Wav2Vec 2.0
Wav2Vec 2.0 bụ Meta Nlereanya okwu nke AI na-ahụ maka onwe ya nke na-amụta ihe ngosi ọdịyo siri ike site na ndekọ raw, enweghị akara.
Nchịkọta
It matters because it slashed the amount of transcribed audio needed to build accurate speech recognizers, unlocking ASR for low-resource languages.
Ime miri emi
Facebook webatara (Meta) AI n'afọ 2020, Wav2Vec 2.0 gbochiri mkpọmkpọ isi na njirimara okwu: ọdịyo akpọrọ dị ụkọ ma dị ọnụ, ebe ọdịyo dị ukwuu dị ukwuu. Ihe nlereanya ahụ na-ebu ụzọ zụọ ụzọ na puku kwuru puku awa nke okwu a na-edeghị aha site n'ịmụta imeju akụkụ nke mgbama ahụ, na-ewulite nghọta ime mmụọ bara ụba nke usoro ụda olu. Naanị mgbe nke ahụ gasịrị, a na-edobe ya nke ọma na ntakịrị data edegharịrị. N'ịbụ nke a ma ama, na-eji naanị nkeji 10 nke egwu egwu gbakwunyere ọzụzụ izizi buru ibu, ọ ruru ọnụego mperi okwu enwere ike iji na akara ngosi LibriSpeech. Ntụziaka a mebere ASR n'ọchịchị onye kwuo uche ya, na-eme ka ederede dị mma maka asụsụ na olumba na-enweghị nnukwu nkọwapụta.
Nghọta nka nka
Wav2Vec 2.0 na-azụ ụdị ebili mmiri sitere na njiri mkpuchi CNN multi-layer, wee kpuchie mkpuchi nke vectors latent na-esi na ya pụta. Onye ntụgharị na-agụ ihe mkpuchi mkpuchi ma ga-achọpụtarịrị ihe nnochite anya nha nke akụkụ ọ bụla nwere mkpuchi site na ihe ndọpụ uche, na-eji mfu dị iche. Akwụkwọ koodu amụtara na-ewepụ ọdịyo na-aga n'ihu ka ọ bụrụ nkeji nkeji okwu, na-enye ọrụ dị iche iche akọwapụtara nke ọma iji buo amụma.
Mmetụta atụmatụ
Nweta na iru
Ọ na-eme ka nnweta ya dịkwuo mma site na ndegharị, ịkọ akụkọ, na ntụgharị olu.
Ọnụ ego na mmefu ego
Ndị otu mgbasa ozi nwere ike ibubata ọdịyo a na-egbu maramara ngwa ngwa site na iji obere mmefu ego.
Ọsọ na ọnụ ọgụgụ
Sistemụ na-eche ihu ndị ahịa nwere ike hazie mkparịta ụka n'ọtụtụ buru ibu.
Ọdịnihu nke Wav2Vec 2.0
Wav2Vec 2.0 weputara ezinaụlọ dum nwere ụdị ikwu okwu na-elekọta onwe ya yana ọtụtụ asụsụ XLS-R, nke gbasara asụsụ 128. Ụzọ a na-agakọ na ihe ngbanwe okwu zuru ụwa ọnụ nke na-ebufe na njirimara, ntụgharị asụsụ, nchọpụta mmetụta, na ọrụ ọkà okwu site n'otu ebe a zụrụ azụ. Na-atụ anya uru na-aga n'ihu maka asụsụ ndị nọ n'ihe ize ndụ na nke dị ala, gbakwunyere njikọ siri ike nke njirimara ọdịyo na-elekọta onwe ya n'ime usoro multimodal na-ejikọta ọnụ maka okwu, ederede, na akara ndị ọzọ.
Mmejuputa n'ezie n'ụwa
Wulite ihe nghota okwu maka asụsụ ndị nwere obere ihe nwere naanị nkeji nke ọdịyo edegharịrị
Ịmalite ntinye koodu ọdịyo zuru ụwa ọnụ ka emechara nke ọma maka ndegharị oku ekwentị
Na-ewepụta atụmatụ okwu maka mmetụta mmetụta ma ọ bụ usoro nnabata ọkà okwu
Na-akwado ụdị XLS-R nke ọtụtụ asụsụ na-edegharị n'ofe asụsụ 100+
Ihe ize ndụ & okporo ụzọ nche
Iji olu eme ihe na ihe egwu mpụta ga-abawanye mgbe nkwenye na-efu.
Izi ezi nwere ike ịdaba n'ofe ụda olu, olumba ma ọ bụ gburugburu mkpọtụ.
Enwere ike imehie ọdịyo sịntetik dị ka ezigbo okwu na-enweghị akara doro anya.
Map mmejuputa
Nweta nkwenye doro anya maka ijide olu, imechi, na ijigharị.
Nwale ogo n'ofe ndị na-ekwu okwu dị iche iche yana ọnọdụ ndabere.
Kọwaa mgbe mmadụ ga-enyocha ma ọ bụ kwado nsonye.
Deba aha ọdịyo sịntetik ma debe ndekọ ihe ndekọ maka ịza ajụjụ.
Nọgide na-eme nchọpụta
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
Ntuziaka na-esote
Vocoders akwara
Ajụjụ a na-ajụkarị
What is Wav2Vec 2.0?
Wav2Vec 2.0 bụ Meta Nlereanya okwu nke AI na-ahụ maka onwe ya nke na-amụta ihe ngosi ọdịyo siri ike site na ndekọ raw, enweghị akara. Ọ dị mkpa n'ihi na ọ belatara ọnụọgụ ọdịyo edegharịrị achọrọ iji wuo ndị na-achọpụta okwu ziri ezi, na-emeghe ASR maka asụsụ ndị nwere obere akụrụngwa.
Kedu isi nsogbu dị na njirimara okwu e mere Wav2Vec 2.0 iji lebara ya anya?
Wav2Vec 2.0 na-ebelata ịdabere na ọdịyo edegharịrị dị oke ọnụ site n'ịzụ ụzọ n'ọtụtụ okwu enweghị aha.
Kedu ụdị ebumnuche mmụta Wav2Vec 2.0 na-eji n'oge ọzụzụ izizi?
Ọ na-ekpuchi obosara nke vectors audio latent ma na-eji mfu dị iche iche were were ọnụọgụ ọnụọgụgụ ziri ezi n'etiti ihe ndọpụ uche.
Kedu ihe na-ebu ụzọ na-ebu ụzọ na-ebu ụzọ na Wav2Vec 2.0?
Otutu n'ígwé nke convolutional na-etinye koodu nke raw waveform ka ọ bụrụ ihe na-eme ka ọ bụrụ ihe na-adịghị mma tupu ikpuchi ihe na ihe ntụgharị.
Obere ọdịyo akpọrọ obere ka Wav2Vec 2.0 kwesịrị ka ama ama iji ruo izi ezi enwere ike na LibriSpeech?
Site na ọzụzụ izizi buru ibu gbakwunyere naanị nkeji 10 nke data akpọrọ, ọ nwetara ọnụego mperi okwu dị ntakịrị, na-egosipụta arụmọrụ labelụ.
Kedu ọrụ nke akwụkwọ koodu amụtara na Wav2Vec 2.0?
Akwụkwọ koodu ahụ na-akọwapụta ihe ngosi na-aga n'ihu n'ime nkeji ole na ole nwere oke, na-enye ebumnuche maka ebumnuche dị iche.