SpecAugment don gane Magana
SpecAugment hanya ce mai sauƙi amma mai ƙarfi ta haɓaka bayanai wacce ke rufe fuska da kuma karkatar da simintin magana don sa ƙirar ƙira ta fi ƙarfi.
Dubawa
It boosted accuracy on benchmarks without any new audio or model changes.
Zurfafa nutsewa
SpecAugment, wanda Google Brain (Park et al.) ya gabatar a cikin 2019, yana haɓaka horarwar fahimtar magana ta hanyar gyara ma'aunin log-mel kai tsaye maimakon sigar ƙaƙƙarfan motsi. Yana aiwatar da ayyuka guda uku: rikicewar lokaci, wanda ɗan shimfiɗawa ko matsawa sauti tare da axis lokaci; mita masking, wanda zeroes fitar da makada na mita tashoshi; da masking lokaci, wanda ke ɓoye tsawon matakan matakan lokaci. Ta hanyar tilasta ƙirar don gane magana ko da an ɓoye ɓangarori na spectrogram, SpecAugment yana aiki azaman daidaitawa kuma yana hana wuce gona da iri. Ya kasance mai arha sosai kuma mai inganci, yana taimaka wa ƙirar LAS-style su kai sa'an nan ƙimar kuskuren kalma na zamani akan LibriSpeech da Switchboard, kuma ya kasance babban sinadari a cikin bututun horo na ASR na zamani.
Fahimtar Fasaha
SpecAugment yana aiki akan sikirin 2D kamar hoto ne. Matsakaicin abin rufe fuska yana cire bazuwar toshe tashoshi na mel-mel; masking lokaci yana cire bazuwar toshe na firam masu yawa; Warping lokaci yana canza wurin da aka zaɓa tare da axis lokaci ta amfani da interpolation. Ana iya amfani da masks da yawa kowace magana. Saboda masks suna canza kowane zamani, ƙirar tana ganin bambance-bambance marasa iyaka na kowane misali, inganta haɓakawa ba tare da tattara sabbin bayanai 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 Bayanan Bayani don Gane Magana
SpecAugment ya zama tsoho na kusa-duniya a cikin fahimtar magana kuma yana yaduwa zuwa wasu ayyuka masu jiwuwa kamar tantancewar lasifika da rarraba sauti. Aiki na gaba yana kunna manufofin rufe fuska ta atomatik ko daidaita su yayin horo, kuma yana haɗa abin rufe fuska da maƙasudin horarwa na kulawa da kai. Kamar yadda samfura ke girma, haɓaka mai arha wanda ke ƙara ƙarfi ba tare da ƙarin alamar sauti ba ya kasance mai ƙima sosai, musamman ga ƙananan yarukan da ke da ƙarancin bayanai.
Aiwatar da Gaskiyar Duniya
Haɓaka ƙimar kuskuren kalma akan LibriSpeech ta hanyar rufe maƙallan spectrogram yayin horo
Tsayar da samfuran ASR na ƙarshe zuwa ƙarshe kamar LAS ko Conformer don rage wuce gona da iri
Ƙirƙirar ƙayyadaddun bayanan bayanai don ƙananan albarkatun harsuna ba tare da yin rikodin sabon sauti ba
Daidaita ra'ayin abin rufe fuska zuwa tabbatarwar lasifika da rarrabuwar taron sauti
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 SpecAugment for Speech Recognition 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
Kayan aikin Gane Magana Kaldi
Tambayoyin da ake yawan yi
What is SpecAugment for Speech Recognition?
SpecAugment hanya ce mai sauƙi amma mai ƙarfi ta haɓaka bayanai wacce ke rufe fuska da kuma karkatar da simintin magana don sa ƙirar ƙira ta fi ƙarfi. Ya haɓaka daidaito akan ma'auni ba tare da wani sabon sauti ko canji na ƙira ba.
Menene SpecAugment ke aiki akai?
SpecAugment yana gyara yanayin bakan (wakilcin-lokaci-lokaci) kai tsaye maimakon danyen kalaman kalaman.
Wanne ne daga cikin waɗannan ayyuka uku na SpecAugment?
Ayyuka guda uku sune rikicewar lokaci, masking mita, da rufe lokaci.
Menene ainihin manufar SpecAugment?
Ta hanyar ɓoye ɓangarori na spectrogram, yana tilasta ƙirar don haɓakawa, rage wuce gona da iri da rage ƙimar kuskure.
Menene 'time masking' yake yi?
Maƙerin lokaci yana fitar da bazuwar juzu'in firam ɗin jere tare da kullin lokaci na spectrogram.
Me yasa canza masks kowane zamani yana taimakawa?
Daban-daban masks na bazuwar kowane zamani yana nufin samfurin ya gamu da bambance-bambance mara iyaka, inganta haɓakawa ba tare da sabbin bayanai ba.