Mozilla Common Voice 7.0 Hloov tshiab

NVIDIA thiab Mozilla tau tshaj tawm qhov hloov tshiab rau lawv cov ntaub ntawv Cov Lus Sib Tham, uas suav nrog 182 tus neeg hais lus piv txwv, nce 25% los ntawm 6 lub hlis dhau los. Cov ntaub ntawv yog luam tawm raws li pej xeem sau (CC0). Cov txheej txheem npaj tuaj yeem siv rau hauv tshuab kev kawm los tsim kev paub txog kev hais lus thiab cov qauv tsim qauv.

Piv rau qhov hloov tshiab dhau los, qhov loj ntawm cov khoom siv hais lus hauv kev sau tau nce los ntawm 9 txog 13.9 txhiab teev ntawm kev hais lus. Tus naj npawb ntawm cov lus txhawb tau nce los ntawm 60 mus rau 76, suav nrog thawj zaug kev txhawb nqa rau Belarusian, Kazakh, Uzbek, Bulgarian, Armenian, Azerbaijani thiab Bashkir lus. Cov txheej txheem rau Lavxias teb sab lus suav nrog 2136 tus neeg koom thiab 173 teev ntawm kev hais lus (muaj 1412 tus neeg koom thiab 111 teev), thiab rau cov lus Ukrainian - 615 tus neeg koom thiab 66 teev (muaj 459 tus neeg koom thiab 30 teev).

Ntau tshaj 75 txhiab tus neeg tau koom nrog hauv kev npaj cov ntaub ntawv ua lus Askiv, dictating 2637 teev ntawm cov lus pom zoo (muaj 66 txhiab tus neeg koom thiab 1686 teev). Interestingly, cov lus nyob rau hauv qhov chaw thib ob nyob rau hauv cov nqe lus ntawm tus nqi ntawm cov ntaub ntawv sau yog Rwanda, uas 2260 teev tau sau. Qhov no yog ua raws li German (1040), Catalan (920) thiab Esperanto (840). Ntawm cov dynamically nce qhov loj ntawm lub suab ntaub ntawv yog cov lus Thaib (20-fold nce nyob rau hauv lub hauv paus, los ntawm 12 mus rau 250 teev), Luganda (los ntawm 8 mus rau 80 teev), Esperanto (los ntawm 100 mus rau 840 teev) thiab Tamil ( los ntawm 24 mus rau 220 teev).

Raws li ib feem ntawm nws txoj kev koom tes hauv Kev Ua Haujlwm Lub Suab, NVIDIA tau npaj cov qauv kev cob qhia npaj rau kev kawm tshuab (txhawb los ntawm PyTorch) raws li cov ntaub ntawv khaws tseg. Cov qauv tau muab faib ua ib feem ntawm cov khoom siv pub dawb thiab qhib NVIDIA NeMo, uas, piv txwv li, twb tau siv rau hauv cov kev pabcuam suab nrov ntawm MTS thiab Sberbank. Cov qauv no yog tsim los siv rau hauv kev paub txog kev hais lus, kev hais lus sib txuas lus, thiab cov txheej txheem ua lus ntuj, thiab tej zaum yuav muaj txiaj ntsig zoo rau cov kws tshawb fawb tsim lub suab sib tham, cov txheej txheem hloov pauv, thiab cov chaw hu xovtooj. Tsis zoo li tej yaam num yav dhau los, cov qauv luam tawm tsis txwv rau kev lees paub lus Askiv thiab npog ntau hom lus, accents thiab hom kev hais lus.

Cia peb nco ntsoov koj tias Lub Hom Phiaj Lub Suab yog txhawm rau txhim kho kev sib koom ua haujlwm txhawm rau txhawm rau khaws cov ntaub ntawv ntawm lub suab qauv uas coj mus rau hauv tus account ntau haiv neeg ntawm lub suab thiab kev hais lus. Cov neeg siv tau raug caw tuaj koom lub suab kab lus tso tawm ntawm qhov screen lossis ntsuas qhov zoo ntawm cov ntaub ntawv ntxiv los ntawm lwm tus neeg siv. Cov ntaub ntawv khaws tseg nrog cov ntaub ntawv ntawm ntau cov lus hais ntawm cov kab lus ntawm tib neeg hais lus tuaj yeem siv yam tsis muaj kev txwv hauv tshuab kev kawm thiab hauv cov haujlwm tshawb fawb.

Raws li tus kws sau ntawv ntawm lub tsev qiv ntawv Vosk kev lees paub kev hais lus tsis tu ncua, qhov tsis zoo ntawm Cov Lus Sib Tham yog qhov ib sab ntawm cov khoom siv suab (qhov tseem ceeb ntawm cov txiv neej hnub nyoog 20-30 xyoo, thiab tsis muaj cov khoom siv nrog lub suab ntawm cov poj niam. , cov menyuam yaus thiab cov neeg laus), qhov tsis muaj qhov hloov pauv hauv phau ntawv txhais lus (rov ua dua ntawm tib nqe lus) thiab kev faib tawm cov ntaub ntawv hauv MP3 distorting.

Tau qhov twg los: opennet.ru

Ntxiv ib saib