Ubuyobozi bwa Audio AI

Umva Kwitabira no Kwandika

Umva, Kwitabira no Kwandika (LAS) numuyoboro wingenzi wa 2015 uhuza imvugo mu nyuguti, nta nkoranyamagambo yubatswe n'intoki cyangwa imvugo itandukanye.

2 min somaIbiherutse kuvugururwa

Incamake

It showed that a single end-to-end model could do speech recognition.

Kwibira cyane

Umva, Kwitabira no Kuroga, byatangijwe na Google abashakashatsi Chan, Jaitly, Le, na Vinyals muri 2015, yari umwe mubambere bamenye imvugo yanyuma-iherezo. Ifite ibice bibiri: 'Umva,' LSTM ya piramide igereranya amajwi mugihe igabanya igihe, hamwe na 'Speller,' ishingiye kuri decoder ya LSTM isohora inyuguti imwe imwe. Uburyo bwo kwitondera butuma Speller yibanda ku gice kijyanye n'amajwi kuri buri rwandiko rusohoka. Bitandukanye n'imiyoboro ya HMM-DNN ishaje, LAS ntikeneye inkoranyamagambo ya foneme, nta guhuza ku gahato, kandi nta n'indimi yatojwe itandukanye; yiga imyandikire, imbibi zijambo, hamwe na acoustics hamwe uhereye kumajwi yandukuwe. Byashishikarije mu buryo butaziguye sisitemu igezweho-ikurikiranye kandi ishingiye kuri sisitemu ya ASR.

Ubushishozi

LAS ikomatanya kodegisi-decoder hamwe no kwitondera. Kodegisi ya piramide LSTM igabanya kabiri igihe cyo gukemura kuri buri cyiciro, igabanya urutonde rurerure rwa acoustique mu burebure bushobora gucungwa bityo kwitabwaho kurikwega. Kuri buri ntambwe ya decodeur Speller ibara uburemere bwibisobanuro kuri leta zose za kodegisi, ikabihuza murwego rwimiterere, kandi igahanura imiterere ikurikira. Amahugurwa yerekana amahirwe menshi yimiterere ikurikirana; gahunda iteganijwe-yo gutoranya igabanya gari ya moshi / ikizamini kidahuye.

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.

Kazoza Kumva Witabe kandi Wamagane

LAS ubu ni amateka, ariko ADN yayo inyura muri sisitemu igezweho ya ASR. Igitekerezo cyacyo gishingiye kuri encoder-decoder igitekerezo cyahindutse mubimenyekanisha ba Transformer na Conformer, mugihe inzira zijyanye na RNN-Transducer imbaraga kubikoresho byabigenewe. Sisitemu yigihe kizaza ikomeza inzira yanyuma-iherezo, guhuza kumenyekanisha hamwe no gusobanura no gusobanukirwa murugero rumwe rwindimi nyinshi, kandi ugasunika kugana, kwandikirana kworoheje LAS, kuba idatemba, idashobora gutanga mbere.

Gushyira mu bikorwa Isi

Kwandukura icyongereza kivugwa mu nyuguti zidafite inkoranyamagambo

Gukora nkibisobanuro byibanze kubitekerezo-bishingiye kumajwi hamwe na sisitemu yo kwandika

Kwerekana amahugurwa arangiza-arangije amasomo yo kumenya-kumenyekanisha amasomo n'ibipimo

Ibyitegererezo bikurikirana-bikurikiranye nyuma byakoreshejwe mumvugo yo guhindura imvugo

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 Listen Attend and Spell 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

Umuziki Auto-Tagging

Ibibazo bikunze kubazwa

What is Listen Attend and Spell?

Umva, Kwitabira no Kwandika (LAS) numuyoboro wingenzi wa 2015 uhuza imvugo mu nyuguti, nta nkoranyamagambo yubatswe n'intoki cyangwa imvugo itandukanye. Yerekanye ko icyitegererezo kimwe cyanyuma-kirangira gishobora gukora imvugo.

Nibihe bintu bibiri byingenzi bigize moderi ya LAS?

LAS igizwe na kodegisi ya 'Umva' itunganya amajwi na 'Speller' ishingiye kuri decoder yohereza inyuguti.

Niki Speller muri LAS isohoka?

Speller ni decoder itanga inyuguti zandikirwa imiterere, yiga imyandikire itaziguye.

Kuki kodegisi ya LAS yitwa 'piramide'?

Buri cyiciro cya piramide BLSTM igabanya kabiri uburebure bwikurikiranya, igabanya amajwi maremare akurikirana kuburyo kwitabwaho bishoboka kubara.

Nibihe bintu bishaje LAS idasanzwe idasaba?

Bitandukanye n'imiyoboro ya HMM-DNN isanzwe, LAS yiga biturutse kumajwi-ku-jambo nta nkoranyamagambo ya foneme cyangwa guhuza ku gahato.

Ni ubuhe buryo butuma Speller yibanda ku gice kijyanye n'amajwi kuri buri nyuguti?

Uburyo bwo kwitondera bugereranya uburemere hejuru ya kodegisi, ikora imiterere ya vector decoder ikoresha kuri buri ntambwe.