Ubuyobozi bwa Audio AI

Ibisobanuro byo Kumenyekanisha Imvugo

SpecAugment nuburyo bworoshye ariko bukomeye bwo kongera amakuru uburyo bwo guhisha no guhinduranya spekrogramu yimvugo kugirango moderi yo kumenyekana irusheho gukomera.

2 min somaIbiherutse kuvugururwa

Incamake

It boosted accuracy on benchmarks without any new audio or model changes.

Kwibira cyane

SpecAugment, yatangijwe na Google Ubwonko (Parike et al.) Muri 2019, byongera amahugurwa yo kumenyekanisha imvugo uhindura log-mel spectrogramme aho kuba umurongo wibanze. Ikoresha ibikorwa bitatu: igihe cyo gufata umwanya, kirambuye gato cyangwa kigabanya amajwi mugihe cyerekezo; guhisha inshuro, zeru zitsinda imirongo yimiyoboro yumurongo; na masking igihe, ikuraho umwanya wintambwe. Muguhatira icyitegererezo kumenya imvugo niyo mugihe uduce twa spekrogramu twihishe, SpecAugment ikora nkibisanzwe kandi ikabuza gukabya. Byari bihendutse kandi byiza, bifasha moderi yuburyo bwa LAS kugera icyo gihe igipimo cyamagambo yamagambo yibeshya kuri LibriSpeech na Switchboard, kandi iracyari ikintu cyambere muburyo bwa ASR butoza.

Ubushishozi

SpecAugment ikora kuri 2D spekrogramu nkaho ari ishusho. Guhisha inshuro nyinshi bikuraho umurongo uteganijwe wa mel-frequency; igihe cyo gukuramo gikuraho ikintu kidasanzwe cyama kadamu; igihe cyo guhinduranya gihindura ingingo yatoranijwe mugihe cyigihe ukoresheje interpolation. Masike nyinshi zirashobora gukoreshwa kumvugo. Kuberako masike ihindura ibihe byose, icyitegererezo kibona neza itandukaniro ridashira rya buri ngero, kunoza rusange muri rusange udakusanyije amakuru mashya.

Ingaruka z'Ingamba

Kugera no kugera

Itezimbere kugerwaho binyuze mu kwandukura, kuvuga, no guhuza amajwi.

Igiciro na bije

Amatsinda yibitangazamakuru arashobora kohereza amajwi yihuse hamwe na bije nto.

Umuvuduko n'igipimo

Sisitemu ireba abakiriya irashobora gutunganya imikoranire ivugwa murwego runini.

Ejo hazaza h'Icyerekezo cyo Kumenyekanisha Imvugo

SpecAugment yahindutse hafi-yisi yose muburyo bwo kumenyekanisha imvugo kandi ikwirakwira mubindi bikorwa byamajwi nko kugenzura abavuga no gutondekanya amajwi. Akazi kazoza gahuza politiki yo guhisha mu buryo bwikora cyangwa kuyihuza mugihe cy'amahugurwa, kandi igahuza masike ya spekrogramu hamwe nintego yo kwiyobora wenyine. Mugihe icyitegererezo gikura, kwiyongera bihendutse byongeramo imbaraga nta majwi yongeyeho amajwi bikomeza kuba iby'igiciro cyinshi, cyane cyane ku ndimi nkeya zifite aho amakuru ari make.

Gushyira mu bikorwa Isi

Kunoza igipimo cyamakosa kuri LibriSpeech muguhisha bande ya spekrogramu mugihe cy'amahugurwa

Guhindura impera-zanyuma za ASR nka LAS cyangwa Conformer kugirango ugabanye gukabya

Gutezimbere imibare mike yindimi-nke zindimi zitanditse amajwi mashya

Guhuza igitekerezo cyo guhishira abavuga kugenzura no kwerekana amajwi y'ibyiciro

Ingaruka & Kurinda

Gukoresha nabi amajwi no kwigira ibyago byiyongera mugihe uruhushya rubuze.

Ukuri kurashobora kugabanuka hejuru yimvugo, imvugo, cyangwa urusaku rwibidukikije.

Amajwi yubukorikori arashobora kwibeshya kumvugo yukuri nta kirango gisobanutse.

Igishushanyo mbonera

1

Shaka uruhushya rusobanutse rwo gufata amajwi, gukoroniza, no gukoresha.

2

Ikizamini cyiza mubiganiro bitandukanye hamwe nuburyo bwimbere.

3

Sobanura igihe umuntu agomba gusuzuma cyangwa kwemeza ibisubizo.

4

Andika amajwi yubukorikori kandi ugumane inyandiko zerekana kubazwa.

Komeza Ubushakashatsi

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.

Tangira ikibazo

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

Ubuyobozi bukurikira

Igikoresho cyo Kumenyekanisha Kaldi

Ibibazo bikunze kubazwa

What is SpecAugment for Speech Recognition?

SpecAugment nuburyo bworoshye ariko bukomeye bwo kongera amakuru uburyo bwo guhisha no guhinduranya spekrogramu yimvugo kugirango moderi yo kumenyekana irusheho gukomera. Yazamuye ukuri ku bipimo nta majwi mashya cyangwa amajwi yahindutse.

Niki SpecAugment ikora?

SpecAugment ihindura spekrogramu (igihe-inshuro zerekana) mu buryo butaziguye aho kuba icyatsi kibisi.

Niki muribi murimwe mubikorwa bitatu bya SpecAugment?

Ibikorwa bitatu ni igihe cyo gufata umwanya, guhisha inshuro, no guhisha igihe.

Niyihe ntego yibanze ya SpecAugment?

Muguhisha ibice bya spekrogramu, bihatira icyitegererezo muri rusange, kugabanya gukabya no kugabanya ibipimo byamakosa.

'Guhisha umwanya' gukora iki?

Igihe cyo guhisha zeru hanze yikurikiranya yikurikiranya yamakadiri akurikiranye mugihe cyerekezo cya ecran ya ecran.

Kuki guhindura masike buri gihe bifasha?

Mask zitandukanye zidasanzwe buri gihe bivuze ko icyitegererezo gihura nimpinduka zidashira, kunoza rusange muri rusange nta makuru mashya.