubc1b\uace0 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n","protected":false},"featured_media":2897,"template":"","tags":[85,76,57,56],"acf":[],"_links":{"self":[{"href":"https:\/\/kakaoenterprise.com\/wp-json\/wp\/v2\/about\/2916"}],"collection":[{"href":"https:\/\/kakaoenterprise.com\/wp-json\/wp\/v2\/about"}],"about":[{"href":"https:\/\/kakaoenterprise.com\/wp-json\/wp\/v2\/types\/about"}],"version-history":[{"count":2,"href":"https:\/\/kakaoenterprise.com\/wp-json\/wp\/v2\/about\/2916\/revisions"}],"predecessor-version":[{"id":2927,"href":"https:\/\/kakaoenterprise.com\/wp-json\/wp\/v2\/about\/2916\/revisions\/2927"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/kakaoenterprise.com\/wp-json\/wp\/v2\/media\/2897"}],"wp:attachment":[{"href":"https:\/\/kakaoenterprise.com\/wp-json\/wp\/v2\/media?parent=2916"}],"wp:term":[{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/kakaoenterprise.com\/wp-json\/wp\/v2\/tags?post=2916"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}