Conformer Architecture
Conformer shine toshewar hanyar sadarwa na jijiyoyi wanda ke haɗa juzu'i tare da kulawar kai, yana ɗaukar nau'ikan sauti na gida masu kyau da kuma mahallin dogon zango a cikin layi ɗaya.
Dubawa
It became the de facto standard encoder for state-of-the-art speech recognition.
Zurfafa nutsewa
An gabatar da shi ta Google a cikin 2020, Conformer ya amsa wata mahimmin tashin hankali a cikin ƙirar sauti: Hankalin kai (daga Transformers) yana da kyau a mahallin duniya amma yana da rauni a cikin gida, ƙirar ƙira mai bambance-bambancen sautin waya, yayin da rikice-rikicen ya yi fice a cikin gida amma yana fafutukar ganin tsawon lokaci mai tsawo. Toshe Conformer yana ɗinke su tare a cikin ƙirar 'sandwich': tsarin ciyar da gaba na rabin mataki, sannan tsarin kulawa da kai da yawa, sannan tsarin jujjuyawar, sannan tsarin ciyar da gaba na rabin mataki na biyu, tare da daidaitawar Layer da sauran haɗin gwiwa gaba ɗaya. Tsarin juyi yana amfani da juzu'i mai zurfi mai iya rabuwa da naúrar layi mai gate. Ta hanyar shiga tsakani aiki na gida da na duniya a cikin kowane toshe, Conformer encoders sun yanke ƙimar kuskuren kalma sosai akan tsantsar Mai Canjawa ko ingantaccen tsarin juyin juya hali akan alamomi kamar LibriSpeech.
Fahimtar Fasaha
Tsarin 'Macaron' na sa hannu ya ƙunshi hankali da juzu'i tsakanin nau'ikan ciyarwar gaba guda biyu, kowanne yana ba da gudummawar ragowar rabin nauyi (factor 0.5), wanda aka yi wahayi ta hanyar nazarin nau'ikan FFN Transformer. Tsarin jujjuyawar galibi yana ɗaure juzu'i mai ma'ana tare da kunna GLU, juyi mai zurfi, daidaita tsari, kunna Swish, da juzu'in juzu'i na ƙarshe - ingantacciyar hanya don ƙirar mahallin gida ba tare da fashe kirgawa ba.
Dabarun Tasiri
Shiga ku isa
Yana inganta samun dama ta hanyar rubutu, ba da labari, da mu'amalar murya.
Kudin da kasafin kuɗi
Ƙungiyoyin kafofin watsa labaru na iya jigilar sauti mai gogewa cikin sauri tare da ƙaramin kasafin kuɗi.
Gudu da sikelin
Tsarin fuskantar abokin ciniki na iya aiwatar da hulɗar magana a mafi girman ma'auni.
Makomar Conformer Architecture
Conformers yanzu suna aiki azaman mai rikodin kashin baya don transducer da CTC/hankali ASR, kuma ƙirar ta yadu zuwa fassarar magana, ƙwarewar magana, da gano taron sauti. Bincike mai aiki yana ba da hankali ga dogon sauti (hankali na madaidaiciya da chunked don yawo), yana lalata Conformers don amfani da na'ura, kuma ya haɗa su tare da horarwa na kulawa da kai. Bambance-bambance kamar Squeezeformer da Ingantattun Conformer suna tura daidaito-da-ƙididdigar cinikin gaba.
Aiwatar da Gaskiyar Duniya
Yin aiki azaman mai rikodin a cikin samar da tsarin ASR mai yawo a bayan mataimakan murya da dictation
Ƙarfafa ƙirar fassarar magana waɗanda ke yin kwafi da fassara harshen magana daga ƙarshe zuwa ƙarshe
Kashin baya don tantancewar lasifika da diarization, gano wanda yayi magana lokacin da ake taro
Taron odiyo da rarraba sauti, kamar gano ƙararrawa, magana, ko kiɗa a cikin rafi
Hatsari & Tsare-tsare
Rashin amfani da murya da haɗarin kwaikwaya yana ƙaruwa lokacin da aka rasa izini.
Daidaituwa na iya faɗuwa cikin lafuzza, yaruka, ko mahalli masu hayaniya.
Ana iya kuskuren sauti na roba don ingantacciyar magana ba tare da bayyananniyar lakabi ba.
Taswirar Hanya
Sami tabbataccen izini don ɗaukar murya, cloning, da sake amfani.
Gwajin ingantattun masu magana daban-daban da yanayin baya.
Ƙayyade lokacin da dole ne ɗan adam ya duba ko ya amince da abubuwan da aka fitar.
Yi lakabin sauti na roba da kuma adana bayanan da aka tabbatar don yin lissafi.
Ci gaba da Bincike
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 Conformer Architecture 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
Jagora na gaba
DeepSpeech Architecture
Tambayoyin da ake yawan yi
What is Conformer Architecture?
Conformer shine toshewar hanyar sadarwa na jijiyoyi wanda ke haɗa juzu'i tare da kulawar kai, yana ɗaukar nau'ikan sauti na gida masu kyau da kuma mahallin dogon zango a cikin layi ɗaya. Ya zama madaidaicin madaidaicin madaidaicin madaidaicin don ƙwarewar magana ta zamani.
Wadanne hanyoyi biyu ne tsarin gine-ginen Conformer ya haɗu a cikin toshe ɗaya?
Conformer yana haɗa juzu'i (don ƙirar gida) tare da kai (don mahallin duniya) a cikin kowane shinge.
Me yasa hada juyi da hankali ke da amfani musamman ga magana?
Juye-juye sun yi fice a kan kyawawan halaye na cikin gida waɗanda ke bambance sautin wayoyi, yayin da hankalin kai ya ƙirƙira abin dogaro a duk faɗin; Conformer yana samun duka biyu.
Menene tsarin 'Macaron' ko sanwici na toshe Conformer?
Conformer yana sanya hankalin kai da tsarin juzu'i tsakanin nau'ikan ciyarwa guda biyu, kowanne an yi amfani da shi tare da ragowar rabin nauyi.
Wace kungiya ce ta gabatar da Conformer, kuma a wace shekara?
Google masu bincike ne suka gabatar da Conformer a cikin 2020 kuma cikin sauri ya zama madaidaicin ma'aunin fahimtar magana.
Menene tsarin juzu'i a cikin Conformer yawanci ya haɗa?
Tsarin juyi yana haɗa juzu'in juzu'i mai ma'ana tare da gated naúrar layi, juzu'i mai zurfi, daidaita tsari, da kunna Swish, yana ƙarewa a cikin wani juzu'i mai ma'ana.