Mozilla Common Voice 7.0 inosvika neanopfuura maawa zviuru gumi nematatu ezwi rezwi

Munguva pfupi yapfuura NVIDIA neMozilla vakazivisa kuburitswa kweshanduro nyowani ye "Mozilla Common Voice 7.0" iyo inomiririra anopfuura maawa zviuru gumi nematatu ezwi rezwi rekubatana mavambo uye kuwedzerwa kweimwe mitauro gumi nematanhatu uye izvo uchienzaniswa neyekupedzisira kugadzirisa, saizi yeiyo vhoriyamu yezvinhu achitaura muunganidzwa yakawedzera neinenge 50% zvimwe.

Uyewo, mitauro yakatsigirwa yakawedzera kubva pa60 kusvika pa76, kusanganisira yekuwedzera rutsigiro rweBelarusian, Kazakh, Uzbek, Bulgarian, Armenian, Azerbaijani, uye mitauro yeBashkir kekutanga.

Kune avo vasina kujaira Common Voice, vanofanirwa kuziva kuti eIyi yakavhurika dhata izwi data data chikuru pasirese uye chakagadzirirwa kuita demokrasi izwi tekinoroji. Iyo inoshandiswa nevanoongorora, vadzidzi uye vanogadzira kubva kumativi ose enyika.

Vashandi vanounganidza matunhu avo kuti vape data rezwi kune dhatabhesi yeveruzhinji yeMCV, iyo chero munhu anogona kushandisa kudzidzisa tekinoroji-inogoneswa tekinoroji. Sechikamu chekushandira pamwe kweNVIDIA cpaMozilla Common Voice, mamodheru akadzidziswa mune izvi uye mamwe eruzhinji data seti anowanikwa mahara kuburikidza neakavhurika sosi yekushandisa inonzi NVIDIA NeMo.

Iyo purojekiti inotarisira kuronga basa rekubatana kuunganidza dhatabhesi remazwi matemplate, uchifunga nezve akasiyana manzwi uye nzira dzekutaura. Iyo yakaunganidzwa dhatabhesi ine marekodhi emataurirwo akasiyana emitsara akajairwa ekutaura kwevanhu anogona kushandiswa pasina zvirambidzo mumasisitimu ekudzidza emuchina uye mumapurojekiti ekutsvaga.

Sekureva kwemunyori weVosk inoenderera mberi yekuziva raibhurari, zvikanganiso zveCommon Voice zvakagadziriswa ndeimwe-sidedness yezvinhu zvekutaura (kunyanya kwevarume vane makore makumi maviri nemakumi matatu uye kushomeka kwezvinhu nezwi revakadzi, vana nevakuru), kushomeka kweshanduko yemazwi (kudzokororwa kwemitsara imwe chete) uye kugoverwa kwe MP20 marekodhi anokanganisa kukanganisa.

Nezve iyo nyowani vhezheni yeCommunity Voice 7.0

Mune iyi vhezheni itsva vanhu vanopfuura zviuru makumi manomwe nezvishanu vakatora chikamu mukugadzirira kwezvinhu muChirungu, kuraira maawa mazana maviri nemakumi manomwe neshanu ekutaura kwakasimbiswa (paive nezviuru mazana matanhatu zvevatori vechikamu nemaawa 2637).

Zvakare sekutaura kwatamboita pakutanga, iyi vhezheni itsva inosuma mitauro mitsva gumi nemitanhatu kupinda muCommunity Voice dhataset yemitauro inosvika makumi manomwe nenomwe, mitauro mishanu yepamusoro nemaawa ese chiRungu (maawa mazana maviri nemazana matanhatu nemakumi matatu), Kinyarwanda (76), ChiGerman (2.630), chiCatalan (2.260) neEsperanto (1.040).

Mitauro yakawedzera zvakanyanya muzana muzana chiThai (kanokwana kuita kakapetwa makumi maviri, kubva pamaawa gumi nemaviri kusvika pa20 maawa), luganda (12-fold kukura, kubva maawa masere kusvika kumaawa makumi masere), Esiperando (kukura kanopfuura ka7, kubva pamaawa zana kusvika maawa mazana masere nemakumi mana) uye Tamil (kukura kweanopfuura masere masere, kubva pamaawa makumi maviri nemana kusvika pa8). Sezvineiwo, U Rwanda rwoherereye ibiri ku rwego rw'ubukungu, iyo maawa makumi maviri nemakumi maviri akaunganidzwa. Ivo vanoteverwa neGerman (2260), Catalan (1040) uye Esperanto (920). Iyo dhataset ikozvino yaonesa pamusoro pe840 yakasarudzika manzwi, iyo 182,000% kukura munharaunda yemubhadhari mumwedzi mitanhatu chete.

Izvo zvinotaurwa zvakare kuti sechikamu chekutora chikamu kwavo muchirongwa, NVIDIA yakagadzirira yakagadzirira-kushandisa-yakadzidziswa mamodheru emachina ekufunda masisitimu zvichibva pane yakaunganidzwa data (inoenderana nePyTorch). Iwo mamodheru anogoverwa sechikamu chemahara uye chakavhurika cheNVIDIA NeMo chishandiso, icho, semuenzaniso, chatove chishandiswa mumashandisi ezwi anoshanda eMTS neSberbank.

Mhando idzi inotarisirwa kuzivikanwa kwekutaura, synthesis yekutaura uye masisitimu ekugadzirisa ruzivo mumutauro wechisikigo uye zvinogona kubatsira kune vanoongorora mukugadzirwa kwenzwi dialog masisitimu, zvinyorwa zvekunyora, uye otomatiki mafoni nzvimbo. Kusiyana nemapurojekiti akange aripo, mhando dzakaburitswa hadzina kungogumira pakuzivikanwa kweChirungu uye dzinovhara mitauro yakasiyana siyana, maaccents uye mafomu ekutaura.

Finalmente kana iwe uchifarira kuziva zvakawanda nezvazvo, unogona kutarisa ruzivo mu inotevera chinongedzo.


Siya yako yekutaura

Your kero e havazobvumirwi ichibudiswa. Raida minda anozivikanwa ne *

*

*

  1. Inotarisira data: AB Internet Networks 2008 SL
  2. Chinangwa cheiyo data: Kudzora SPAM, manejimendi manejimendi.
  3. Legitimation: Kubvuma kwako
  4. Kutaurirana kwedata
  5. Dhata yekuchengetedza: Dhatabhesi inobatwa neOccentus Networks (EU)
  6. Kodzero: Panguva ipi neipi iwe unogona kudzora, kupora uye kudzima ruzivo rwako