Wav2Letter Convolutional ASR
Wav2Letter bụ usoro njirimara okwu njedebe na njedebe sitere na Facebook AI nke na-eji naanị netwọkụ akwara convolutional, enweghị nlọghachite.
Nchịkọta
It mattered as a fast, simple alternative that proved CNNs alone could transcribe speech competitively.
Ime miri emi
Nchọpụta Facebook AI webatara na 2016, Wav2Letter mebiri site na usoro na-emekarị ugboro ugboro na HMM site na ịdabere kpamkpam na netwọkụ akwara na-agbanwe agbanwe iji mapụta ọdịyo ozugbo na mkpụrụedemede (akwụkwọ ozi), yabụ aha ya. Ọ zụrụ azụ site na mfu AutoSegCriterion (ASG) omenala, ụzọ dị mfe karịa mfu CTC na-ahụkarị nke tụfuru akara oghere na ntụgharị mkpụrụedemede ozugbo. Edere ya na C++ site na iji Flashlight/ArrayFire backend, e mere ya maka ọsọ na ma CPU na GPU. Ụdị ndị ọzọ mechara, Wav2Letter++ na ụdị mgbanwe ngbanwe zuru oke, gbagoro na nnukwu dataset wee nweta ọnụego njehie okwu asọmpi na Librispeech. Nhazi ya-naanị convolution mere ka ọ bụrụ nke a na-emekọrịta ihe nke ukwuu yana mmetụ enyi ma e jiri ya tụnyere ndị nrụpụta RNN usoro.
Nghọta nka nka
Wav2Letter na-achịkọta mgbagha oge 1D maka njirimara ụda olu, na oyi akwa ọ bụla na-agbasawanye ogige nnabata nke mere na nchịkọta miri emi na-ejide ogologo ọnọdụ na-enweghị nlọghachi. N'ihi na convolutions na-ahazi usoro oge niile n'otu oge, ọzụzụ na ntinye aka na-adị ngwa ngwa. Mfu ASG mbụ dị ka CTC mana ọ na-ewepụ akara oghere wee gbakwunye akara mgbanwe mkpụrụedemede na mkpụrụedemede doro anya, na-ewepụta usoro usoro dị iche iche nke na-ahazi ọdịyo ogologo ogologo na mmepụta agwa na-enweghị akara ọ bụla.
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 Wav2Letter Convolutional ASR
Usoro ọmụmụ nke Wav2Letter na-ebi na ọkụ ọkụ, ọbá akwụkwọ mmụta igwe C++ nke Facebook, wee gwa ụdị ndị na-ahụ maka onwe ha wav2vec na-achị ugbu a. Ihe mmụta ka ukwuu, na convolution na ndị yiri architectures nwere ike dakọtara na nlọghachi azụ, nri ozugbo n'ime transformer dabeere ASR. Na-atụ anya ka usoro n'ọdịnihu ga-anọgide na-agbazinye ego Wav2Letter na-arụ ọrụ nke ọma, n'otu aka ahụ, nke nwere ike ịdị iche zuru oke na njedebe na njedebe mgbe ị na-etinye aka na ọzụzụ onwe onye na-elekọta maka asụsụ ndị dị ala.
Mmejuputa n'ezie n'ụwa
Ndegharị oge n'ezie ebe ọ na-adị ala, ntinye n'otu aka ahụ bara uru karịa isi ihe ole na ole ziri ezi
Nchọpụta okwu na ngwaọrụ ma ọ bụ CPU nke na-enweghị ike ị nweta ihe nrụpụta dị arọ na-emegharị ugboro ugboro
Usoro nyocha na-atụnyere convolutional ASR megide RNN na sistemụ ntụgharị na Librispeech
Na-eje ozi dị ka ntọala injinia maka ọbaakwụkwọ ọkụ ọkụ Facebook na ụdị wav2vec mechara
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 Wav2Letter Convolutional ASR 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
Netwọk Neural Convolutional
Ajụjụ a na-ajụkarị
What is Wav2Letter Convolutional ASR?
Wav2Letter bụ usoro njirimara okwu njedebe na njedebe sitere na Facebook AI nke na-eji naanị netwọkụ akwara convolutional, enweghị nlọghachite. Ọ dị mkpa dị ka ngwa ngwa, ụzọ dị mfe nke gosipụtara na CNN naanị nwere ike ịdegharị okwu na asọmpi.
Kedu ụlọ ọrụ netwọkụ akwara Wav2Letter na-adabere naanị?
Wav2Letter bụ ihe ama ama maka iji naanị netwọkụ akwara na-eme mgbanwe, na-ezere nlọghachi kpamkpam.
Kedu nzukọ mebere Wav2Letter?
Facebook AI Research webatara Wav2Letter na 2016 wee mesịa ghọọ ọbá akwụkwọ ọkụ ọkụ.
Kedu ihe 'leta' dị na Wav2Letter na-ezo aka?
Wav2Letter na-esetịpụ ụda ụda olu ozugbo gaa n'usoro mkpụrụedemede (akwụkwọ ozi), ụzọ ngwụcha na njedebe.
Kedu ka mfu AutoSegCriterion (ASG) mbụ si dị iche na mfu CTC na-adịkarị?
ASG tụfuru oghere oghere CTC kama na-agbakwụnye akara mgbanwe doro anya n'etiti mkpụrụedemede ka ọ na-adịkwa iche.
Kedu ihe bụ isi uru bara uru nke naanị nhazi nke Wav2Letter?
N'adịghị ka usoro RNN nke usoro, enwere ike ịgbakọ convolutions n'otu oge n'ogologo, na-eme ka usoro ahụ dị ngwa na ịrụ ọrụ nke ọma.