Isibuyekezo se-Mozilla Common Voice 8.0

I-Mozilla ikhiphe isibuyekezo kumadathasethi ayo e-Common Voice, afaka amasampula okuphimisa avela kubantu abacishe babe ngu-200. Idatha ishicilelwa njengesizinda somphakathi (CC0). Amasethi ahlongozwayo angasetshenziswa kumasistimu okufunda omshini ukuze akhe ukunakwa kwenkulumo namamodeli okuhlanganiswa. Uma kuqhathaniswa nokubuyekezwa kwangaphambilini, umthamo wezinto zokukhuluma eqoqweni lenyuke ngo-30% - ukusuka ku-13.9 kuya ku-18.2 wamahora ayizinkulungwane zokukhuluma. Inani lezilimi ezisekelwayo lenyuke lisuka ku-67 laya ku-87.

Isethi yolimi lwesiRashiya ihlanganisa abahlanganyeli abangu-2452 namahora angu-193 wezinto zokukhuluma (kwakukhona ababambiqhaza abangu-2136 namahora angu-173), ngolimi lwaseBelarusian - abahlanganyeli abangu-6160 namahora angu-987 (kwakunabahlanganyeli abangu-3831 namahora angu-356), ngolimi lwase-Ukraine - Abahlanganyeli abangama-684 namahora angama-76 (kwakunabahlanganyeli abangama-615 namahora angama-66). Abantu abangaphezu kwezinkulungwane ezingama-79 babambe iqhaza ekulungiseleleni izinto ngesiNgisi, besho amahora angama-2886 enkulumo eqinisekisiwe (kwakukhona ababambiqhaza abayizinkulungwane ezingama-75 namahora angama-2637).

Ake sikukhumbuze ukuthi iphrojekthi Yezwi Elivamile ihloselwe ukuhlela umsebenzi ohlanganyelwe ukuze kuqoqwe isizindalwazi samaphethini ezwi acabangela ukwehlukahlukana kwamazwi nezitayela zokukhuluma. Abasebenzisi bayamenywa ukuthi bakhulume imishwana yezwi eboniswe esikrinini noma bahlole ikhwalithi yedatha engezwe abanye abasebenzisi. Isizindalwazi esiqoqiwe esinamarekhodi okuphimisela okuhlukahlukene kwemishwana evamile yenkulumo yomuntu ingasetshenziswa ngaphandle kwemikhawulo ezinhlelweni zokufunda zomshini kanye namaphrojekthi ocwaningo. Ngokusho kombhali womtapo wolwazi weVosk oqhubekayo wokuqashelwa kwenkulumo, ububi besethi ye-Common Voice wuhlangothi olulodwa lwezwi lezwi (ubukhulu babantu besilisa abaneminyaka engama-20-30 ubudala, kanye nokuntuleka kwezinto ezinamazwi abesifazane. , izingane kanye nabantu abadala), ukuntuleka kokuhlukahluka kusichazamazwi (ukuphindaphindwa kwemisho efanayo) kanye nokusatshalaliswa kokurekhodiwe ngefomethi ye-MP3 ehlanekezelwe.

Ukwengeza, singaqaphela ukukhishwa kwekhithi yamathuluzi ye-NVIDIA NeMo 1.6, ehlinzeka ngezindlela zokufunda zomshini zokudala amasistimu okubona inkulumo, ukuhlanganiswa kwenkulumo nokucubungula ulimi lwemvelo. I-NeMo ihlanganisa amamodeli alungele ukusetshenziswa aqeqeshelwe amasistimu okufunda omshini asuselwe kuhlaka lwe-PyTorch, olulungiswe i-NVIDIA isebenzisa idatha yenkulumo Yezwi Elivamile futhi ihlanganisa izilimi ezihlukahlukene, amaphimbo nezindlela zokukhuluma. Amamodeli angase abe usizo kubacwaningi abakha amasistimu engxoxo asekelwe ezwini, izinkundla zokuloba, nezikhungo zezingcingo ezizenzakalelayo. Isibonelo, i-NVIDIA NeMo isetshenziswa ezinsizeni zezwi ezizenzakalelayo ze-MTS ne-Sberbank. Ikhodi ye-NeMo ibhalwe nge-Python kusetshenziswa i-PyTorch futhi isatshalaliswa ngaphansi kwelayisensi ye-Apache 2.0.

Source: opennet.ru

Engeza amazwana