Eyona Ezintsha UGoogle usebenzisa ubuGcisa bobuGcisa ukwenza utshintsho olukhulu kwisixhobo sakhe sokuGuqulela

UGoogle usebenzisa ubuGcisa bobuGcisa ukwenza utshintsho olukhulu kwisixhobo sakhe sokuGuqulela

Akukho zinguqulelo zimbi.I-Pixabay



Oomatshini bokufundisa ukuze baqonde ngokwenyani ulwimi lwendalo kuye kwaba yenye yemiceli mngeni ejongene nososayensi abasebenza ukuqhubela phambili ubukrelekrele bokulinganisa. Kodwa uGoogle wenze inkqubela phambili yokwenyani ekwenzeni ukuba iikhompyuter zijonge kulwimi njengebhegi yamagama, kwaye ezi nkqubela phambili ngoku zingena kwimveliso yazo.

Iinguqulelo zikaGoogle, umzekelo, ufumana ubuchwephesha obungeniso kunye nokwaziswa kweNeural Machine Translation (NMT). Ukuqala namhlanje, uyakuqaphela ukuphuculwa okukhulu kwazo naziphi na iinguqulelo ngesiHindi, isiRashiya kunye nesiVietnam. Oku kulandela ukuqala kokusebenzisa i-NMT kwiToliki ka-Novemba ophelileyo, xa isiNgesi, isiFrentshi, isiJamani, iSpanish, isiPhuthukezi, isiTshayina, isiJapan, isiKorea nesiTurkey bonke babone ukuphucuka okufanayo.

Sineelwimi ezili-103 zizonke, kwaye injongo yethu kukuba zonke zisebenze ngeminatha ye-neural, isithethi sikaGoogle sixelele uMbonisi. Wongeze ukukhutshwa kweelwimi ezishiyekileyo kuyakwenzeka ngaphezulu kweenyanga ezininzi, kodwa ixesha elichanekileyo alaziwa kuba uGoogle umisela nje naliphi na ixesha xa ekwazi ukugqwesa inkqubo yangoku. Ngamanye amaxesha oku kuya kuba mbalwa ngaxeshanye, njengokungeniswa kwanamhlanje kokuphuculwa kwesiHindi, isiRashiya kunye neVietnam.

Inguqulelo yakudala xa ithelekiswa.uphando



Itoliki kaGoogle ihlala iluncedo kodwa ehhh iyonke. Ungayisebenzisa ukufumana ukuqonda ukuba yintoni enye into ethetha olunye ulwimi, kodwa nantoni na engaphezulu kwebinzana elilula ayinakuba yinguqulelo echanekileyo. Kodwa ngale ndlela intsha, iinguqulelo kuKhangelo lukaGoogle, translate.google.com, iiapp zikaGoogle kwaye, ekugqibeleni, ukuguqulelwa kwamaphepha okuzenzekelayo kwiChannel kuya kuba ngcono kwaye ekugqibeleni kubonise ulwimi lwendalo.

Ukuguqulelwa kwe-Neural kungcono kakhulu kunetekhnoloji yethu yangaphambili kuba siguqulela izivakalisi zonke ngexesha, endaweni yeziqwenga zesivakalisi, u-Barak Turovsky, okhokelayo kwimveliso kuToliko lukaGoogle, wabhala kwiposti yebhlog evakalisa iindaba.

Ngaphambili, uGoogle wayethembele kuNguqulelo loMatshini oluSekwe kumatshini (i-PBMT), eyaphula isivakalisi esifakelweyo kumagama nakwimabinzana anokuguqulelwa ngokuzimeleyo. I-NMT entsha, nangona kunjalo, isithatha sonke isivakalisi njengegalelo kwaye isiguqulela njengesinye. I-NMT isebenzisa uthungelwano olunzulu lwe-neural, olwenza ukuba ikhompyuter iqonde iimeko ezingazibonanga ngaphambili ngokufunda, ixesha elongezelelekileyo, kolunye ulwazi. Kule meko, olunye uqeqesho luseta inkqubo olusebenzisayo ukufunda idatha evela kwifayile ye- Indawo yoToliko kaGoogle , apho abasebenzisi bemihla ngemihla abasuka kwihlabathi liphela baguqulela izivakalisi kwiilwimi zabo kwaye bathelekise neenguqulelo.

Ngelixa zonke iilwimi zingayi kwenza utshintsho kwiinyanga ezininzi, ibhetshi elandelayo kulindeleke kwiiveki ezimbalwa.



Amanqaku Umdla