Wav2Vec 2.0
Wav2Vec 2.0 ni Meta AI yerekana imvugo yigenga yigenga yiga amajwi akomeye avuye mumajwi mbisi, atanditse.
Incamake
It matters because it slashed the amount of transcribed audio needed to build accurate speech recognizers, unlocking ASR for low-resource languages.
Kwibira cyane
Yatangijwe na Facebook (Meta) AI muri 2020, Wav2Vec 2.0 yakemuye ikibazo gikomeye mu kumenyekanisha imvugo: amajwi yanditseho ni make kandi ahenze, mu gihe amajwi mbisi ari menshi. Icyitegererezo kibanza kwitegura kumasaha ibihumbi yamagambo atanditse yiga kuzuza ibice byapfunditswe byerekana ibimenyetso, byubaka imbere imbere muburyo bwimiterere yimiterere. Gusa nyuma yaho, itunganijwe neza kumubare muto wamakuru yandukuwe. Icyamamare, hamwe niminota 10 gusa yanditseho amajwi wongeyeho nini-nini yo kwitegura, yageze ku gipimo cyamagambo yamagambo akoreshwa ku gipimo cya LibriSpeech. Iyi resept ya demokarasi ASR, ituma abantu bandika neza indimi n'imvugo idafite corpora nini isobanurwa.
Ubushishozi
Wav2Vec 2.0 igaburira imivumba mbisi ikoresheje kodegisi ya CNN igizwe na kodegisi nyinshi, hanyuma ikerekana masike yibisubizo byihishe. Transformer isoma ibipfukisho kandi igomba kwerekana neza ibipimo byerekana neza buri gice cyashizwe kumurongo uhereye kubirangaza, ukoresheje igihombo gitandukanye. Igitabo cyize cyanditse cyerekana amajwi akomeje mubice bitagira ingano byamagambo, bigatanga imirimo itandukanye igamije intego yo guhanura.
Ingaruka z'Ingamba
Kugera no kugera
Itezimbere kugerwaho binyuze mu kwandukura, kuvuga, no guhuza amajwi.
Igiciro na bije
Amatsinda yibitangazamakuru arashobora kohereza amajwi yihuse hamwe na bije nto.
Umuvuduko n'igipimo
Sisitemu ireba abakiriya irashobora gutunganya imikoranire ivugwa murwego runini.
Kazoza ka Wav2Vec 2.0
Wav2Vec 2.0 yabibye umuryango wose wubwoko bwigenga bwigenga hamwe na XLS-R mu ndimi nyinshi, zikoresha indimi 128. Uburyo burimo guhurira kuri kodegisi yimvugo yisi yose yimurwa kumenyekana, guhindurwa, gutahura amarangamutima, hamwe nakazi kavuga uhereye kumurongo umwe wabigenewe. Witege ko uzakomeza kunguka indimi zibangamiwe kandi zifite amikoro make, wongeyeho guhuza cyane kwijwi ryigenga ryigenga muri sisitemu ya multimodal ihuza ibitekerezo kumvugo, inyandiko, nibindi bimenyetso.
Gushyira mu bikorwa Isi
Kubaka imvugo imenyekanisha indimi nkeya hamwe niminota mike y amajwi yandukuwe
Gutegura amajwi ya kodegisi yisi yose nyuma ikurikiranwa neza kugirango uhamagare kuri terefone
Gukuramo imvugo yibiranga amarangamutima cyangwa sisitemu yo kumenya-kuvuga
Guha imbaraga indimi nyinshi XLS-R yerekana indimi 100+
Ingaruka & Kurinda
Gukoresha nabi amajwi no kwigira ibyago byiyongera mugihe uruhushya rubuze.
Ukuri kurashobora kugabanuka hejuru yimvugo, imvugo, cyangwa urusaku rwibidukikije.
Amajwi yubukorikori arashobora kwibeshya kumvugo yukuri nta kirango gisobanutse.
Igishushanyo mbonera
Shaka uruhushya rusobanutse rwo gufata amajwi, gukoroniza, no gukoresha.
Ikizamini cyiza mubiganiro bitandukanye hamwe nuburyo bwimbere.
Sobanura igihe umuntu agomba gusuzuma cyangwa kwemeza ibisubizo.
Andika amajwi yubukorikori kandi ugumane inyandiko zerekana kubazwa.
Komeza Ubushakashatsi
Free newsletter
Get the daily AI briefing
Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.
One email each weekday. Unsubscribe in one click. We never sell or share your address.
Test yourself
Take the Wav2Vec 2.0 quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
Ubuyobozi bukurikira
Vocoders
Ibibazo bikunze kubazwa
What is Wav2Vec 2.0?
Wav2Vec 2.0 ni Meta AI yerekana imvugo yigenga yigenga yiga amajwi akomeye uhereye kumajwi mbisi, atanditse. Nibyingenzi kuko yagabanije umubare wamajwi yandukuwe akenewe kugirango yubake imvugo imenya neza, ifungura ASR indimi nkeya.
Ni ikihe kibazo cy'ibanze mu kumenyekanisha imvugo Wav2Vec 2.0 yagenewe gukemura?
Wav2Vec 2.0 igabanya kwishingikiriza kumajwi ahenze yanditswemo mbere yo kuvuga kumagambo menshi atanditse.
Ni ubuhe buryo bwo kwiga Wav2Vec 2.0 ikoresha mugihe cyo kwitegura?
Ihisha ibice byamajwi yihishe kandi ikoresha igihombo gitandukanye kugirango uhitemo ibice bikwiye mubirangaza.
Nibihe bice byambere bitunganya imiterere mbisi muri Wav2Vec 2.0?
Ikirindiro cyibice byemeza kodegisi ya flim ya feri muburyo bwihishe mbere yo guhisha hamwe na Transformer.
Ni amajwi make yanditseho amajwi Wav2Vec 2.0 azwi cyane yari akeneye kugera kubintu byukuri kuri LibriSpeech?
Hamwe nini-nini yo kwitegura hiyongereyeho iminota 10 gusa yamakuru yanditseho, yageze ku buryo butangaje igipimo cyo hasi cyamakosa, cyerekana label ikora neza.
Ni uruhe ruhare rwigitabo cyize muri Wav2Vec 2.0?
Codebook igereranya guhora igereranya mubice bitagira ingano, itanga intego kubintu bitandukanye.