Kudala sazi ukuthi elinye lamaphuzu lapho iGoogle ibonakala inentshisekelo enkulu yokwethula uhlelo lwayo lwe- ukuhlakanipha okufakelwayo kungokubili ezinjinini zakho zokusesha nakuhumushi lwakho. Ngenxa yalokhu futhi ngenxa yentuthuko enkulu etholakele, evela kwaGoogle, emasontweni ambalwa edlule, basitshele ukuthi bahlanganisa uhlelo lwe-neural network enkonzweni yabo yokuhumusha ulimi, yona ebinikeza imiphumela efanelekile.
Manje, ngemuva kwaso sonke lesi sikhathi sokuqalisa nokuhlola, sifunda ukuthi lolu hlelo alukwazi kuphela ukuhumusha noma yiluphi ulimi ngejubane nangokunemba okufanele ukunconywa, kepha manje selukwazi dala ulimi lwakho. Konke lokhu kudalwe yindlela yokusebenza kwensiza uqobo lwayo lapho, kuze kube manje, isiNgisi besisetshenziswa njengebhuloho eliphakathi lezilimi ezinjengesiJapane nesiKorea, ulimi oluphakathi oluthi, njengesivivinyo, lwasuswa ku-algorithm yilabo ababhekene nomsebenzi.
I-Google ithola ukuthi uhlelo lwayo lobuhlakani bokuzenzela lukwazile ukudala ulimi lwayo.
Umphumela walolu vivinyo wawuhlukile impela ngoba uhlelo lokuhumusha lwe-Google ngokusobala lwadala i-interlingua, okuwulimi lwayo lokuzenzela ukuze kuqedelwe ukuhumusha. Ngeshwa, futhi ngaphandle kokuthi abadali bohlelo lokuhumusha kweGoogle bakholelwa ngokuqinisekile ukuthi ipulatifomu izakhele ulimi lwayo, okwamanje, akunakwenzeka ukwazi okuningi ngalolu limi lokuzenzela ngoba izinqubo zangaphakathi ezisetshenziswa amanethiwekhi we-neural kunzima kakhulu ukuzichaza.
Njengoba kunengqondo, ngalezi zindaba kube nabaningi abasungule izinsimbi endizeni, ngaphandle kokuqhubeka, ukusuka I-TechCrunch, baqinisekisa ukuthi lolu hlobo lwenqubekela phambili ngeke lwenze lutho ngaphandle kokukhombisa lokho namuhla imishini ingahle ikhulume yodwa singanakile futhi, njengoba kunjalo, uma sakuqaphela, asikwazanga ukukuqonda abakushoyo.
Ukwaziswa okwengeziwe: I-Google Research Blog
Noma ngubani ofuna ukwenza lokho