Ụdị Ntugharị RNN
The RNN-Transducer (RNN-T) bụ nkwanye okwu enyi na enyi ije ije na-edozi nnukwu adịghị ike CTC - ya enweghị ike nlereanya ndabere n'etiti mmepụta tokens.
Nchịkọta
It powers much of the on-device 'live' speech recognition you use every day.
Ime miri emi
Alex Graves (2012) webakwara ya, RNN-Transducer na-ejikọta akụkụ atọ. Ihe ngbanwe (netwọọdụ ederede) na-ahazi okpokolo agba ka ọ bụrụ njiri egwu. Netwọk amụma na-eme dị ka ihe nlereanya asụsụ, na-edozi usoro nke akara ederede ewepụtara na mbụ. Otu obere netwọkụ jikọrọ ọnụ wee jikọta nlele ngbanwe nke 'ebe anyị nọ na ọdịyo' na nlele netwọk amụma nke 'ihe anyị kwuru ugbu a' iji akara akara na-esote n'elu okwu nke gụnyere oghere. N'adịghị ka CTC, netwọk amụma na-ewepụ echiche nke nnwere onwe, ya mere RNN-T na-amụta mkpoputa ezi uche na usoro okwu n'ime. Decoding na-aga n'ihu 2D lattice nke oge ọdịyo yana akara mmepụta, na-ewepụta oghere iji gaa n'ihu site na ọdịyo na ezigbo akara iji gaa n'ihu site na ederede - na-akwado mmepụta nkwanye ugwu.
Nghọta nka nka
Ọnwụ RNN-T, dị ka CTC, na-achịkọta ụzọ niile ziri ezi site na nlọghachi azụ azụ, mana n'elu grid akụkụ abụọ (usoro oge site na ọnọdụ mmepụta) karịa otu usoro. Mwepụta nke na-abụghị oghere na-anọ n'otu etiti ọdịyo ahụ ma na-akwalite ndeksi labelụ; na-ewepụta oge ọganihu efu. Nhazi nke otu a, aka ekpe gaa n'aka nri bụ kpọmkwem ihe kpatara RNN-T na-asọ asọ nke ọma na njedebe nwere oke, n'adịghị ka nlebara anya zuru oke nke nwere ike ileba anya na nkwupụta dum.
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 ụdị RNN-Transducer
RNN-T bụ nhọrọ kachasị maka mmepụta ASR na-ebuputa ma na-eji conformer encoders kama LSTM. Nchọpụta na-elekwasị anya n'ịbelata ọnụ ahịa ebe nchekwa ya dị arọ n'oge ọzụzụ, na-achịkwa nkwụsị ikuku ka edemede pụtara ngwa ngwa, na nhazi 'ngwa ngwa'. Na-atụ anya na nchikota na-aga n'ihu na ọzụzụ ọzụzụ na-elekọta onwe ya na ndị ntụgharị asụsụ ọtụtụ asụsụ, gbakwunyere ntinye ngwa ngwa karịa ka a na-atụnye amụma na netwọk mmekọrịta ma kwachaa.
Mmejuputa n'ezie n'ụwa
Nchọpụta okwu Google maka ngwa Gboard na Pixel Recorder, na-agba ọsọ na-anọghị n'ịntanetị.
Nkọwa okwu dị ndụ nke na-ebugharị okwu ka ị na-ekwu kama ichere ka ị mechaa ahịrịokwu
Ndị enyemaka olu na-edegharị iwu n'oge dị ala mgbe ị ka na-ekwu okwu
Nzuko ozugbo na ndegharị oku ebe nsonaazụ akụkụ ga-apụtarịrị mgbe niile
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 RNN-Transducer Models 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
Nkewa RNN-Dual-Path
Ajụjụ a na-ajụkarị
What is RNN-Transducer Models?
The RNN-Transducer (RNN-T) bụ nkwanye okwu enyi na enyi ije ije na-edozi nnukwu adịghị ike CTC - ya enweghị ike nlereanya ndabere n'etiti mmepụta tokens. Ọ na-akwado ọtụtụ njirimara okwu 'ndụ' na ngwaọrụ ị na-eji kwa ụbọchị.
Kedu mmachi nke CTC ka RNN-Transducer na-ekwu kpọmkwem?
RNN-T na-agbakwụnye netwọk amụma na ọnọdụ na tupu akara, na-ewepụ echiche CTC na mmepụta ọ bụla nwere onwe ya nyere ụda.
Kedu ihe bụ isi ihe atọ nke RNN-Transducer?
RNN-T na-ejikọta koodu ntinye ọdịyo yana netwọk amụma nwere ntụ oyi, jikọtara ya na obere netwọk njikọ nke na-ede akara na-esote.
Kedu ọrụ netwọk amụma na-arụ na RNN-T?
Netwọk amụma na-arụ ọrụ dị ka ụdị asụsụ ime n'ime akụkọ ihe mere eme akara ngosi, na-enye RNN-T mkpope akwụkwọ na ụkpụrụ okwu.
Kedu ihe kpatara RNN-T ji akwado nkwanye ala na-adịghị ala ala n'ụzọ nkịtị?
RNN-T na-eje ije n'otu oge na-eji akara ngosi monotonic, ya mere ọ nwere ike iwepụta mmepụta ka ọdịyo na-abịarute na nkwụsịtụ nwere oke kama ichere vidiyo zuru ezu.
Na ngbanwe RNN-T, kedu ihe iwepụta akara oghere na-eme?
Na lattice RNN-T, oghere oghere na-akwalite axis oge (gaa na etiti ọdịyo na-esote), ebe enweghị oghere na-aga n'ihu axis labelụ.