{"id":11109,"date":"2022-06-26T12:00:45","date_gmt":"2022-06-26T12:00:45","guid":{"rendered":"https:\/\/www.orynycz.com\/uncategorized-sk\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\/"},"modified":"2026-01-23T06:13:52","modified_gmt":"2026-01-23T06:13:52","slug":"say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022","status":"publish","type":"post","link":"https:\/\/www.orynycz.com\/sk\/veda\/studia\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\/","title":{"rendered":"Say It Right: Umel\u00fd preklad neur\u00f3nov\u00fdch strojov posil\u0148uje nov\u00fdch hovorcov na o\u017eivenie Lemko (2022)"},"content":{"rendered":"\n<h2 class=\"wp-block-heading\" id=\"h-abstract\">Abstrakt<\/h2>\n\n<p class=\"wp-block-paragraph\">Neur\u00f3nov\u00fd strojov\u00fd preklad poh\u00e1\u0148an\u00fd umelou inteligenciou by mohol \u010doskoro o\u017eivi\u0165 ohrozen\u00e9 jazyky t\u00fdm, \u017ee umo\u017en\u00ed nov\u00fdm hovorcom komunikova\u0165 v re\u00e1lnom \u010dase pomocou viet, ktor\u00e9 s\u00fa kvantitat\u00edvne bli\u017e\u0161ie k liter\u00e1rnej norme ako vety roden\u00fdch hovorcov, a to u\u017e od prv\u00e9ho d\u0148a ich cesty k obnove jazyka. Zatia\u013e \u010do Silicon Valley investuje obrovsk\u00e9 zdroje do technol\u00f3gie neur\u00f3nov\u00e9ho prekladu schopnej nad\u013eudskej r\u00fdchlosti a presnosti pre najpou\u017e\u00edvanej\u0161ie jazyky sveta, 98 % z nich zostalo pozadu, kv\u00f4li nedostatku korpusov: modely neur\u00f3nov\u00e9ho strojov\u00e9ho prekladu sa tr\u00e9nuj\u00fa na mili\u00f3noch slov dvojjazy\u010dn\u00e9ho textu, ktor\u00e9 pre v\u00e4\u010d\u0161inu jazykov jednoducho neexistuj\u00fa a ich zostavenie stoj\u00ed st\u00e1tis\u00edce americk\u00fdch dol\u00e1rov za jeden jazyk. <\/p>\n\n<p class=\"wp-block-paragraph\">Pre jazyky s n\u00edzkymi zdrojmi existuje vynaliezavej\u0161\u00ed pr\u00edstup, ak nie efekt\u00edvnej\u0161\u00ed: prenosov\u00e9 u\u010denie, ktor\u00e9 umo\u017e\u0148uje jazykom s ni\u017e\u0161\u00edmi zdrojmi profitova\u0165 z \u00faspechov jazykov s vy\u0161\u0161\u00edmi zdrojmi. V tomto experimente bola slu\u017eba neur\u00f3nov\u00e9ho prekladu Google z angli\u010dtiny do po\u013e\u0161tiny spojen\u00e1 s moj\u00edm klasick\u00fdm, pravidlami riaden\u00fdm motorom na preklad z angli\u010dtiny do ohrozen\u00e9ho, n\u00edzkoresursov\u00e9ho, v\u00fdchodoslovansk\u00e9ho jazyka Lemko. Syst\u00e9m dosiahol sk\u00f3re kvality dvojjazy\u010dn\u00e9ho hodnotenia (BLEU) 6,28, \u010do je nieko\u013ekon\u00e1sobne lep\u0161ie ako slu\u017eby Google Translate z angli\u010dtiny do \u0161tandardnej ukrajin\u010diny (BLEU 2,17), ru\u0161tiny (BLEU 1,10) a po\u013e\u0161tiny (BLEU 1,70). Nakoniec bol v\u00fdsledok tohto experimentu, prv\u00e1 prekladate\u013esk\u00e1 slu\u017eba z angli\u010dtiny do Lemko na svete, spr\u00edstupnen\u00fd na webovej adrese <code>www.LemkoTran.com<\/code>, aby umo\u017enil nov\u00fdm hovorcom o\u017eivi\u0165 ich jazyk.   <\/p>\n\n<p class=\"wp-block-paragraph\">Nov\u00ed hovorcovia s\u00fa k\u013e\u00fa\u010dom k o\u017eiveniu jazyka a mo\u017enos\u0165 \u201epoveda\u0165 to spr\u00e1vne\u201c v Lemko je teraz na dosah ruky.<\/p>\n\n<p class=\"wp-block-paragraph\"><strong>K\u013e\u00fa\u010dov\u00e9 slov\u00e1:<\/strong> Umel\u00e1 inteligencia zameran\u00e1 na \u010dloveka, revitaliz\u00e1cia jazyka, Lemko.<\/p>\n\n<div class=\"wp-block-file aligncenter\"><a id=\"wp-block-file--media-276bf8bd-1f68-4249-80c2-db9258ee7ce0\" href=\"https:\/\/www.orynycz.com\/wp-content\/uploads\/2024\/02\/OrynyczP_2022_HCI_preprint.pdf\">Z\u00edska\u0165 PDF<\/a><a href=\"https:\/\/www.orynycz.com\/wp-content\/uploads\/2024\/02\/OrynyczP_2022_HCI_preprint.pdf\" class=\"wp-block-file__button wp-element-button\" download=\"\" aria-describedby=\"wp-block-file--media-276bf8bd-1f68-4249-80c2-db9258ee7ce0\">Stiahnu\u0165<\/a><\/div>\n\n<p class=\"alert is-style-text-annotation is-style-text-annotation--1 wp-block-paragraph\">Pros\u00edm, citujte ako: Orynycz, P. (2022). Say It Right: AI Neural Machine Translation Empowers New Speakers to Revitalize Lemko. In: Degen, H., Ntoa, S. (eds) Artificial Intelligence in HCI. HCII 2022. Lecture Notes in Computer Science, vol 13336. Springer, Cham.       <a href=\"https:\/\/doi.org\/10.1007\/978-3-031-05643-7_37\" rel=\"nofollow\">https:\/\/doi.org\/10.1007\/978-3-031-05643-7_37<\/a><\/p>\n\n<p class=\"is-style-text-annotation has-text-color has-background has-link-color wp-elements-a8d599e270e0ef2d2076c9013637ce46 is-style-text-annotation--2 wp-block-paragraph\" style=\"color:#0f5132;background-color:#d1e7dd\">\u2705 T\u00e1to verzia pr\u00edspevku bola prijat\u00e1 na publikovanie po recenzii, ale nie je verziou z\u00e1znamu a neodr\u00e1\u017ea vylep\u0161enia po prijat\u00ed ani \u017eiadne opravy. Verzia z\u00e1znamu je dostupn\u00e1 online na   <a href=\"https:\/\/doi.org\/10.1007\/978-3-031-35894-4_10\"><\/a><a href=\"https:\/\/doi.org\/10.1007\/978-3-031-05643-7_37\" rel=\"nofollow\">https:\/\/doi.org\/10.1007\/978-3-031-05643-7_37<\/a>. Pou\u017e\u00edvanie tejto prijatej verzie podlieha podmienkam pou\u017e\u00edvania prijatej rukopisnej verzie vydavate\u013ea: <a href=\"https:\/\/www.springernature.com\/gp\/open-research\/policies\/accepted-manuscript-terms\" rel=\"nofollow\">https:\/\/www.springernature.com\/gp\/open-research\/policies\/accepted-manuscript-terms<\/a>. <\/p>\n\n<div class=\"wp-block-yoast-seo-table-of-contents yoast-table-of-contents\"><h2>Obsah<\/h2><ul><li><a href=\"#h-abstract\" data-level=\"2\">Abstrakt<\/a><\/li><li><a href=\"#h-1-introduction\" data-level=\"2\">1 \u00davod<\/a><ul><li><a href=\"#h-1-1-problems\" data-level=\"3\">1.1. Probl\u00e9my<\/a><\/li><li><a href=\"#h-1-2-work-so-far\" data-level=\"3\">1.2 Doteraj\u0161ia pr\u00e1ca<\/a><\/li><li><a href=\"#h-1-3-system-under-study\" data-level=\"3\">1.3 \u0160tudovan\u00fd syst\u00e9m<\/a><\/li><li><a href=\"#h-1-4-hypothesis\" data-level=\"3\">1.4 Hypot\u00e9za<\/a><\/li><li><a href=\"#h-1-5-predictions\" data-level=\"3\">1.5 Predpovede<\/a><\/li><li><a href=\"#h-1-6-methods-and-justification\" data-level=\"3\">1.6 Met\u00f3dy a zd\u00f4vodnenie<\/a><\/li><li><a href=\"#h-1-7-principal-results\" data-level=\"3\">1.7 Hlavn\u00e9 v\u00fdsledky<\/a><\/li><\/ul><\/li><li><a href=\"#h-2-materials-and-methods\" data-level=\"2\">2 Materi\u00e1ly a met\u00f3dy<\/a><ul><li><a href=\"#h-2-1-setup\" data-level=\"3\">2.1 Nastavenie<\/a><\/li><\/ul><\/li><li><a href=\"#h-3-results\" data-level=\"2\">3 V\u00fdsledky<\/a><ul><li><a href=\"#h-3-1-results-by-machine-translation-service\" data-level=\"3\">3.1 V\u00fdsledky pod\u013ea slu\u017eby strojov\u00e9ho prekladu<\/a><\/li><\/ul><\/li><li><a href=\"#h-4-discussion\" data-level=\"2\">4 Diskusia<\/a><\/li><li><a href=\"#h-references\" data-level=\"2\">Referencie<\/a><\/li><\/ul><\/div>\n\n<h2 class=\"wp-block-heading\" id=\"h-1-introduction\">1 \u00davod<\/h2>\n\n<h3 class=\"wp-block-heading\" id=\"h-1-1-problems\">1.1. Probl\u00e9my<\/h3>\n\n<p class=\"wp-block-paragraph\">Tento experiment si kladie za cie\u013e prispie\u0165 na miestnej \u00farovni k glob\u00e1lnemu probl\u00e9mu straty jazykov, ku ktorej m\u00f4\u017ee doch\u00e1dza\u0165 r\u00fdchlos\u0165ou jedn\u00e9ho jazyka denne, pri\u010dom pre\u017ei\u0165 m\u00e1 len jeden z desiatich jazykov <a id=\"cite-1\" href=\"#ref-1\">[1, s. 1329]<\/a>. V \u010dase tla\u010de pou\u017e\u00edva SIL International\u2019s <em>Ethnologue<\/em> roz\u0161\u00edren\u00fa stup\u0148ovan\u00fa \u0161k\u00e1lu medzigenera\u010dn\u00e9ho naru\u0161enia Lewis a Simons z roku 2010 na odhad, \u017ee 3 018 jazykov je ohrozen\u00fdch <a id=\"cite-2\" href=\"#ref-2\">[2]<\/a>, \u010do je 43 % zo 7 001 jednotliv\u00fdch \u017eiv\u00fdch jazykov zaznamenan\u00fdch v \u010dase tla\u010de v norme Medzin\u00e1rodnej organiz\u00e1cie pre normaliz\u00e1ciu ISO 639-3 <a id=\"cite-3\" href=\"#ref-3\">[3]<\/a>. Medzit\u00fdm Google Translate obsluhuje len 108 <a id=\"cite-4\" href=\"#ref-4\">[4]<\/a> a Facebook 112 <a href=\"#ref-5\" id=\"cite-5\">[5]<\/a>, \u010do je za\u010diatok. Napriek tomu je teraz jeden jazyk menej nedostato\u010dne obsluhovan\u00fd, ke\u010f\u017ee v\u00fdsledok tohto experimentu bol nasaden\u00fd na webov\u00fd server ako verejn\u00e1 prekladate\u013esk\u00e1 slu\u017eba.   <\/p>\n\n<p class=\"wp-block-paragraph\">Nov\u00e9 technol\u00f3gie umelej inteligencie l\u00e1kaj\u00fa pr\u00eds\u013eubom pomoci, ktor\u00e1 okam\u017eite kompenzuje stratu jazyka prostredn\u00edctvom interakcie \u010dlovek-po\u010d\u00edta\u010d. V mojom predch\u00e1dzaj\u00facom experimente dosiahli neur\u00f3nov\u00e9 motory novej gener\u00e1cie vy\u0161\u0161ie sk\u00f3re kvality pri preklade z ru\u0161tiny a po\u013e\u0161tiny do angli\u010dtiny ako \u013eudsk\u00e1 kontrola <a href=\"#ref-6\" id=\"cite-6-0\">[6, s. 9]<\/a>. Medzit\u00fdm Facebook a Google<sup>1<\/sup> investovali obrovsk\u00e9 zdroje do poskytovania lep\u0161\u00edch ako \u013eudsk\u00fdch automatick\u00fdch prekladate\u013esk\u00fdch syst\u00e9mov s nulov\u00fdmi n\u00e1kladmi pre spotrebite\u013ea.  <\/p>\n\n<p class=\"has-small-font-size wp-block-paragraph\"><sup>1<\/sup> Zverejnenie: Pracujem ako platen\u00fd lingvista a \u0161pecialista na kontrolu kvality prekladu pre projekt Google Translate v ru\u0161tine, po\u013e\u0161tine a ukrajin\u010dine; s\u00eddlo je v San Franciscu.<\/p>\n\n<p class=\"wp-block-paragraph\">Nad\u013eudsk\u00e1 umel\u00e1 inteligencia nie je lacn\u00e1: tr\u00e9ning neur\u00f3nov\u00fdch jazykov\u00fdch modelov si vy\u017eaduje dvojjazy\u010dn\u00e9 korpusy s po\u010dtom slov v stovk\u00e1ch tis\u00edc, a ide\u00e1lne mili\u00f3noch, \u010do by st\u00e1lo st\u00e1tis\u00edce dol\u00e1rov na preklad, sumy presahuj\u00face mo\u017enosti v\u00e4\u010d\u0161iny jazykov\u00fdch komun\u00edt s n\u00edzkymi zdrojmi. Na\u0161\u0165astie, tento experiment ukazuje, \u017ee existuj\u00fa vynaliezavej\u0161ie a efekt\u00edvnej\u0161ie sp\u00f4soby, ako reagova\u0165 na v\u00fdzvu vytv\u00e1rania prekladate\u013esk\u00fdch pom\u00f4cok na revitaliz\u00e1ciu ohrozen\u00fdch jazykov v prostred\u00ed s n\u00edzkymi zdrojmi. <\/p>\n\n<h3 class=\"wp-block-heading\" id=\"h-1-2-work-so-far\">1.2 Doteraj\u0161ia pr\u00e1ca<\/h3>\n\n<p class=\"wp-block-paragraph\">Vytvoril som prv\u00fd syst\u00e9m strojov\u00e9ho prekladu z Lemko do angli\u010dtiny na svete a spr\u00edstupnil som ho verejnosti. Jeho objekt\u00edvne sk\u00f3re kvality prekladu sa zlep\u0161uje: motor dosiahol sk\u00f3re dvojjazy\u010dn\u00e9ho hodnotenia (BLEU) 14,57 v lete 2021, ako bolo prezentovan\u00e9 odborn\u00edkom na konferencii Interservice\/Industry Training, Simulation and Education Conference N\u00e1rodnej asoci\u00e1cie obrann\u00e9ho priemyslu a publikovan\u00e9 v jej zborn\u00edku <a href=\"#ref-6\" id=\"cite-6\">[6]<\/a>. Pre porovnanie, ako \u013eudsk\u00fd prekladate\u013e pracuj\u00faci v ter\u00e9nnych podmienkach, odrezan\u00fd od vonkaj\u0161ieho sveta, som dosiahol BLEU 28,66. Do jesene 2021 motor dosiahol BLEU 15,74, ako bolo ozn\u00e1men\u00e9 lingvistom, akademikom a \u0161ir\u0161ej komunite na podujat\u00ed, ktor\u00e9 usporiadala University of Pittsburgh.<sup>2<\/sup>   <\/p>\n\n<p class=\"has-small-font-size wp-block-paragraph\"><sup>2<\/sup> Zverejnenie: podujatie sponzorovala Karpatsko-rus\u00ednska spolo\u010dnos\u0165 (Pensylv\u00e1nia) a University of Pittsburgh mi zaplatila za moju prezent\u00e1ciu.<\/p>\n\n<h3 class=\"wp-block-heading\" id=\"h-1-3-system-under-study\">1.3 \u0160tudovan\u00fd syst\u00e9m<\/h3>\n\n<p class=\"wp-block-paragraph\">Lemko je definit\u00edvne a\u017e v\u00e1\u017ene ohrozen\u00fd [<a href=\"#ref-6\">6, s. 3<\/a>, <a id=\"cite-7\" href=\"#ref-7\">7, s. 177-178<\/a>], n\u00edzkoresursov\u00fd <a id=\"cite-8\" href=\"#ref-8\">[8]<\/a>, ofici\u00e1lne uznan\u00fd men\u0161inov\u00fd jazyk <a id=\"cite-9\" href=\"#ref-9\">[9]<\/a>, pravdepodobne p\u00f4vodn\u00fd pre cezhrani\u010dn\u00e9 vyso\u010diny ju\u017ene od metropolitn\u00fdch oblast\u00ed Krakova, Tarnova a Rzeszowa; historick\u00e9 vymedzuj\u00face izoglosy bud\u00fa, d\u00fafajme, t\u00e9mou bud\u00facej pr\u00e1ce. Po\u013esk\u00fd \u0161tatistick\u00fd \u00farad v roku 2011 zaznamenal 6 279 obyvate\u013eov, pre ktor\u00fdch bolo Lemko jazykom \u201ezvy\u010dajne pou\u017e\u00edvan\u00fdm doma\u201c (aj ke\u010f okrem po\u013e\u0161tiny) [<a id=\"cite-10\" href=\"#ref-10\">10, s. 3<\/a>], \u010do predstavuje 12 % n\u00e1rast oproti 5 605, pre ktor\u00fdch bolo Lemko \u201enaj\u010dastej\u0161ie hovoren\u00fdm jazykom doma\u201c v roku 2002 [<a id=\"cite-11\" href=\"#ref-11\">11, s. 6<\/a>, <a href=\"#ref-12\" id=\"cite-12\">12, s. 7<\/a>]. V \u010dase tla\u010de sa v\u00fdsledky nov\u00e9ho s\u010d\u00edtania s\u010d\u00edtavaj\u00fa.  <\/p>\n\n<p class=\"wp-block-paragraph\">Lemko je klasifikovate\u013en\u00e9 ako v\u00fdchodoslovansk\u00fd jazyk, preto\u017ee sp\u013a\u0148a obvykl\u00e9 krit\u00e9ri\u00e1 genetick\u00fdch \u0161truktur\u00e1lnych znakov, z ktor\u00fdch najv\u00fdznamnej\u0161\u00edm je pleof\u00f3nia [<a id=\"cite-13\" href=\"#ref-13\">13, s. 20<\/a>], pri ktorej sa predpoklad\u00e1, \u017ee samohl\u00e1ska vznikla v praslovansk\u00fdch sekvenci\u00e1ch spoluhl\u00e1sky  <em><code>C<\/code><\/em>  nasledovanej strednou alebo n\u00edzkou samohl\u00e1skou  <code>V<\/code>  (<code>*e<\/code>, alebo  <code>*o<\/code>, s ktorou sa  <code>*a<\/code>  zl\u00fa\u010dilo <a id=\"cite-14\" href=\"#ref-14\">[14, s. 366]<\/a>), nasledovanej likvidou R (t.j.   <code>*l<\/code>  alebo  <code>*r<\/code>), nasledovanou \u010fal\u0161ou spoluhl\u00e1skou  <em><code>C<\/code><\/em>, t.j. <code>CVRC &gt; CVRVC<\/code>. Na ilustr\u00e1ciu porovnajte staroanglick\u00e9 slovo pre \u201etopi\u0165\u201c, <em>meltan<\/em> (<code>CVRC<\/code>) <a id=\"cite-15-0\" href=\"#ref-15\">[15, s. 718]<\/a> s jeho predpokladan\u00fdm lemkovsk\u00fdm pr\u00edbuzn\u00fdm <em>mo\u0142\u00f3dyj<\/em> [<a id=\"cite-16\" href=\"#ref-16\">16, s. 92<\/a>, <a id=\"cite-17-0\" href=\"#ref-17\">17, s. 150<\/a>] (<code>CVRC<\/code>), \u010do znamen\u00e1 \u201emlad\u00fd\u201c. Medzi \u010fal\u0161ie v\u00fdchodoslovansk\u00e9 pr\u00edbuzn\u00e9 patria ukrajinsk\u00e9 <em>mo\u0142od\u00fdj<\/em> a rusk\u00e9 <em>mo\u0142od\u00f3j<\/em> <a id=\"cite-17-1\" href=\"#ref-17\">[17]<\/a>, obe vykazuj\u00face samohl\u00e1sku po likvide (<code>CVRVC<\/code>).   Medzit\u00fdm z\u00e1padoslovansk\u00e9 jazyky nemaj\u00fa samohl\u00e1sku pred likvidou; porovnajte po\u013esk\u00e9 <em>m\u0142ody<\/em> a slovensk\u00e9 <em>mlad\u00fd<\/em> (obe <code>CRVC<\/code>) <a id=\"cite-17-2\" href=\"#ref-17\">[17]<\/a>. \u010ealej sa predpoklad\u00e1 pr\u00edbuznos\u0165 pre in\u00e9 slov\u00e1 prelo\u017eite\u013en\u00e9 ako \u201emierny\u201c, vr\u00e1tane sanskritsk\u00e9ho <em>m\u1e5bd\u00fa<\/em> (<code>CRC<\/code>) <a id=\"cite-18\" href=\"#ref-18\">[18, s. 830]<\/a> a latinsk\u00e9ho <em>mollis<\/em> (<code>CVRC<\/code> ak z *<em>moldvis<\/em>) [<a id=\"cite-15-1\" href=\"#ref-15\">15<\/a>, <a id=\"cite-17-3\" href=\"#ref-17\">17<\/a>, <a href=\"#ref-19\" id=\"cite-19\">19, s. 323<\/a>]. <\/p>\n\n<p class=\"wp-block-paragraph\">V tomto experimente sa nehodnotilo, ako dobre Lemko sp\u013a\u0148a obvykl\u00e9, modern\u00e9 ukrajinsk\u00e9 krit\u00e9ri\u00e1 genetick\u00fdch \u0161truktur\u00e1lnych znakov. Av\u0161ak, podobnos\u0165 medzi Lemko a \u0161tandardnou ukrajin\u010dinou bola kvantifikovan\u00e1, po prv\u00fdkr\u00e1t v tla\u010di, o ktorej viem. Ni\u017e\u0161ie, m\u00f4j Lemko motor dosiahol sk\u00f3re BLEU 6,28, takmer trikr\u00e1t vy\u0161\u0161ie ako sk\u00f3re ukrajin\u010diny Google Translate s BLEU 2,17. \u010eal\u0161ie experimenty by sa mohli vykona\u0165 za \u00fa\u010delom kvantifik\u00e1cie podobnosti medzi Lemko, \u0161tandardnou ukrajin\u010dinou, po\u013e\u0161tinou a rus\u00edn\u010dinou, ako je kodifikovan\u00e1 na Slovensku, ako aj nov\u00fd poh\u013ead na typologick\u00fa klasifik\u00e1ciu Lemko.   <\/p>\n\n<p class=\"wp-block-paragraph\">Mno\u017estvo a kvalita zdrojov sa zlep\u0161uje, rovnako ako vynaliezavos\u0165 posilnen\u00e1 technol\u00f3giou. V\u0161etky zn\u00e1me dvojjazy\u010dn\u00e9 korpusy, obsahuj\u00face menej ako sedemdesiattis\u00edc lemkovsk\u00fdch slov, boli zhroma\u017eden\u00e9 pre tento experiment. \u010cist\u00edm dvojjazy\u010dn\u00fd korpus prepisov rozhovorov veden\u00fdch s roden\u00fdmi hovorcami v Po\u013esku a mojich prekladov do angli\u010dtiny, ktor\u00e9 mi zaplatil americk\u00fd klient a povolil mi ich pou\u017ei\u0165. Taktie\u017e zostavujem monolingv\u00e1lne korpusy, ktor\u00e9 v \u010dase tla\u010de celkovo obsahuj\u00fa 534 512 slov.   <\/p>\n\n<h3 class=\"wp-block-heading\" id=\"h-1-4-hypothesis\">1.4 Hypot\u00e9za<\/h3>\n\n<p class=\"wp-block-paragraph\">Na z\u00e1klade m\u00f4jho subjekt\u00edvneho dojmu ako profesion\u00e1lneho prekladate\u013ea, \u017ee roden\u00ed hovorcovia Lemko, s ktor\u00fdmi som robil rozhovory v Po\u013esku, s v\u00e4\u010d\u0161ou pravdepodobnos\u0165ou pou\u017e\u00edvali slov\u00e1 s o\u010dividn\u00fdmi po\u013esk\u00fdmi pr\u00edbuzn\u00fdmi ako \u0161tandardn\u00e9 ukrajinsk\u00e9, som predpokladal, \u017ee za inak rovnak\u00fdch podmienok by sa stroj mohol nakonfigurova\u0165 na preklad do Lemko z angli\u010dtiny a dosiahnu\u0165 objekt\u00edvne sk\u00f3re kvality BLEU vy\u0161\u0161ie ako slu\u017eby Google Translate pre ukrajin\u010dinu a ru\u0161tinu.<\/p>\n\n<h3 class=\"wp-block-heading\" id=\"h-1-5-predictions\">1.5 Predpovede<\/h3>\n\n<p class=\"wp-block-paragraph\"><strong>Prekladate\u013esk\u00fd syst\u00e9m Lemko.<\/strong>  Predpokladal som, \u017ee vy\u0161\u0161ie uveden\u00fd prekladate\u013esk\u00fd syst\u00e9m dosiahne sk\u00f3re BLEU 15 pri preklade do Lemko z angli\u010dtiny oproti dvojjazy\u010dn\u00e9mu korpusu.<\/p>\n\n<p class=\"wp-block-paragraph\"><strong>Google Translate.<\/strong><\/p>\n\n<p class=\"wp-block-paragraph\"><em>Slu\u017eba z angli\u010dtiny do ukrajin\u010diny<\/em>. Predpokladal som, \u017ee slu\u017eba Google Translate z angli\u010dtiny do ukrajin\u010diny dosiahne sk\u00f3re BLEU 10 oproti dvojjazy\u010dn\u00e9mu korpusu. <\/p>\n\n<p class=\"wp-block-paragraph\"><em>Slu\u017eba z angli\u010dtiny do ru\u0161tiny<\/em>. Predpokladal som, \u017ee slu\u017eba Google Translate z angli\u010dtiny do ru\u0161tiny dosiahne sk\u00f3re BLEU 1 oproti dvojjazy\u010dn\u00e9mu korpusu. <\/p>\n\n<h3 class=\"wp-block-heading\" id=\"h-1-6-methods-and-justification\">1.6 Met\u00f3dy a zd\u00f4vodnenie<\/h3>\n\n<p class=\"wp-block-paragraph\">V z\u00e1ujme r\u00fdchlosti, \u00faspory zdrojov a robustnosti bol notebook, ktor\u00fd m\u00f4j zamestn\u00e1vate\u013e vyradil ako zastaran\u00fd, nakonfigurovan\u00fd na preklad do Lemko a na volanie slu\u017eby Google Cloud Platform Google Translate, ako aj na vyhodnocovanie uveden\u00fdch prekladov pomocou priemyseln\u00e9ho \u0161tandardu BLEU.<\/p>\n\n<h3 class=\"wp-block-heading\" id=\"h-1-7-principal-results\">1.7 Hlavn\u00e9 v\u00fdsledky<\/h3>\n\n<p class=\"wp-block-paragraph\">Prekladate\u013esk\u00fd syst\u00e9m z angli\u010dtiny do Lemko dosiahol kumulat\u00edvne sk\u00f3re BLEU <code>6.28431824990417<\/code>. Medzit\u00fdm slu\u017eba Google Translate pre ukrajin\u010dinu dosiahla BLEU <code>2.16830846776652<\/code>, jej slu\u017eba pre ru\u0161tinu BLEU <code>1.10424105952048<\/code> a kontrola po\u013e\u0161tiny prep\u00edsanej do cyriliky BLEU <code>1.70036447680114<\/code>. <\/p>\n\n<h2 class=\"wp-block-heading\" id=\"h-2-materials-and-methods\">2 Materi\u00e1ly a met\u00f3dy<\/h2>\n\n<p class=\"wp-block-paragraph\">Vy\u0161\u0161ie uveden\u00e1 hypot\u00e9za bola testovan\u00e1 v\u00fdpo\u010dtom sk\u00f3re kvality BLEU pre ka\u017ed\u00fd prekladate\u013esk\u00fd syst\u00e9m nastaven\u00fd sp\u00f4sobom podrobne op\u00edsan\u00fdm ni\u017e\u0161ie.<\/p>\n\n<h3 class=\"wp-block-heading\" id=\"h-2-1-setup\">2.1 Nastavenie<\/h3>\n\n<p class=\"wp-block-paragraph\"><strong>Hardv\u00e9r.<\/strong>  Experiment sa uskuto\u010dnil na notebooku HP Elitebook 850 G2 s procesorom Core i7-5600U 2,6 GHz a 16 gigabajtami pam\u00e4te RAM. M\u00f4j zamestn\u00e1vate\u013e ho vyradil ako zastaran\u00fd a v \u010dase tla\u010de bol pon\u00fakan\u00fd na predaj za 450 USD. <\/p>\n\n<p class=\"wp-block-paragraph\"><em>Konfigur\u00e1cia<\/em>. V menu z\u00e1kladn\u00e9ho vstupno-v\u00fdstupn\u00e9ho syst\u00e9mu (BIOS) bolo zariadenie nakonfigurovan\u00e9 tak, aby umo\u017e\u0148ovalo technol\u00f3giu virtualiz\u00e1cie (VTx). <\/p>\n\n<p class=\"wp-block-paragraph\"><strong>Opera\u010dn\u00fd syst\u00e9m.<\/strong>  Windows 10 Professional 64 bit bol nain\u0161talovan\u00fd na hol\u00fd hardv\u00e9r. Bolo zabezpe\u010den\u00e9, aby boli povolen\u00e9 funkcie Windows <code>Virtual Machine Platform<\/code> a <code>Windows Subsystem for Linux<\/code>. N\u00e1sledne boli nain\u0161talovan\u00e9 <code>WSL2 Linux kernel update for x64 <\/code> stroje (wsl_update_x64.msi) dostupn\u00e9 od spolo\u010dnosti Microsoft na <code><a href=\"https:\/\/aka.ms\/wsl2kernel\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">https:\/\/aka.ms\/wsl2kernel<\/a><\/code>.  <\/p>\n\n<p class=\"wp-block-paragraph\"><strong>Softv\u00e9r.<\/strong> In\u0161tal\u00e1tor Docker Desktop pre Windows verzie 4.4.3 (73365) bol stiahnut\u00fd z <code><a href=\"https:\/\/www.docker.com\/get-started\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">https:\/\/www.docker.com\/get-started<\/a><\/code> a spusten\u00fd s mo\u017enos\u0165ou <code>Install required Windows components for WSL 2 selected<\/code>.<\/p>\n\n<p class=\"wp-block-paragraph\"><strong>Bal\u00ed\u010dky.<\/strong>  Experiment z\u00e1visel od ni\u017e\u0161ie uveden\u00fdch bal\u00edkov z Python Package Index.<\/p>\n\n<p class=\"wp-block-paragraph\"><strong>SacreBLEU.<\/strong>  Verzia 2.0.0 bola nain\u0161talovan\u00e1 pomocou bal\u00edka Python zdokumentovan\u00e9ho na nasleduj\u00facom univerz\u00e1lnom lok\u00e1tore zdrojov (URL):<br\/><code><a href=\"https:\/\/pypi.org\/project\/sacrebleu\/2.0.0\/\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">https:\/\/pypi.org\/project\/sacrebleu\/2.0.0\/<\/a><\/code><\/p>\n\n<p class=\"wp-block-paragraph\"><em>Klientska kni\u017enica Google Cloud Translation API<\/em>. Verzia 2.0.1 bola nain\u0161talovan\u00e1 pomocou bal\u00edka Python zdokumentovan\u00e9ho na univerz\u00e1lnom lok\u00e1tore zdrojov (URL)   <code><a href=\"https:\/\/pypi.org\/project\/google-cloud-translate\/2.0.1\/\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">https:\/\/pypi.org\/project\/google-cloud-translate\/2.0.1\/<\/a><\/code><\/p>\n\n<p class=\"wp-block-paragraph\">Vy\u0161\u0161ie uveden\u00e9 z\u00e1vislosti boli \u0161pecifikovan\u00e9 v s\u00fabore po\u017eiadaviek nasledovne:<br\/><code>google-cloud-translate==2.0.1<\/code><br\/><code>sacrebleu==2.0.0<\/code>\n<\/p>\n\n<p class=\"wp-block-paragraph\"><strong>Kontajner.<\/strong><\/p>\n\n<p class=\"wp-block-paragraph\"><em>Zostavenie<\/em>. Experiment bol spusten\u00fd v kontajneri Docker s najnov\u0161ou verziou programovacieho jazyka Python, ktor\u00e1 bola v tom \u010dase verzia 3.10.2, be\u017eiaca na opera\u010dnom syst\u00e9me Debian Bullseye 11 Linux architekt\u00fary AMD64, so skr\u00e1ten\u00fdm digestom Secure Hash Algorithm 2 <code>bcb158d5ddb6<\/code>, z\u00edskate\u013en\u00fdm pomocou nasleduj\u00faceho pr\u00edkazu: <br\/><code>docker pull python@sha256:bcb158d5ddb636fa3aa567c987e7fcf61113307820d466813527ca90d60fedc7<\/code><\/p>\n\n<p class=\"wp-block-paragraph\"><em>Runtime<\/em>. Kontajner bol nakonfigurovan\u00fd tak, aby ukladal surov\u00e9 experiment\u00e1lne d\u00e1tov\u00e9 s\u00fabory do lok\u00e1lne pripojen\u00e9ho zv\u00e4zku. <\/p>\n\n<p class=\"wp-block-paragraph\"><strong>Hodnotenie kvality prekladu.<\/strong><br\/>Sk\u00f3re kvality prekladu bolo vypo\u010d\u00edtan\u00e9 pod\u013ea metriky BLEU pomocou verzie 2.0.0 n\u00e1stroja <em>SacreBLEU<\/em>, ktor\u00fd vyna\u0161iel Post <a href=\"#ref-20\" id=\"cite-20\">[20]<\/a>.<\/p>\n\n<p class=\"wp-block-paragraph\"><strong>Citlivos\u0165 na ve\u013ek\u00e9 a mal\u00e9 p\u00edsmen\u00e1.<\/strong>  Hodnotenie sa vykonalo s oh\u013eadom na ve\u013ek\u00e9 a mal\u00e9 p\u00edsmen\u00e1.<\/p>\n\n<p class=\"wp-block-paragraph\"><strong>Tokeniz\u00e1cia.<\/strong>  Segmenty boli tokenizovan\u00e9 pomocou verzie 13a \u0161tandardn\u00e9ho skriptu na hodnotenie Workshop on Statistical Machine Translation, intern\u00e9ho postupu tokeniz\u00e1cie metriky.<\/p>\n\n<p class=\"wp-block-paragraph\"><strong>Met\u00f3da vyhladzovania.<\/strong> Pou\u017eila sa met\u00f3da vyhladzovania vyvinut\u00e1 N\u00e1rodn\u00fdm in\u0161tit\u00fatom pre \u0161tandardy a technol\u00f3gie zamestnancami feder\u00e1lnej vl\u00e1dy Spojen\u00fdch \u0161t\u00e1tov pre ich s\u00fapravu n\u00e1strojov Multimodal Information Group BLEU, ktor\u00e1 je tre\u0165ou technikou op\u00edsanou Chenom a Cherrym <a href=\"#ref-21\" id=\"cite-21\">[21, s. 363]<\/a>, \u0161tandardne.<\/p>\n\n<p class=\"wp-block-paragraph\"><strong>Podpis<\/strong>. Vy\u0161\u0161ie uveden\u00e9 nastavenia vytvorili nasleduj\u00faci podpis:<br\/>n <code>refs:1|case:mixed|eff:no|tok:13a|smooth:exp|version:2.0.0<\/code><\/p>\n\n<p class=\"wp-block-paragraph\"><strong>Kalibr\u00e1cia<\/strong>. Nakonfigurovan\u00fd ako vy\u0161\u0161ie, stroj produkuje nasleduj\u00faci v\u00fdstup: <\/p>\n\n<div>\n<em>Segment 1031.<\/em><figure class=\"wp-block-table is-style-stripes\"><table><tbody><tr><td>Anglick\u00fd zdroj<\/td><td colspan=\"2\"><code>Everything was there.<\/code><\/td><\/tr><tr><td>Lemko referencia a transliter\u00e1cia<\/td><td><code>\u0412\u0448\u044b\u0442\u043a\u043e \u0442\u0430\u043c \u0431\u044b\u043b\u043e.<\/code><\/td><td><code>V\u0161\u0177tko tam b\u0177lo.<\/code><\/td><\/tr><tr><td><code>Lemkotran.com<\/code>  hypot\u00e9za a transliter\u00e1cia<\/td><td><code>\u0412\u0448\u044b\u0442\u043a\u043e \u0442\u0430\u043c \u0431\u044b\u043b\u043e.<\/code><\/td><td><code>V\u0161\u0177tko tam b\u0177lo.<\/code><\/td><\/tr><tr><td>Sk\u00f3re<\/td><td colspan=\"2\"><code>BLEU = 100.00 100.0\/100.0\/100.0\/100.0 (BP = 1.000 ratio = 1.000 hyp_len = 4 ref_len = 4)<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n<\/div>\n\n<p class=\"wp-block-paragraph\"><em>Vysvetlenie<\/em>. Hypotetick\u00fd segment bol identick\u00fd s referen\u010dn\u00fdm a stroj dosiahol perfektn\u00e9 sk\u00f3re BLEU 100. <\/p>\n\n<div>\n<em>Segment 179.<\/em><figure class=\"wp-block-table is-style-stripes\"><table><tbody><tr><td>Anglick\u00fd zdroj<\/td><td colspan=\"2\"><code>I don't remember what year.<\/code><\/td><\/tr><tr><td>Lemko referencia a transliter\u00e1cia<\/td><td><code>\u041d\u0435 \u043f\u0430\u043c\u044f\u0442\u0430\u043c \u0432 \u043a\u043e\u0442\u0440\u044b\u043c \u0440\u043e\u0446\u0456.<\/code><\/td><td><code>Ne pamjatam v kotr\u0177m roci.<\/code><\/td><\/tr><tr><td><code>Lemkotran.com<\/code>  hypot\u00e9za a transliter\u00e1cia<\/td><td><code>\u041d\u0456 \u043f\u0430\u043c\u044f\u0442\u0430\u043c, \u0432 \u043a\u043e\u0442\u0440\u044b\u043c \u0440\u043e\u0446\u0456.<\/code><\/td><td><code>Ni pamjatam, v kotr\u0177m roci.<\/code><\/td><\/tr><tr><td>Sk\u00f3re<\/td><td colspan=\"2\"><code>BLEU = 43.47 71.4\/50.0\/40.0\/25.0 (BP = 1.000 ratio = 1.167 hyp_len = 7 ref_len = 6)<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n<\/div>\n\n<p class=\"wp-block-paragraph\"><em>Vysvetlenie<\/em>. Hypot\u00e9za sa l\u00ed\u0161ila od referencie o dva znaky. Stroj nespr\u00e1vne prelo\u017eil \u010dasticu neguj\u00facu sloveso, pou\u017eil slovo pre \u201enie\u201c (<em>ni<\/em>) namiesto o\u010dak\u00e1van\u00e9ho slova pre \u201enie\u201c (<em>ne<\/em>). To sa odvtedy do zna\u010dnej miery opravilo. Stroj tie\u017e pridal \u010diarku za <em>pamjatam<\/em>, \u010do znamen\u00e1 \u201epam\u00e4t\u00e1m si\u201c. To zn\u00ed\u017eilo sk\u00f3re z perfektn\u00e9ho sk\u00f3re 100 na 43,47.     <\/p>\n\n<p class=\"wp-block-paragraph\"><strong>Kontrola<\/strong>. Ke\u010f\u017ee korpus je zalo\u017een\u00fd na rozhovoroch uskuto\u010dnen\u00fdch v Po\u013esku, preklady do po\u013e\u0161tiny boli pou\u017eit\u00e9 ako kontrola. Boli transliterovan\u00e9 do cyriliky obr\u00e1ten\u00edm pravidiel pre transliter\u00e1ciu mien Lemko, ktor\u00e9 stanovilo po\u013esk\u00e9 Ministerstvo vn\u00fatra a administrat\u00edvy <a id=\"cite-22\" href=\"#ref-22\">[22, str. 6564]<\/a>. Po\u013esk\u00e9 nosov\u00e9 samohl\u00e1sky boli rozlo\u017een\u00e9 na samohl\u00e1sku plus nosov\u00fa z\u00e1verov\u00fa spoluhl\u00e1sku, okrem pr\u00edpadov pred aproximantmi, kde boli priamo denazalizovan\u00e9. Na konci slova bola predn\u00e1 nosov\u00e1 samohl\u00e1ska \/\u0119\/ jednoducho denazalizovan\u00e1 a zadn\u00e1 \/\u0105\/ bola transliterovan\u00e1, akoby po nej nasledovala zubn\u00e1 z\u00e1verov\u00e1 spoluhl\u00e1ska.    <\/p>\n\n<h2 class=\"wp-block-heading\" id=\"h-3-results\">3 V\u00fdsledky<\/h2>\n\n<p class=\"wp-block-paragraph\">Motor dostupn\u00fd verejnosti na <code>www.LemkoTran.com<\/code> obsadil prv\u00e9 miesto s kumulat\u00edvnym sk\u00f3re kvality prekladu BLEU 6,28, \u010do je takmer trojn\u00e1sobok sk\u00f3re druh\u00e9ho v porad\u00ed, slu\u017eby Google Translate z angli\u010dtiny do ukrajin\u010diny (BLEU 2,17). \u010ealej nasledovala jej slu\u017eba z angli\u010dtiny do po\u013e\u0161tiny (BLEU 1,70) a jej slu\u017eba z angli\u010dtiny do ru\u0161tiny bola na poslednom mieste (BLEU 1,10). <\/p>\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"716\" src=\"https:\/\/www.orynycz.com\/wp-content\/uploads\/2022\/02\/chart-1024x716.png\" alt=\"\" class=\"wp-image-2157\" srcset=\"https:\/\/www.orynycz.com\/wp-content\/uploads\/2022\/02\/chart-1024x716.png 1024w, https:\/\/www.orynycz.com\/wp-content\/uploads\/2022\/02\/chart-300x210.png 300w, https:\/\/www.orynycz.com\/wp-content\/uploads\/2022\/02\/chart-768x537.png 768w, https:\/\/www.orynycz.com\/wp-content\/uploads\/2022\/02\/chart.png 1338w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><figcaption class=\"wp-element-caption\"><strong>Tabu\u013eka 1<\/strong>. Kvalita prekladu z angli\u010dtiny do Lemko: LemkoTran.com verzus Google Translate <\/figcaption><\/figure>\n\n<h3 class=\"wp-block-heading\" id=\"h-3-1-results-by-machine-translation-service\">3.1 V\u00fdsledky pod\u013ea slu\u017eby strojov\u00e9ho prekladu<\/h3>\n\n<p class=\"wp-block-paragraph\"><strong>Kontrola.<\/strong> Pri transliter\u00e1cii do cyriliky dosiahli preklady Google Translate do \u0161tandardnej po\u013e\u0161tiny sk\u00f3re BLEU na \u00farovni korpusu 1,70. Uk\u00e1\u017eky jeho v\u00fdkonov s\u00fa nasledovn\u00e9: <\/p>\n\n<div>\n<em>Segment 2174.<\/em><figure class=\"wp-block-table is-style-stripes\"><table><tbody><tr><td>Anglick\u00fd zdroj<\/td><td colspan=\"2\"><code>We had still been in Izby, right.<\/code><\/td><\/tr><tr><td>Lemko referen\u010dn\u00fd text a transliter\u00e1cia<\/td><td><code>\u0422\u043e \u043c\u044b \u0456\u0449\u044b \u0431\u044b\u043b\u0438 \u0432 \u0406\u0437\u0431\u0430\u0445, \u0442\u0430\u043a.<\/code><\/td><td><code>To m\u0177 i\u0161\u010d\u0177 b\u0177ly v Izbach, tak.<\/code><\/td><\/tr><tr><td>Po\u013esk\u00e1 hypot\u00e9za a transliter\u00e1cia<\/td><td><code>\u0411\u0438\u043b\u0456\u0441\u044c\u043c\u0438 \u0454\u0449\u0435 \u0432 \u0406\u0437\u0431\u0430\u0445, \u0442\u0430\u043a.<\/code><\/td><td><code>Byli\u015bmy jeszcze w Izbach, tak.<\/code><\/td><\/tr><tr><td>Sk\u00f3re<\/td><td colspan=\"2\"><code>BLEU = 46.20<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n<\/div>\n\n<div>\n<em>Segment 854.<\/em><figure class=\"wp-block-table is-style-stripes\"><table><tbody><tr><td>Anglick\u00fd zdroj<\/td><td colspan=\"2\"><code>And that's what it's all about.<\/code><\/td><\/tr><tr><td>Lemko referen\u010dn\u00fd text a transliter\u00e1cia<\/td><td><code>\u0406 \u043e \u0442\u043e \u0445\u043e\u0434\u0438\u0442.<\/code><\/td><td><code>I o to chodyt.<\/code><\/td><\/tr><tr><td>Po\u013esk\u00e1 hypot\u00e9za a transliter\u00e1cia<\/td><td><code>\u0406 \u043e \u0442\u043e \u0432\u043b\u0430\u0441\u044c\u043d\u0454 \u0445\u043e\u0434\u0437\u0456.<\/code><\/td><td><code>I o to w\u0142a\u015bnie chodzi.<\/code><\/td><\/tr><tr><td>Sk\u00f3re<\/td><td colspan=\"2\"><code>BLEU = 32.47<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n<\/div>\n\n<div>\n<em>Segment 217.<\/em><figure class=\"wp-block-table is-style-stripes\"><table><tbody><tr><td>Anglick\u00fd zdroj<\/td><td colspan=\"2\"><code>And that's what it's all about.<\/code><\/td><\/tr><tr><td>Lemko referen\u010dn\u00fd text a transliter\u00e1cia<\/td><td><code>\u0422\u0430\u043a \u043c\u0456 \u043f\u043e\u0432\u0456\u043b.<\/code><\/td><td><code>Tak mi povil.<\/code><\/td><\/tr><tr><td>Po\u013esk\u00e1 hypot\u00e9za a transliter\u00e1cia<\/td><td><code>\u0422\u0430\u043a \u043c\u0456 \u043f\u043e\u0432\u0454\u0434\u0437\u044f\u043b.<\/code><\/td><td><code>Tak mi powiedzia\u0142.<\/code><\/td><\/tr><tr><td>Sk\u00f3re<\/td><td colspan=\"2\"><code>BLEU = 35.36<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n<\/div>\n\n<p class=\"wp-block-paragraph\"><strong>Hybridn\u00fd anglicko-Lemko motor<\/strong>. Motor vo\u013ene dostupn\u00fd verejnosti na URL adrese <code>www.LemkoTran.com<\/code> dosiahol sk\u00f3re BLEU na \u00farovni korpusu 6,28. <\/p>\n\n<div>\n<em>Segment 1031.<\/em><figure class=\"wp-block-table is-style-stripes\"><table><tbody><tr><td>Anglick\u00fd zdroj<\/td><td colspan=\"2\"><code>Everything was there.<\/code><\/td><\/tr><tr><td>Lemko referen\u010dn\u00fd text a transliter\u00e1cia<\/td><td><code>\u0412\u0448\u044b\u0442\u043a\u043e \u0442\u0430\u043c \u0431\u044b\u043b\u043e.<\/code><\/td><td><code>V\u0161\u0177tko tam b\u0177lo.<\/code><\/td><\/tr><tr><td><code>Lemkotran.com<\/code>  hypot\u00e9za a transliter\u00e1cia<\/td><td><code>\u0412\u0448\u044b\u0442\u043a\u043e \u0442\u0430\u043c \u0431\u044b\u043b\u043e.<\/code><\/td><td><code>V\u0161\u0177tko tam b\u0177lo.<\/code><\/td><\/tr><tr><td>Sk\u00f3re<\/td><td colspan=\"2\"><code>BLEU = 100.00<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n<\/div>\n\n<div>\n<em>Segment 1445.<\/em><figure class=\"wp-block-table is-style-stripes\"><table><tbody><tr><td>Anglick\u00fd zdroj<\/td><td colspan=\"2\"><code>But that officer took that medal and said,<\/code><\/td><\/tr><tr><td>Lemko referen\u010dn\u00fd text a transliter\u00e1cia<\/td><td><code>\u0410\u043b\u0435 \u0442\u043e\u0442 \u043e\u0444\u0456\u0446\u0435\u0440 \u0432\u0437\u044f\u043b \u0442\u043e\u0442 \u043c\u0435\u0434\u0430\u043b\u044c \u0456 \u043f\u043e\u0432\u0456\u0434\u0430\u0442:<\/code><\/td><td><code>Ale tot oficer vzial tot medal' i povidat:<\/code><\/td><\/tr><tr><td><code>Lemkotran.com<\/code>  hypot\u00e9za a transliter\u00e1cia<\/td><td><code>\u0410\u043b\u0435 \u0442\u043e\u0442 \u043e\u0444\u0456\u0446\u0435\u0440 \u0432\u0437\u044f\u043b \u0442\u043e\u0442 \u043c\u0435\u0434\u0430\u043b\u044c \u0456 \u043f\u043e\u0432\u0456\u043b:<\/code><\/td><td><code>Ale tot oficer vzial tot medal' i povil:<\/code><\/td><\/tr><tr><td>Sk\u00f3re<\/td><td colspan=\"2\"><code>BLEU = 75.06<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n<\/div>\n\n<div>\n<em>Segment 217.<\/em><figure class=\"wp-block-table is-style-stripes\"><table><tbody><tr><td>Anglick\u00fd zdroj<\/td><td colspan=\"2\"><code>That's what he said to me.<\/code><\/td><\/tr><tr><td>Lemko referen\u010dn\u00fd text a transliter\u00e1cia<\/td><td><code>\u0422\u0430\u043a \u043c\u0456 \u043f\u043e\u0432\u0456\u043b.<\/code><\/td><td><code>Tak mi povil.<\/code><\/td><\/tr><tr><td><code>Lemkotran.com<\/code>  hypot\u00e9za a transliter\u00e1cia<\/td><td><code>\u0422\u0430\u043a \u043c\u0456 \u043f\u043e\u0432\u0456\u043b.<\/code><\/td><td><code>Tak mi povil.<\/code><\/td><\/tr><tr><td>Sk\u00f3re<\/td><td colspan=\"2\"><code>BLEU = 100.00<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n<\/div>\n\n<p class=\"wp-block-paragraph\"><strong>Ukrajin\u010dina<\/strong>. Preklady Google Translate do \u0161tandardnej ukrajin\u010diny dosiahli sk\u00f3re BLEU na \u00farovni korpusu 2,35. <\/p>\n\n<div>\n<em>Segment 2419.<\/em><figure class=\"wp-block-table is-style-stripes\"><table><tbody><tr><td>Anglick\u00fd zdroj<\/td><td colspan=\"2\"><code>Where and when?<\/code><\/td><\/tr><tr><td>Lemko referen\u010dn\u00fd text a transliter\u00e1cia<\/td><td><code>\u0414\u0435 \u0456 \u043a\u043e\u043b\u0438?<\/code><\/td><td><code>De i koly?<\/code><\/td><\/tr><tr><td>Ukrajinsk\u00e1 hypot\u00e9za a transliter\u00e1cia<\/td><td><code>\u0414\u0435 \u0456 \u043a\u043e\u043b\u0438?<\/code><\/td><td><code>De i koly?<\/code><\/td><\/tr><tr><td>Sk\u00f3re<\/td><td colspan=\"2\"><code>BLEU = 100.00<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n<\/div>\n\n<div>\n<em>Segment 1096.<\/em><figure class=\"wp-block-table is-style-stripes\"><table><tbody><tr><td>Anglick\u00fd zdroj<\/td><td colspan=\"2\"><code>We were there for three months.<\/code><\/td><\/tr><tr><td>Lemko referen\u010dn\u00fd text a transliter\u00e1cia<\/td><td><code>\u0422\u0430\u043c \u0437\u043c\u0435 \u0431\u044b\u043b\u0438 \u0442\u0440\u0438 \u043c\u0456\u0441\u044f\u0446\u0456.<\/code><\/td><td><code>Tam zme b\u0177ly try misiaci.<\/code><\/td><\/tr><tr><td>Ukrajinsk\u00e1 hypot\u00e9za a transliter\u00e1cia<\/td><td><code>\u041c\u0438 \u0431\u0443\u043b\u0438 \u0442\u0430\u043c \u0442\u0440\u0438 \u043c\u0456\u0441\u044f\u0446\u0456.<\/code><\/td><td><code>My buly tam try misjaci.<\/code><\/td><\/tr><tr><td>Sk\u00f3re<\/td><td colspan=\"2\"><code>BLEU = 30.21<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n<\/div>\n\n<div>\n<em>Segment 2513.<\/em><figure class=\"wp-block-table is-style-stripes\"><table><tbody><tr><td>Anglick\u00fd zdroj<\/td><td colspan=\"2\"><code>Well, here to the west.<\/code><\/td><\/tr><tr><td>Lemko referen\u010dn\u00fd text a transliter\u00e1cia<\/td><td><code>\u041d\u043e \u0442\u043e \u0442\u0443 \u043d\u0430 \u0437\u0430\u0445\u0456\u0434.<\/code><\/td><td><code>No to tu na zachid.<\/code><\/td><\/tr><tr><td>Ukrajinsk\u00e1 hypot\u00e9za a transliter\u00e1cia<\/td><td><code>\u041d\u0443, \u0442\u0443\u0442 \u043d\u0430 \u0437\u0430\u0445\u0456\u0434.<\/code><\/td><td><code>Nu, tut na zachid.<\/code><\/td><\/tr><tr><td>Sk\u00f3re<\/td><td colspan=\"2\"><code>BLEU = 30.21<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n<\/div>\n\n<p class=\"wp-block-paragraph\"><strong>Ru\u0161tina<\/strong>. Slu\u017eba Google Translate z angli\u010dtiny do ru\u0161tiny dosiahla sk\u00f3re BLEU na \u00farovni korpusu 1,10. <\/p>\n\n<div>\n<em>Segment 432.<\/em><figure class=\"wp-block-table is-style-stripes\"><table><tbody><tr><td>Anglick\u00fd zdroj<\/td><td colspan=\"2\"><code>Nobody knew.<\/code><\/td><\/tr><tr><td>Lemko referen\u010dn\u00fd text a transliter\u00e1cia<\/td><td><code>\u041d\u0438\u0445\u0442\u043e \u043d\u0435 \u0437\u043d\u0430\u043b.<\/code><\/td><td><code>Nychto ne znal.<\/code><\/td><\/tr><tr><td>Rusk\u00e1 hypot\u00e9za a transliter\u00e1cia<\/td><td><code>\u041d\u0438\u043a\u0442\u043e \u043d\u0435 \u0437\u043d\u0430\u043b.<\/code><\/td><td><code>Nikto ne znal.<\/code><\/td><\/tr><tr><td>Sk\u00f3re<\/td><td colspan=\"2\"><code>BLEU = 59.46<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n<\/div>\n\n<div>\n<em>Segment 2751.<\/em><figure class=\"wp-block-table is-style-stripes\"><table><tbody><tr><td>Anglick\u00fd zdroj<\/td><td colspan=\"2\"><code>What did they expel us for?<\/code><\/td><\/tr><tr><td>Lemko referen\u010dn\u00fd text a transliter\u00e1cia<\/td><td><code>\u0417\u0430 \u0448\u0442\u043e \u043d\u0430\u0441 \u0432\u044b\u0433\u043d\u0430\u043b\u0438?<\/code><\/td><td><code>Za \u0161to nas v\u0177hnaly?<\/code><\/td><\/tr><tr><td>Rusk\u00e1 hypot\u00e9za a transliter\u00e1cia<\/td><td><code>\u0417\u0430 \u0447\u0442\u043e \u043d\u0430\u0441 \u0432\u044b\u0433\u043d\u0430\u043b\u0438?<\/code><\/td><td><code>Za \u010dto nas vygnali?<\/code><\/td><\/tr><tr><td>Sk\u00f3re<\/td><td colspan=\"2\"><code>BLEU = 42.73<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n<\/div>\n\n<div>\n<em>Segment 2164.<\/em><figure class=\"wp-block-table is-style-stripes\"><table><tbody><tr><td>Anglick\u00fd zdroj<\/td><td colspan=\"2\"><code>Brother went off to war.<\/code><\/td><\/tr><tr><td>Lemko referen\u010dn\u00fd text a transliter\u00e1cia<\/td><td><code>\u0411\u0440\u0430\u0442 \u043f\u0456\u0448\u043e\u043b \u043d\u0430 \u0432\u043e\u0439\u043d\u0443.<\/code><\/td><td><code>Brat pi\u0161ol na vojnu.<\/code><\/td><\/tr><tr><td>Rusk\u00e1 hypot\u00e9za a transliter\u00e1cia<\/td><td><code>\u0411\u0440\u0430\u0442 \u0443\u0448\u0435\u043b \u043d\u0430 \u0432\u043e\u0439\u043d\u0443.<\/code><\/td><td><code>Brat u\u0161el na vojnu.<\/code><\/td><\/tr><tr><td>Sk\u00f3re<\/td><td colspan=\"2\"><code>BLEU = 42.73<\/code><\/td><\/tr><\/tbody><\/table><\/figure>\n<\/div>\n\n<h2 class=\"wp-block-heading\" id=\"h-4-discussion\">4 Diskusia<\/h2>\n\n<p class=\"wp-block-paragraph\">Sk\u00f3re BLEU na \u00farovni korpusu pre prekladov\u00fd syst\u00e9m Lemko 6,28 nazna\u010duje, \u017ee hoci je e\u0161te ve\u013ea pr\u00e1ce, veci s\u00fa na spr\u00e1vnej ceste. \u0160tandardn\u00e9 rusk\u00e9 sk\u00f3re BLEU 1,10 nazna\u010duje, \u017ee Lemko je menej podobn\u00e9 ru\u0161tine ako po\u013e\u0161tine (BLEU 1,70). Mo\u017eno by pou\u017eitie predrevolu\u010dnej ortografie mohlo zv\u00fd\u0161i\u0165 sk\u00f3re ru\u0161tiny, ale to by bol drah\u00fd experiment s mal\u00fdm zjavn\u00fdm pr\u00ednosom.  <\/p>\n\n<p class=\"wp-block-paragraph\">Transliterovan\u00e9 \u0161tandardn\u00e9 po\u013esk\u00e9 kontroln\u00e9 sk\u00f3re podobnosti BLEU 1,70 nazna\u010duje men\u0161ie ru\u0161enie zo strany dominantn\u00e9ho jazyka v Po\u013esku, ne\u017e by sa dalo o\u010dak\u00e1va\u0165. Bolo by zauj\u00edmav\u00e9 prepracova\u0165 experiment, kde by sa na po\u013e\u0161tinu aplikovalo nieko\u013eko v\u00fdpo\u010dtovo nen\u00e1ro\u010dn\u00fdch a zjavn\u00fdch zvukov\u00fdch kore\u0161pondenci\u00ed (napr\u00edklad denazaliz\u00e1cia *\u0119 na \/ja\/ a *\u01eb na \/u\/, retrakcia *i na \/y\/ a zmena *g na \/h\/ <a href=\"#ref-23\" id=\"cite-23\">[23]<\/a>), aby sa zistilo, \u010di by potom dosiahla vy\u0161\u0161ie sk\u00f3re ako \u0161tandardn\u00e1 ukrajin\u010dina. <\/p>\n\n<p class=\"wp-block-paragraph\">Zhrnutie: Lemko bolo syntetizovan\u00e9 v laborat\u00f3riu a mo\u017enos\u0165 jeho produkcie bola dan\u00e1 do r\u00fak nov\u00fdm aj roden\u00fdm hovorcom. Po d\u00f4kladnej gener\u00e1lnej oprave motora a roz\u0161\u00edren\u00ed glos\u00e1ra je \u010fal\u0161\u00edm krokom objekt\u00edvne zmera\u0165 a, ak je to mo\u017en\u00e9, necha\u0165 hovorcami subjekt\u00edvne ohodnoti\u0165 kvalitu syntetick\u00e9ho Lemko v porovnan\u00ed s t\u00fdm, ktor\u00e9 produkuj\u00fa roden\u00ed hovorcovia. De\u0148, ke\u010f nov\u00ed hovorcovia jazykov s n\u00edzkymi zdrojmi m\u00f4\u017eu pou\u017ei\u0165 strojov\u00fd preklad na to, aby za\u010dali komunikova\u0165 vo svojom jazyku cez noc, je bli\u017e\u0161ie, rovnako ako de\u0148, ke\u010f sa jazyk Lemko pripoj\u00ed k radom t\u00fdch, ktor\u00e9 boli predt\u00fdm ohrozen\u00e9, ale teraz s\u00fa revitalizovan\u00e9.  <\/p>\n\n<p class=\"wp-block-paragraph\"><strong>Po\u010fakovanie.<\/strong> R\u00e1d by som po\u010fakoval svojmu kolegovi Mingovi Qianovi z Peraton Labs za in\u0161pir\u00e1ciu k uskuto\u010dneniu tohto experimentu a Brianovi Stensrudovi zo Soar Technology, Inc. za to, \u017ee n\u00e1s predstavil, ako aj za jeho povzbudenie.<\/p>\n\n<p class=\"wp-block-paragraph\">Taktie\u017e by som r\u00e1d po\u010fakoval svojej priate\u013eke Corinne Caudill za jej povzbudenie a osobn\u00fd z\u00e1ujem o projekt, ako aj za to, \u017ee ma predstavila prezidentke Karpatsko-rus\u00ednskej spolo\u010dnosti Maryann Sivak z University of Pittsburgh, ktorej by som r\u00e1d po\u010fakoval za pr\u00edle\u017eitos\u0165 prezentova\u0165 moju pr\u00e1cu.<\/p>\n\n<p class=\"wp-block-paragraph\">Taktie\u017e by som r\u00e1d po\u010fakoval Marii Silvestri z nad\u00e1cie John and Helen Timo Foundation za uskuto\u010dnenie rozhovorov s roden\u00fdmi hovorcami Lemko a darovanie prepisov a mojich prekladov na v\u00fdskum a v\u00fdvoj.<\/p>\n\n<p class=\"wp-block-paragraph\">R\u00e1d by som po\u010fakoval Achimovi Rabusovi z Univerzity vo Freiburgu a Yvesovi Scherrerovi z Helsinskej univerzity za ich z\u00e1ujem o projekt a n\u00e1pady.<\/p>\n\n<p class=\"wp-block-paragraph\">Taktie\u017e by som r\u00e1d po\u010fakoval Myhal&#8217;ovi L\u0177\u017ee\u010dkovi z blogu o technol\u00f3gi\u00e1ch men\u0161inov\u00fdch jazykov InterFyisa za jeho skor\u00fd z\u00e1ujem o projekt a komunitn\u00fa osvetu.<\/p>\n\n<p class=\"wp-block-paragraph\">Taktie\u017e by som r\u00e1d po\u010fakoval kolegovi, rod\u00e1kovi zo Zahoczewie, Markovi \u0141yszykovi za jeho z\u00e1ujem o projekt a komunitn\u00fa osvetu.<\/p>\n\n<p class=\"wp-block-paragraph\">Na z\u00e1ver by som r\u00e1d po\u010fakoval svojmu spoluautorovi a kolegovi z Antech Systems Inc. Tomovi Dobrymu za jeho povzbudenie a vedenie.<\/p>\n\n<h2 class=\"wp-block-heading\" id=\"h-references\">Referencie<\/h2>\n\n<p class=\"wp-block-paragraph\" id=\"ref-1\">1. <a href=\"#cite-1\">^<\/a> Graddol, D.: Bud\u00facnos\u0165 jazyka. Science, 303(5662), 1329-1331 (2004). <a href=\"https:\/\/doi.org\/10.1126\/science.1096546\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">https:\/\/doi.org\/10.1126\/science.1096546<\/a><\/p>\n\n<p class=\"wp-block-paragraph\" id=\"ref-2\">2. <a href=\"#cite-2\">^<\/a> Eberhard, D. M., Simons, G. F., &amp; Fennig, C. D.: Ethnologue: Jazyky sveta, SIL International. Dvadsiate \u0161tvrt\u00e9 vydanie. SIL International, Dallas (2021). Online verzia: Ko\u013eko jazykov je ohrozen\u00fdch?, <a href=\"https:\/\/www.ethnologue.com\/guides\/how-many-languages-endangered\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">https:\/\/www.ethnologue.com\/guides\/how-many-languages-endangered<\/a>, naposledy pr\u00edstupn\u00e9 11. 2. 2022.<\/p>\n\n<p class=\"wp-block-paragraph\" id=\"ref-3\">3. <a href=\"#cite-3\">^<\/a> K\u00f3dov\u00e9 tabu\u013eky ISO 639, <a href=\"https:\/\/iso639-3.sil.org\/code_tables\/639\/data\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">https:\/\/iso639-3.sil.org\/code_tables\/639\/data<\/a>, naposledy pr\u00edstupn\u00e9 11. 2. 2022.<\/p>\n\n<p class=\"wp-block-paragraph\" id=\"ref-4\">4. <a href=\"#cite-4\">^<\/a> Jazykov\u00e1 podpora, <a href=\"https:\/\/cloud.google.com\/translate\/docs\/languages\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">https:\/\/cloud.google.com\/translate\/docs\/languages<\/a>, naposledy pr\u00edstupn\u00e9 11. 2. 2022.<\/p>\n\n<p class=\"wp-block-paragraph\" id=\"ref-5\">5. <a href=\"#cite-5\">^<\/a> Vybra\u0165 jazyk, <a href=\"https:\/\/m.facebook.com\/language.php\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">https:\/\/m.facebook.com\/language.php<\/a>, naposledy pr\u00edstupn\u00e9 11. 2. 2022.<\/p>\n\n<p class=\"wp-block-paragraph\" id=\"ref-6\">6. <a href=\"#cite-6-0\">^<\/a> <a href=\"#cite-6\">^<\/a> Orynycz, P., Dobry, T., Jackson, A., &amp; Litzenberg, K.: \u00c1no, hovor\u00edm\u2026 Neur\u00f3nov\u00fd strojov\u00fd preklad AI vo viacjazy\u010dnom tr\u00e9ningu. In: Zborn\u00edk pr\u00edspevkov z konferencie Interservice\/Industry Training, Simulation, and Education Conference (I\/ITSEC) 2021, pr\u00edspevok \u010d. 21176. National Training and Simulation Association, Orlando (2021). <a href=\"https:\/\/www.xcdsystem.com\/iitsec\/proceedings\/index.cfm?Year=2021&amp;AbID=96953&amp;CID=862\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">https:\/\/www.xcdsystem.com\/iitsec\/proceedings\/index.cfm?Year=2021&amp;AbID=96953&amp;CID=862<\/a><\/p>\n\n<p class=\"wp-block-paragraph\" id=\"ref-7\">7. <a href=\"#cite-7\">^<\/a> Du\u0107-Fajfer, O.: <em>Literatura a proces rozwoju i rewitalizacja to\u017csamo\u015bci j\u0119zykowej na przyk\u0142adzie literatury \u0142emkowskiej<\/em>. In: Olko, J., Wicherkiewicz, T., Borges, R. (eds.), Integrovan\u00e9 strat\u00e9gie pre revitaliz\u00e1ciu jazyka, str. 175\u2013200. Prv\u00e9 vydanie. Fakulta \u201eArtes Liberales\u201c, Var\u0161avsk\u00e1 univerzita, Var\u0161ava (2016).   <\/p>\n\n<p class=\"wp-block-paragraph\" id=\"ref-8\">8. <a href=\"#cite-8\">^<\/a> Scherrer, Y., Rabus, A.: Neur\u00f3nov\u00e9 morfosyntaktick\u00e9 zna\u010dkovanie pre rus\u00edn\u010dinu. In: Mitkov, R., Tait, J., Boguraev, B. (eds.), Natural Language Engineering, 25(5), 633\u2013650. Cambridge University Press, Cambridge (2019). <a href=\"https:\/\/doi.org\/10.1017\/S1351324919000287\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">https:\/\/doi.org\/10.1017\/S1351324919000287<\/a><\/p>\n\n<p class=\"wp-block-paragraph\" id=\"ref-9\">9. <a href=\"#cite-9\">^<\/a> V\u00fdhrady a vyhl\u00e1senia k Zmluve \u010d. 148 \u2013 Eur\u00f3pska charta region\u00e1lnych alebo men\u0161inov\u00fdch jazykov (ETS \u010d. 148), <a href=\"https:\/\/www.coe.int\/en\/web\/conventions\/full-list?module=declarations-by-treaty&amp;numSte=148&amp;codeNature=1&amp;codePays=POL,%20last%20accessed%202022\/02\/11\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">https:\/\/www.coe.int\/en\/web\/conventions\/full-list?module=declarations-by-treaty&amp;numSte=148&amp;codeNature=1&amp;codePays=POL, naposledy pr\u00edstupn\u00e9 11. 2. 2022<\/a>.<\/p>\n\n<p class=\"wp-block-paragraph\" id=\"ref-10\">10. <a href=\"#cite-10\">^<\/a> <em>Formularz indywidualny<\/em>, <a href=\"https:\/\/stat.gov.pl\/download\/gfx\/portalinformacyjny\/pl\/defaultstronaopisowa\/5781\/1\/1\/nsp_2011_badanie__pelne_wykaz_pytan.pdf\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">https:\/\/stat.gov.pl\/download\/gfx\/portalinformacyjny\/pl\/defaultstronaopisowa\/5781\/1\/1\/nsp_2011_badanie__pelne_wykaz_pytan.pdf<\/a>, naposledy pr\u00edstupn\u00e9 11. 2. 2022.<\/p>\n\n<p class=\"wp-block-paragraph\" id=\"ref-11\">11. <a href=\"#cite-11\">^<\/a> <em>Narodowy Spis Powszechny Ludno\u015bci i Mieszka\u0144 2002 r. z 20 maja (formularz A)<\/em> <a href=\"https:\/\/stat.gov.pl\/gfx\/portalinformacyjny\/userfiles\/_public\/spisy_powszechne\/nsp2002-form-a.pdf\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">https:\/\/stat.gov.pl\/gfx\/portalinformacyjny\/userfiles\/_public\/spisy_powszechne\/nsp2002-form-a.pdf<\/a>, naposledy pr\u00edstupn\u00e9 11. 2. 2022.<\/p>\n\n<p class=\"wp-block-paragraph\" id=\"ref-12\">12. <a href=\"#cite-12\">^<\/a> <em>IV Raport dotycz\u0105cy sytuacji mniejszo\u015bci narodowych i etnicznych oraz j\u0119zyka regionalnego w Rzeczypospolitej Polskiej \u2013 2013<\/em>, <a href=\"http:\/\/mniejszosci.narodowe.mswia.gov.pl\/download\/86\/14637\/TekstIVRaportu.pdf\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">http:\/\/mniejszosci.narodowe.mswia.gov.pl\/download\/86\/14637\/TekstIVRaportu.pdf<\/a>, naposledy pr\u00edstupn\u00e9 11. 2. 2022.<\/p>\n\n<p class=\"wp-block-paragraph\" id=\"ref-13\">13. <a href=\"#cite-13\">^<\/a> Va\u0148ko, J.: Jazyk slovensk\u00fdch Rus\u00ednov. East European Monographs, New York (2000).<\/p>\n\n<p class=\"wp-block-paragraph\" id=\"ref-14\">14. <a href=\"#cite-14\">^<\/a> Forston, B., IV: Indoeur\u00f3psky jazyk a kult\u00fara. Blackwell Publishing, Oxford (2004). <\/p>\n\n<p class=\"wp-block-paragraph\" id=\"ref-15\">15. <a href=\"#cite-15-0\">^<\/a> <a href=\"#cite-15-1\">^<\/a> Pokorny, J.: <em>Indogermanisches etymologisches W\u00f6rterbuch<\/em>, Bern, 1959.<\/p>\n\n<p class=\"wp-block-paragraph\" id=\"ref-16\">16. <a href=\"#cite-16\">^<\/a> Horoszczak, J.: <em>S\u0142ownik \u0142emkowsko-polski, polsko-\u0142emkowski<\/em>. Rutenika, Var\u0161ava (2004). <\/p>\n\n<p class=\"wp-block-paragraph\" id=\"ref-17\">17. <a href=\"#cite-17-0\">^<\/a> <a href=\"#cite-17-1\">^<\/a> <a href=\"#cite-17-2\">^<\/a> <a href=\"#cite-17-3\">^<\/a> Vasmer, M. <em>Russisches etymologisches W\u00f6rterbuch<\/em>. <em>Zweiter Band<\/em>. Carl Winter, Universit\u00e4tsverlag, Heidelberg (1955). <\/p>\n\n<p class=\"wp-block-paragraph\" id=\"ref-18\">18. <a href=\"#cite-18\">^<\/a> Monier-Williams, M.: Sanskrt-anglick\u00fd slovn\u00edk etymologicky a filologicky usporiadan\u00fd so zvl\u00e1\u0161tnym zrete\u013eom na pr\u00edbuzn\u00e9 indoeur\u00f3pske jazyky, The Clarendon Press, Oxford (1899).<\/p>\n\n<p class=\"wp-block-paragraph\" id=\"ref-19\">19. <a href=\"#cite-19\">^<\/a> Derksen, R.: Etymologick\u00fd slovn\u00edk slovanskej zdedenej lexiky. In: Lubotsky, A. (ed.) Leiden Indo-European Etymological Dictionary Series, vol. 4, Koninklijke Brill, Leiden (2008).<\/p>\n\n<p class=\"wp-block-paragraph\" id=\"ref-20\">20. <a href=\"#cite-20\">^<\/a> Post, M.: V\u00fdzva na jasnos\u0165 pri uv\u00e1dzan\u00ed sk\u00f3re BLEU. In: Zborn\u00edk pr\u00edspevkov z Tretej konferencie o strojovom preklade (WMT), vol. 1, str. 186\u2013191. Association for Computational Linguistics, Brusel (2018). <a href=\"https:\/\/aclanthology.org\/W18-63\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">https:\/\/aclanthology.org\/W18-63<\/a><\/p>\n\n<p class=\"wp-block-paragraph\" id=\"ref-21\">21. <a href=\"#cite-21\">^<\/a> Chen B., Cherry, C.: Systematick\u00e9 porovnanie vyhladzovac\u00edch techn\u00edk pre BLEU na \u00farovni viet. In: Zborn\u00edk pr\u00edspevkov z Deviateho workshopu o \u0161tatistickom strojovom preklade, str. 362\u2013367. Association for Computational Linguistics, Baltimore (2014). <a href=\"http:\/\/dx.doi.org\/10.3115\/v1\/W14-33\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">http:\/\/dx.doi.org\/10.3115\/v1\/W14-33<\/a><\/p>\n\n<p class=\"wp-block-paragraph\" id=\"ref-22\">22. <a href=\"#cite-22\">^<\/a> Ministerstvo vn\u00fatra a administrat\u00edvy: <em>Rozporz\u0105dzenie Ministra Spraw Wewn\u0119trznych i Administracji z dnia 30 maja 2005 r. w sprawie sposobu transliteracji imion i nazwisk os\u00f3b nale\u017c\u0105cych do mniejszo\u015bci narodowych i etnicznych zapisanych w alfabecie innym ni\u017c alfabet \u0142aci\u0144ski<\/em>. In: Dziennik Ustaw \u010d. 102, str. 6560\u20136573. Rz\u0105dowe Centrum Legislacji, Var\u0161ava (2005).  <\/p>\n\n<p class=\"wp-block-paragraph\" id=\"ref-23\">23. <a href=\"#cite-23\">^<\/a> Shevelov, G.: O chronol\u00f3gii H a nov\u00e9ho G v ukrajin\u010dine. In: Harvard Ukrainian Studies, vol. 1, \u010d. 2, str. 137\u2013152. Harvard Ukrainian Research Institute, Cambridge (1977). <a href=\"https:\/\/www.jstor.org\/stable\/40999942\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">https:\/\/www.jstor.org\/stable\/40999942<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Abstrakt Neur\u00f3nov\u00fd strojov\u00fd preklad poh\u00e1\u0148an\u00fd umelou inteligenciou by mohol \u010doskoro o\u017eivi\u0165 ohrozen\u00e9 jazyky t\u00fdm, \u017ee umo\u017en\u00ed nov\u00fdm hovorcom komunikova\u0165 v re\u00e1lnom \u010dase pomocou viet, ktor\u00e9 s\u00fa kvantitat\u00edvne bli\u017e\u0161ie k liter\u00e1rnej norme ako vety roden\u00fdch hovorcov, a to u\u017e od prv\u00e9ho d\u0148a ich cesty k obnove jazyka. Zatia\u013e \u010do Silicon Valley investuje obrovsk\u00e9 zdroje do technol\u00f3gie [&hellip;]<\/p>\n","protected":false},"author":2,"featured_media":11113,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"inline_featured_image":false,"footnotes":""},"categories":[445],"tags":[392,427,428],"class_list":["post-11109","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-studia","tag-lemko","tag-neuronovy-strojovy-preklad-nmt","tag-revitalizacia-jazyka"],"yoast_head":"<!-- This site is optimized with the Yoast SEO Premium plugin v27.3 (Yoast SEO v27.6) - https:\/\/yoast.com\/product\/yoast-seo-premium-wordpress\/ -->\n<title>Say It Right: Lemko AI neur\u00f3nov\u00fd strojov\u00fd preklad<\/title>\n<meta name=\"description\" content=\"Pr\u00e1ca o tom, ako som vytvoril preklada\u010d, ktor\u00fd pom\u00e1ha nov\u00fdm hovorcom o\u017eivi\u0165 Lemko\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.orynycz.com\/sk\/veda\/studia\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\/\" \/>\n<meta property=\"og:locale\" content=\"sk_SK\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Say It Right: Lemko AI neur\u00f3nov\u00fd strojov\u00fd preklad\" \/>\n<meta property=\"og:description\" content=\"Pr\u00e1ca o tom, ako som vytvoril preklada\u010d, ktor\u00fd pom\u00e1ha nov\u00fdm hovorcom o\u017eivi\u0165 Lemko\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.orynycz.com\/sk\/veda\/studia\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\/\" \/>\n<meta property=\"og:site_name\" content=\"Petro Orynycz\" \/>\n<meta property=\"article:publisher\" content=\"https:\/\/www.facebook.com\/orynycz\" \/>\n<meta property=\"article:author\" content=\"https:\/\/www.facebook.com\/orynycz\" \/>\n<meta property=\"article:published_time\" content=\"2022-06-26T12:00:45+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2026-01-23T06:13:52+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/www.orynycz.com\/wp-content\/uploads\/2022\/02\/2022-say-fi-en-v00.jpg\" \/>\n\t<meta property=\"og:image:width\" content=\"1200\" \/>\n\t<meta property=\"og:image:height\" content=\"628\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/jpeg\" \/>\n<meta name=\"author\" content=\"Admin\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:title\" content=\"Say It Right: Lemko AI neur\u00f3nov\u00fd strojov\u00fd preklad\" \/>\n<meta name=\"twitter:description\" content=\"Pr\u00e1ca o tom, ako som vytvoril preklada\u010d, ktor\u00fd pom\u00e1ha nov\u00fdm hovorcom o\u017eivi\u0165 Lemko.\" \/>\n<meta name=\"twitter:creator\" content=\"@OrynyczP\" \/>\n<meta name=\"twitter:label1\" content=\"Autor\" \/>\n\t<meta name=\"twitter:data1\" content=\"Admin\" \/>\n\t<meta name=\"twitter:label2\" content=\"Predpokladan\u00fd \u010das \u010d\u00edtania\" \/>\n\t<meta name=\"twitter:data2\" content=\"21 min\u00fat\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/veda\\\/studia\\\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/veda\\\/studia\\\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\\\/\"},\"author\":{\"name\":\"Admin\",\"@id\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/#\\\/schema\\\/person\\\/81acf6b0d8344b8d8832f55a0e4a9f63\"},\"headline\":\"Say It Right: Umel\u00fd preklad neur\u00f3nov\u00fdch strojov posil\u0148uje nov\u00fdch hovorcov na o\u017eivenie Lemko (2022)\",\"datePublished\":\"2022-06-26T12:00:45+00:00\",\"dateModified\":\"2026-01-23T06:13:52+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/veda\\\/studia\\\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\\\/\"},\"wordCount\":3888,\"commentCount\":0,\"publisher\":{\"@id\":\"https:\\\/\\\/www.orynycz.com\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/veda\\\/studia\\\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/www.orynycz.com\\\/wp-content\\\/uploads\\\/2022\\\/02\\\/2022-say-fi-en-v00.jpg\",\"keywords\":[\"Lemko\",\"Neur\u00f3nov\u00fd strojov\u00fd preklad (NMT)\",\"Revitaliz\u00e1cia jazyka\"],\"articleSection\":[\"Recenzovan\u00e9 vedeck\u00e9 pr\u00e1ce\"],\"inLanguage\":\"sk-SK\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/veda\\\/studia\\\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\\\/#respond\"]}],\"copyrightYear\":\"2022\",\"copyrightHolder\":{\"@id\":\"https:\\\/\\\/www.orynycz.com\\\/#organization\"},\"accessibilityFeature\":[\"tableOfContents\"]},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/veda\\\/studia\\\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\\\/\",\"url\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/veda\\\/studia\\\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\\\/\",\"name\":\"Say It Right: Lemko AI neur\u00f3nov\u00fd strojov\u00fd preklad\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/veda\\\/studia\\\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/veda\\\/studia\\\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/www.orynycz.com\\\/wp-content\\\/uploads\\\/2022\\\/02\\\/2022-say-fi-en-v00.jpg\",\"datePublished\":\"2022-06-26T12:00:45+00:00\",\"dateModified\":\"2026-01-23T06:13:52+00:00\",\"description\":\"Pr\u00e1ca o tom, ako som vytvoril preklada\u010d, ktor\u00fd pom\u00e1ha nov\u00fdm hovorcom o\u017eivi\u0165 Lemko\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/veda\\\/studia\\\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\\\/#breadcrumb\"},\"inLanguage\":\"sk-SK\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/veda\\\/studia\\\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"sk-SK\",\"@id\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/veda\\\/studia\\\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\\\/#primaryimage\",\"url\":\"https:\\\/\\\/www.orynycz.com\\\/wp-content\\\/uploads\\\/2022\\\/02\\\/2022-say-fi-en-v00.jpg\",\"contentUrl\":\"https:\\\/\\\/www.orynycz.com\\\/wp-content\\\/uploads\\\/2022\\\/02\\\/2022-say-fi-en-v00.jpg\",\"width\":1200,\"height\":628,\"caption\":\"Povedzte to spr\u00e1vne v Lemko s neur\u00f3nov\u00fdm prekladom poh\u00e1\u0148an\u00fdm AI.\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/veda\\\/studia\\\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Domov\",\"item\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Veda\",\"item\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/moc\\\/veda\\\/\"},{\"@type\":\"ListItem\",\"position\":3,\"name\":\"Recenzovan\u00e9 vedeck\u00e9 pr\u00e1ce\",\"item\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/moc\\\/veda\\\/studia\\\/\"},{\"@type\":\"ListItem\",\"position\":4,\"name\":\"Say It Right: Umel\u00fd preklad neur\u00f3nov\u00fdch strojov posil\u0148uje nov\u00fdch hovorcov na o\u017eivenie Lemko (2022)\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/#website\",\"url\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/\",\"name\":\"Orynycz.com\",\"description\":\"Vedec. In\u017einier umelej inteligencie. Lingvista.\",\"publisher\":{\"@id\":\"https:\\\/\\\/www.orynycz.com\\\/#organization\"},\"alternateName\":\"\u041e\u0440\u0438\u043d\u0438\u0447.com\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"sk-SK\"},{\"@type\":\"Organization\",\"@id\":\"https:\\\/\\\/www.orynycz.com\\\/#organization\",\"name\":\"orynycz.com\",\"alternateName\":\"\u041e\u0440\u0438\u043d\u0438\u0447.com\",\"url\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"sk-SK\",\"@id\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/#\\\/schema\\\/logo\\\/image\\\/\",\"url\":\"https:\\\/\\\/www.orynycz.com\\\/wp-content\\\/uploads\\\/2025\\\/12\\\/logo-1.jpg\",\"contentUrl\":\"https:\\\/\\\/www.orynycz.com\\\/wp-content\\\/uploads\\\/2025\\\/12\\\/logo-1.jpg\",\"width\":512,\"height\":512,\"caption\":\"orynycz.com\"},\"image\":{\"@id\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/#\\\/schema\\\/logo\\\/image\\\/\"},\"sameAs\":[\"https:\\\/\\\/www.facebook.com\\\/orynycz\"],\"publishingPrinciples\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/zasady\\\/\",\"ownershipFundingInfo\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/zasady\\\/\",\"actionableFeedbackPolicy\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/zasady\\\/\",\"correctionsPolicy\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/zasady\\\/\",\"ethicsPolicy\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/zasady\\\/\",\"diversityPolicy\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/zasady\\\/\",\"diversityStaffingReport\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/zasady\\\/\"},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/www.orynycz.com\\\/sk\\\/#\\\/schema\\\/person\\\/81acf6b0d8344b8d8832f55a0e4a9f63\",\"name\":\"Admin\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"sk-SK\",\"@id\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/9a38941f48011a0de6d533516cefcfcbff0b865d9bbce556ca1778430b8139cf?s=96&d=initials&r=pg&initials=p\",\"url\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/9a38941f48011a0de6d533516cefcfcbff0b865d9bbce556ca1778430b8139cf?s=96&d=initials&r=pg&initials=p\",\"contentUrl\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/9a38941f48011a0de6d533516cefcfcbff0b865d9bbce556ca1778430b8139cf?s=96&d=initials&r=pg&initials=p\",\"caption\":\"Admin\"},\"description\":\"I am a scientist (ORCID iD 0000-0003-3094-9156), software engineer, computational linguist, localization and natural language engineer, Slavist, and Silicon Valley consultant. His research currently focuses on artificial intelligence (AI), neural machine translation (NMT), and hybrid systems to revitalize endangered, indigenous languages like Lemko. He received his degree in Russian at the Institute of East Slavic Philology of Jagiellonian University in Cracow, Poland, where he worked for Google amid its 2016 neural machine translation artificial intelligence breakthrough. His engines were recently mentioned in the Cambridge University Press journal Natural Language Engineering (Volume 25, Issue 5, page 634). Mr. Orynycz also has two decades of transatlantic experience as a linguist specializing in Russian, Polish, Ukrainian, Rusyn and Lemko for top language service providers, national defense, heavy industry, Raytheon, Amazon, Siemens, Mercedes-Benz, Daimler, investigators, philanthropists, and scientists.\",\"sameAs\":[\"https:\\\/\\\/www.orynycz.com\",\"https:\\\/\\\/www.facebook.com\\\/orynycz\",\"https:\\\/\\\/www.linkedin.com\\\/in\\\/orynycz\\\/\",\"https:\\\/\\\/x.com\\\/OrynyczP\",\"https:\\\/\\\/www.youtube.com\\\/channel\\\/UCOBzL010xr3XfzJcEZbaZyQ\"],\"gender\":\"male\",\"award\":[\"Engines mentioned in the Cambridge University Press journal Natural Language Engineering (Volume 25\",\"Issue 5\",\"page 634)\"],\"knowsAbout\":[\"Natural Language Processing\",\"Machine Translation\"],\"knowsLanguage\":[\"Ukrainian\",\"Lemko\",\"English\",\"Polish\",\"Russian\",\"Hungarian\"],\"jobTitle\":\"scientist\",\"worksFor\":\"Orynycz.com\"}]}<\/script>\n<!-- \/ Yoast SEO Premium plugin. -->","yoast_head_json":{"title":"Say It Right: Lemko AI neur\u00f3nov\u00fd strojov\u00fd preklad","description":"Pr\u00e1ca o tom, ako som vytvoril preklada\u010d, ktor\u00fd pom\u00e1ha nov\u00fdm hovorcom o\u017eivi\u0165 Lemko","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/www.orynycz.com\/sk\/veda\/studia\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\/","og_locale":"sk_SK","og_type":"article","og_title":"Say It Right: Lemko AI neur\u00f3nov\u00fd strojov\u00fd preklad","og_description":"Pr\u00e1ca o tom, ako som vytvoril preklada\u010d, ktor\u00fd pom\u00e1ha nov\u00fdm hovorcom o\u017eivi\u0165 Lemko","og_url":"https:\/\/www.orynycz.com\/sk\/veda\/studia\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\/","og_site_name":"Petro Orynycz","article_publisher":"https:\/\/www.facebook.com\/orynycz","article_author":"https:\/\/www.facebook.com\/orynycz","article_published_time":"2022-06-26T12:00:45+00:00","article_modified_time":"2026-01-23T06:13:52+00:00","og_image":[{"width":1200,"height":628,"url":"https:\/\/www.orynycz.com\/wp-content\/uploads\/2022\/02\/2022-say-fi-en-v00.jpg","type":"image\/jpeg"}],"author":"Admin","twitter_card":"summary_large_image","twitter_title":"Say It Right: Lemko AI neur\u00f3nov\u00fd strojov\u00fd preklad","twitter_description":"Pr\u00e1ca o tom, ako som vytvoril preklada\u010d, ktor\u00fd pom\u00e1ha nov\u00fdm hovorcom o\u017eivi\u0165 Lemko.","twitter_creator":"@OrynyczP","twitter_misc":{"Autor":"Admin","Predpokladan\u00fd \u010das \u010d\u00edtania":"21 min\u00fat"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/www.orynycz.com\/sk\/veda\/studia\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\/#article","isPartOf":{"@id":"https:\/\/www.orynycz.com\/sk\/veda\/studia\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\/"},"author":{"name":"Admin","@id":"https:\/\/www.orynycz.com\/sk\/#\/schema\/person\/81acf6b0d8344b8d8832f55a0e4a9f63"},"headline":"Say It Right: Umel\u00fd preklad neur\u00f3nov\u00fdch strojov posil\u0148uje nov\u00fdch hovorcov na o\u017eivenie Lemko (2022)","datePublished":"2022-06-26T12:00:45+00:00","dateModified":"2026-01-23T06:13:52+00:00","mainEntityOfPage":{"@id":"https:\/\/www.orynycz.com\/sk\/veda\/studia\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\/"},"wordCount":3888,"commentCount":0,"publisher":{"@id":"https:\/\/www.orynycz.com\/#organization"},"image":{"@id":"https:\/\/www.orynycz.com\/sk\/veda\/studia\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\/#primaryimage"},"thumbnailUrl":"https:\/\/www.orynycz.com\/wp-content\/uploads\/2022\/02\/2022-say-fi-en-v00.jpg","keywords":["Lemko","Neur\u00f3nov\u00fd strojov\u00fd preklad (NMT)","Revitaliz\u00e1cia jazyka"],"articleSection":["Recenzovan\u00e9 vedeck\u00e9 pr\u00e1ce"],"inLanguage":"sk-SK","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/www.orynycz.com\/sk\/veda\/studia\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\/#respond"]}],"copyrightYear":"2022","copyrightHolder":{"@id":"https:\/\/www.orynycz.com\/#organization"},"accessibilityFeature":["tableOfContents"]},{"@type":"WebPage","@id":"https:\/\/www.orynycz.com\/sk\/veda\/studia\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\/","url":"https:\/\/www.orynycz.com\/sk\/veda\/studia\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\/","name":"Say It Right: Lemko AI neur\u00f3nov\u00fd strojov\u00fd preklad","isPartOf":{"@id":"https:\/\/www.orynycz.com\/sk\/#website"},"primaryImageOfPage":{"@id":"https:\/\/www.orynycz.com\/sk\/veda\/studia\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\/#primaryimage"},"image":{"@id":"https:\/\/www.orynycz.com\/sk\/veda\/studia\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\/#primaryimage"},"thumbnailUrl":"https:\/\/www.orynycz.com\/wp-content\/uploads\/2022\/02\/2022-say-fi-en-v00.jpg","datePublished":"2022-06-26T12:00:45+00:00","dateModified":"2026-01-23T06:13:52+00:00","description":"Pr\u00e1ca o tom, ako som vytvoril preklada\u010d, ktor\u00fd pom\u00e1ha nov\u00fdm hovorcom o\u017eivi\u0165 Lemko","breadcrumb":{"@id":"https:\/\/www.orynycz.com\/sk\/veda\/studia\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\/#breadcrumb"},"inLanguage":"sk-SK","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.orynycz.com\/sk\/veda\/studia\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\/"]}]},{"@type":"ImageObject","inLanguage":"sk-SK","@id":"https:\/\/www.orynycz.com\/sk\/veda\/studia\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\/#primaryimage","url":"https:\/\/www.orynycz.com\/wp-content\/uploads\/2022\/02\/2022-say-fi-en-v00.jpg","contentUrl":"https:\/\/www.orynycz.com\/wp-content\/uploads\/2022\/02\/2022-say-fi-en-v00.jpg","width":1200,"height":628,"caption":"Povedzte to spr\u00e1vne v Lemko s neur\u00f3nov\u00fdm prekladom poh\u00e1\u0148an\u00fdm AI."},{"@type":"BreadcrumbList","@id":"https:\/\/www.orynycz.com\/sk\/veda\/studia\/say-it-right-umely-preklad-neuronovych-strojov-posilnuje-novych-hovorcov-na-ozivenie-lemko-2022\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Domov","item":"https:\/\/www.orynycz.com\/sk\/"},{"@type":"ListItem","position":2,"name":"Veda","item":"https:\/\/www.orynycz.com\/sk\/moc\/veda\/"},{"@type":"ListItem","position":3,"name":"Recenzovan\u00e9 vedeck\u00e9 pr\u00e1ce","item":"https:\/\/www.orynycz.com\/sk\/moc\/veda\/studia\/"},{"@type":"ListItem","position":4,"name":"Say It Right: Umel\u00fd preklad neur\u00f3nov\u00fdch strojov posil\u0148uje nov\u00fdch hovorcov na o\u017eivenie Lemko (2022)"}]},{"@type":"WebSite","@id":"https:\/\/www.orynycz.com\/sk\/#website","url":"https:\/\/www.orynycz.com\/sk\/","name":"Orynycz.com","description":"Vedec. In\u017einier umelej inteligencie. Lingvista.","publisher":{"@id":"https:\/\/www.orynycz.com\/#organization"},"alternateName":"\u041e\u0440\u0438\u043d\u0438\u0447.com","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.orynycz.com\/sk\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"sk-SK"},{"@type":"Organization","@id":"https:\/\/www.orynycz.com\/#organization","name":"orynycz.com","alternateName":"\u041e\u0440\u0438\u043d\u0438\u0447.com","url":"https:\/\/www.orynycz.com\/sk\/","logo":{"@type":"ImageObject","inLanguage":"sk-SK","@id":"https:\/\/www.orynycz.com\/sk\/#\/schema\/logo\/image\/","url":"https:\/\/www.orynycz.com\/wp-content\/uploads\/2025\/12\/logo-1.jpg","contentUrl":"https:\/\/www.orynycz.com\/wp-content\/uploads\/2025\/12\/logo-1.jpg","width":512,"height":512,"caption":"orynycz.com"},"image":{"@id":"https:\/\/www.orynycz.com\/sk\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/www.facebook.com\/orynycz"],"publishingPrinciples":"https:\/\/www.orynycz.com\/sk\/zasady\/","ownershipFundingInfo":"https:\/\/www.orynycz.com\/sk\/zasady\/","actionableFeedbackPolicy":"https:\/\/www.orynycz.com\/sk\/zasady\/","correctionsPolicy":"https:\/\/www.orynycz.com\/sk\/zasady\/","ethicsPolicy":"https:\/\/www.orynycz.com\/sk\/zasady\/","diversityPolicy":"https:\/\/www.orynycz.com\/sk\/zasady\/","diversityStaffingReport":"https:\/\/www.orynycz.com\/sk\/zasady\/"},{"@type":"Person","@id":"https:\/\/www.orynycz.com\/sk\/#\/schema\/person\/81acf6b0d8344b8d8832f55a0e4a9f63","name":"Admin","image":{"@type":"ImageObject","inLanguage":"sk-SK","@id":"https:\/\/secure.gravatar.com\/avatar\/9a38941f48011a0de6d533516cefcfcbff0b865d9bbce556ca1778430b8139cf?s=96&d=initials&r=pg&initials=p","url":"https:\/\/secure.gravatar.com\/avatar\/9a38941f48011a0de6d533516cefcfcbff0b865d9bbce556ca1778430b8139cf?s=96&d=initials&r=pg&initials=p","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/9a38941f48011a0de6d533516cefcfcbff0b865d9bbce556ca1778430b8139cf?s=96&d=initials&r=pg&initials=p","caption":"Admin"},"description":"I am a scientist (ORCID iD 0000-0003-3094-9156), software engineer, computational linguist, localization and natural language engineer, Slavist, and Silicon Valley consultant. His research currently focuses on artificial intelligence (AI), neural machine translation (NMT), and hybrid systems to revitalize endangered, indigenous languages like Lemko. He received his degree in Russian at the Institute of East Slavic Philology of Jagiellonian University in Cracow, Poland, where he worked for Google amid its 2016 neural machine translation artificial intelligence breakthrough. His engines were recently mentioned in the Cambridge University Press journal Natural Language Engineering (Volume 25, Issue 5, page 634). Mr. Orynycz also has two decades of transatlantic experience as a linguist specializing in Russian, Polish, Ukrainian, Rusyn and Lemko for top language service providers, national defense, heavy industry, Raytheon, Amazon, Siemens, Mercedes-Benz, Daimler, investigators, philanthropists, and scientists.","sameAs":["https:\/\/www.orynycz.com","https:\/\/www.facebook.com\/orynycz","https:\/\/www.linkedin.com\/in\/orynycz\/","https:\/\/x.com\/OrynyczP","https:\/\/www.youtube.com\/channel\/UCOBzL010xr3XfzJcEZbaZyQ"],"gender":"male","award":["Engines mentioned in the Cambridge University Press journal Natural Language Engineering (Volume 25","Issue 5","page 634)"],"knowsAbout":["Natural Language Processing","Machine Translation"],"knowsLanguage":["Ukrainian","Lemko","English","Polish","Russian","Hungarian"],"jobTitle":"scientist","worksFor":"Orynycz.com"}]}},"_links":{"self":[{"href":"https:\/\/www.orynycz.com\/sk\/wp-json\/wp\/v2\/posts\/11109","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.orynycz.com\/sk\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.orynycz.com\/sk\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.orynycz.com\/sk\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.orynycz.com\/sk\/wp-json\/wp\/v2\/comments?post=11109"}],"version-history":[{"count":1,"href":"https:\/\/www.orynycz.com\/sk\/wp-json\/wp\/v2\/posts\/11109\/revisions"}],"predecessor-version":[{"id":11119,"href":"https:\/\/www.orynycz.com\/sk\/wp-json\/wp\/v2\/posts\/11109\/revisions\/11119"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.orynycz.com\/sk\/wp-json\/wp\/v2\/media\/11113"}],"wp:attachment":[{"href":"https:\/\/www.orynycz.com\/sk\/wp-json\/wp\/v2\/media?parent=11109"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.orynycz.com\/sk\/wp-json\/wp\/v2\/categories?post=11109"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.orynycz.com\/sk\/wp-json\/wp\/v2\/tags?post=11109"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}