I-Amazon ipapashe phantsi kwe-CC BY 4.0 ilayisenisi ye-"MASSIVE" (Iilwimi ezininzi ze-Amazon SLURP yokuzaliswa kwe-Slot, uHlelo lweNjongo, kunye noVavanyo lwe-Virtual-assistant) dataset, imodeli yeenkqubo zokufunda ngomatshini, kunye nezixhobo zokuqeqesha iimodeli zakho ezinokuthi zisetyenziswe ukuqonda ulwazi ngolwimi lwendalo (NLU, Natural Language Understanding). Olu luhlu luquka amagama angaphezu kwesigidi aneenkcazo namagama ahleliweyo alungiselelwe iilwimi ezingama-51.
Ingqokelela ye-SLURP, eyayifumaneka ngesiNgesi, yasetyenziswa njengesalathiso sokwakha iseti ye-MASSIVE, eyathi yahlala kwezinye iilwimi ezingama-50 kusetyenziswa abaguquleli abaziingcali. I-Alexa yokuqonda ulwimi lwendalo (i-NLU) iteknoloji iqala ukuguqula intetho kwisicatshulwa, emva koko isebenzise iimodeli ezininzi ze-NLU kwisicatshulwa esihlalutya ubukho bamagama angundoqo ukugqiba undoqo wombuzo womsebenzisi.
Enye yeenjongo zokwenza kunye nokupapasha iseti kukulungelelanisa abancedisi belizwi ukucubungula ulwazi ngeelwimi ezininzi ngaxeshanye, kunye nokukhuthaza abaphuhlisi beqela lesithathu ukuba benze usetyenziso kunye neenkonzo ezandisa izakhono zabancedisi belizwi. Ukutsala umdla wabaphuhlisi, iAmazon yaqalisa ukhuphiswano lokwenza eyona modeli igqwesileyo yegeneric isebenzisa iseti yedatha epapashiweyo.
Okwangoku, abancedisi belizwi baxhasa kuphela iilwimi ezimbalwa kwaye basebenzise iimodeli zokufunda ngomatshini ezibotshelelwe kulwimi oluthile. Iprojekthi ye-MASSIVE ijolise ekupheliseni le ntsilelo ngokudala imifuziselo yehlabathi jikelele kunye neenkqubo zokufunda ngoomatshini ezikwaziyo ukwahlula kunye nokucubungula ulwazi ngeelwimi ezahlukeneyo.
umthombo: opennet.ru