Audio AI JAGORA

Wav2Letter Convolutional ASR

Wav2Letter shine tsarin gane magana daga ƙarshen zuwa-ƙarshen daga Facebook AI wanda yayi amfani da hanyoyin sadarwa na juyi kawai, babu maimaituwa.

2 min karatuAn sabunta ta ƙarshe

Dubawa

It mattered as a fast, simple alternative that proved CNNs alone could transcribe speech competitively.

Zurfafa nutsewa

Binciken Facebook AI ya gabatar da shi a cikin 2016, Wav2Letter ya karye daga mafi girman kai-tsaye da hanyoyin tushen HMM ta hanyar dogaro gabaɗaya akan hanyoyin sadarwa na juzu'i don taswirar sauti kai tsaye zuwa haruffa (haruffa), saboda haka sunan. An fara horar da shi tare da asarar AutoSegCriterion (ASG) na al'ada, madadin mafi sauƙi ga asarar CTC gama gari wanda ya watsar da alamar mara kyau da ƙirar wasiƙa kai tsaye. An rubuta shi a cikin C ++ ta amfani da bangon Flashlight/ArrayFire, an yi shi don saurin gudu akan CPU da GPU. Sigar baya, Wav2Letter++ da cikakken bambance-bambancen juyin juya hali, an daidaita su zuwa manyan ma'ajin bayanai kuma sun sami ƙimar kuskuren kalmomin gasa akan Librispeech. Ƙirar juyi-kawai ya sanya ta zama mai kama da juna sosai kuma mai sauƙin fahimta idan aka kwatanta da na'urori na RNN na jeri.

Fahimtar Fasaha

Wav2Letter ya tattara juzu'i na wucin gadi na 1D akan fasalulluka na sauti, tare da kowane Layer yana faɗaɗa filin karɓa don haka zurfafan rikodi suna ɗaukar dogon mahallin ba tare da maimaituwa ba. Saboda rikice-rikice suna aiwatar da duk matakan lokaci a layi daya, horarwa da tunani suna da sauri. Asarar ASG ta asali tana kama da CTC amma tana cire alamar da ba komai ba kuma tana ƙara madaidaicin madaidaicin wasiƙa zuwa wasiƙa, yana samar da cikakkiyar ma'aunin ma'auni mai banbanta wanda ke daidaita sauti mai tsayi mai tsayi zuwa fitowar hali ba tare da kowane lakabin firam 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 Wav2Letter Convolutional ASR

Lantarki kai tsaye na Wav2Letter yana rayuwa cikin Hasken walƙiya, ɗakin karatu na koyon injin C++ na Facebook, kuma ya sanar da samfuran wav2vec masu sarrafa kansu waɗanda yanzu suka mamaye. Babban darasi, cewa juyin juya hali da tsarin gine-gine masu kama da juna na iya dacewa da maimaitawa, ciyar da kai tsaye zuwa tushen ASR. Yi tsammanin tsarin nan gaba don ci gaba da rancen fifikon Wav2Letter akan ingantaccen, layi daya, cikakken bambance-bambancen bututun daga ƙarshen zuwa-ƙarshe yayin da ake sa ido kan horar da kai don ƙananan albarkatu.

Aiwatar da Gaskiyar Duniya

Rubutun lokaci na ainihi inda rashin jinkiri, magana mai kama da juna ya fi kima fiye da ƴan maki na daidaito

Ganewar magana ta kan na'ura ko mai ɗaure CPU waɗanda ba za su iya ba da manyan na'urori masu maimaitawa ba

Tushen bincike da ke kwatanta ASR na juyin juya hali akan RNN da tsarin canji akan Librispeech

Yin hidima a matsayin tushen injiniya don ɗakin karatu na Flashlight na Facebook da kuma samfuran wav2vec daga baya

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

1

Sami tabbataccen izini don ɗaukar murya, cloning, da sake amfani.

2

Gwajin ingantattun masu magana daban-daban da yanayin baya.

3

Ƙayyade lokacin da dole ne ɗan adam ya duba ko ya amince da abubuwan da aka fitar.

4

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 Wav2Letter Convolutional ASR quiz

Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.

Fara tambayoyi

Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation

Jagora na gaba

Hanyoyin Sadarwar Jijiya Na Juyin Halitta

Tambayoyin da ake yawan yi

What is Wav2Letter Convolutional ASR?

Wav2Letter shine tsarin gane magana daga ƙarshen zuwa-ƙarshen daga Facebook AI wanda yayi amfani da hanyoyin sadarwa na juyi kawai, babu maimaituwa. Yana da mahimmanci azaman madadin sauri, sauƙi mai sauƙi wanda ya tabbatar CNNs kaɗai zai iya rubuta magana cikin gasa.

Wane tsarin gine-ginen cibiyar sadarwa na Wav2Letter ya dogara musamman?

Wav2Letter sananne ne don amfani da cibiyoyin sadarwa na jujjuyawar juyi kawai, guje wa maimaitawa gaba ɗaya.

Wace kungiya ce ta fara haɓaka Wav2Letter?

Facebook AI Research ne ya gabatar da Wav2Letter a cikin 2016 kuma daga baya ya samo asali zuwa ɗakin karatu na Tocila.

Menene 'wasiƙa' a Wav2Letter ke nufi?

Wav2Letter yana tsara tsarin motsin sauti kai tsaye zuwa jerin haruffa (haruffa), hanya ta ƙarshe zuwa ƙarshe.

Ta yaya Asarar AutoSegCriterion (ASG) ta asali ta bambanta da mafi yawan asarar CTC?

ASG ta jefar da babur alama ta CTC kuma a maimakon haka tana ƙara madaidaicin maki tsakanin haruffa yayin da ya kasance mai banbanta.

Menene babban fa'ida mai amfani na ƙirar Wav2Letter's convolution-kawai?

Ba kamar jerin RNNs ba, ana iya ƙididdige juzu'ai a cikin layi daya a cikin lokaci, yana sa tsarin ya yi sauri da inganci.