Kusintha kwa Mozilla Common Voice 7.0

NVIDIA ndi Mozilla atulutsa zosintha zamaseti awo a Common Voice, omwe akuphatikiza zitsanzo zolankhula za anthu 182, kukwera 25% kuchokera miyezi 6 yapitayo. Zambiri zimasindikizidwa ngati gulu la anthu (CC0). Ma seti omwe akuganiziridwawo atha kugwiritsidwa ntchito pamakina ophunzirira kuti apange kuzindikira kwamawu ndi mitundu yophatikizika.

Poyerekeza ndi zosintha zam'mbuyomu, kukula kwa zolankhulidwa zomwe zasonkhanitsidwa zawonjezeka kuchokera ku 9 mpaka 13.9 maola masauzande olankhula. Chiwerengero cha zilankhulo zothandizidwa chawonjezeka kuchokera ku 60 mpaka 76, kuphatikiza kwa nthawi yoyamba kuthandizira zilankhulo za Chibelarusi, Kazakh, Uzbek, Chibugariya, Chiameniya, Chiazabajani ndi Bashkir. Kukonzekera kwa chinenero cha Chirasha kumakhudza anthu 2136 ndi maola 173 a zolankhula (anali 1412 ndi maola 111), ndi chinenero cha Chiyukireniya - 615 ndi maola 66 (anali 459 ndi maola 30).

Anthu opitilira 75 adatenga nawo gawo pokonzekera zida zachingerezi, kulamula maola 2637 akulankhula kotsimikizika (anali 66 otenga nawo gawo ndi maola 1686). Chochititsa chidwi n'chakuti chinenero chomwe chili m'malo achiwiri malinga ndi kuchuluka kwa deta yomwe yasonkhanitsidwa ndi Rwanda, yomwe maola a 2260 asonkhanitsidwa. Izi zikutsatiridwa ndi German (1040), Catalan (920) ndi Esperanto (840). Zina zomwe zikuchulukirachulukira kukula kwamawu ndi chilankhulo cha Thai (kuwonjezeka kowirikiza kawiri m'munsi, kuchokera pa maola 20 mpaka 12), Luganda (kuchokera maola 250 mpaka 8), Esperanto (kuchokera maola 80 mpaka 100) ndi Chitamil ( kuyambira maola 840 mpaka 24).

Monga gawo lakuchita nawo pulojekiti ya Common Voice, NVIDIA inakonza zitsanzo zophunzitsidwa zokonzekera zamakina ophunzirira makina (othandizidwa ndi PyTorch) kutengera zomwe zasonkhanitsidwa. Mitunduyi imagawidwa ngati gawo la zida zaulere komanso zotseguka za NVIDIA NeMo, zomwe, mwachitsanzo, zimagwiritsidwa ntchito kale pamawu omvera a MTS ndi Sberbank. Zitsanzozi zimapangidwira kuti zigwiritsidwe ntchito pozindikira malankhulidwe, kaphatikizidwe ka mawu, ndi kachitidwe ka chilankhulo chachilengedwe, ndipo zitha kukhala zothandiza kwa ofufuza omwe amamanga makina olumikizirana ndi mawu, nsanja zolembera, ndi malo oimbira mafoni. Mosiyana ndi mapulojekiti omwe analipo kale, zitsanzo zosindikizidwa sizongozindikirika ndi Chingerezi komanso zilankhulo zosiyanasiyana, kamvekedwe ka mawu ndi kalankhulidwe kosiyanasiyana.

Tikukumbutseni kuti pulojekiti ya Common Voice ndi cholinga chokonzekera ntchito yophatikizana kuti ipeze mndandanda wa machitidwe a mawu omwe amaganizira za kusiyanasiyana kwa mawu ndi kalembedwe. Ogwiritsa ntchito amapemphedwa kuti azilankhula mawu omwe akuwonetsedwa pazenera kapena kuwunika kuchuluka kwa data yomwe yawonjezeredwa ndi ogwiritsa ntchito ena. Dongosolo lankhokwe losanjidwa lomwe lili ndi katchulidwe kosiyanasiyana ka mawu amunthu amatha kugwiritsidwa ntchito popanda zoletsa pamakina ophunzirira pamakina ndi ntchito zofufuza.

Malinga ndi mlembi wa laibulale yozindikiritsa mawu yopitilira ya Vosk, zovuta za Common Voice set ndi mbali imodzi ya mawu (kuchuluka kwa amuna azaka za 20-30, komanso kusowa kwa zinthu ndi mawu a azimayi. .

Source: opennet.ru

Kuwonjezera ndemanga