Moderi ya RNN-Transducer
RNN-Transducer (RNN-T) ni uburyo bworoshye bwo kuvuga imvugo yubaka ikosora intege nke za CTC - kuba idashobora kwerekana urugero rushingiye ku bimenyetso bisohoka.
Incamake
It powers much of the on-device 'live' speech recognition you use every day.
Kwibira cyane
Yatangijwe na Alex Graves (2012), RNN-Transducer ihuza ibice bitatu. Kodegisi (umuyoboro wandikirwa) itunganya amajwi mumajwi ya acoustic. Urusobe rwo guhanura rukora nkurugero rwururimi, rutondekanya kurukurikirane rwibimenyetso byasohotse mbere. Umuyoboro muto uhuriweho noneho uhuza kodegisi ya 'aho turi mumajwi' hamwe numuyoboro wo guhanura kureba 'ibyo twavuze kugeza ubu' kugirango utange ikimenyetso gikurikira hejuru yamagambo arimo ubusa. Bitandukanye na CTC, umuyoboro wo guhanura ukuraho igitekerezo-cyigenga, bityo RNN-T yiga imyandikire ifatika hamwe nijambo imbere. Decoding igenda 2D lattice ya majwi-igihe hamwe nibisohoka-ibimenyetso, gusohora ubusa kugirango utere imbere ukoresheje amajwi nibimenyetso nyabyo kugirango utere imbere ukoresheje inyandiko - mubisanzwe ushyigikira ibisohoka.
Ubushishozi
Igihombo cya RNN-T, kimwe na CTC, igereranya inzira zose zemewe zo guhuza binyuze mu gusubira inyuma-gusubira inyuma, ariko hejuru ya gride-ebyiri (igihe intambwe ku myanya isohoka) aho kuba urukurikirane rumwe. Gusohora ibitari ubusa biguma kumurongo umwe wamajwi kandi biteza imbere ikirango; gusohora ubusa. Iyi monotonic, ibumoso-iburyo niyo mpamvu rwose imigezi ya RNN-T isukuye neza hamwe nubukererwe bwimbibi, bitandukanye nubwitonzi bwuzuye bushobora kureba kumvugo yose.
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 ka RNN-Transducer Model
RNN-T niyo yiganje mugutunganya umusaruro ASR kandi igenda ikoresha kodegisi ya Conformer aho gukoresha LSTMs. Ubushakashatsi bwibanze ku kugabanya igiciro cyinshi cyo kwibuka mugihe cyamahugurwa, kugenzura ubukererwe bw’ibyuka bihumanya bityo ibisobanuro bigaragara vuba, hamwe na 'emit yihuta'. Tegereza gukomeza guhuza hamwe no kwiyobora kwiyobora no guhinduranya indimi nyinshi, hiyongereyeho uburyo bwo kohereza ibikoresho nkuko guhanura no guhuza imiyoboro yabaruwe kandi igacibwa.
Gushyira mu bikorwa Isi
__AIU_PROTECTED_11 __ 'ku bikoresho byerekana imvugo kubikoresho bya Gboard hamwe na Pixel Recorder, ikora kumurongo wose
Kwandika neza byerekana amagambo uko uvuga aho gutegereza ko urangiza interuro
Abafasha b'ijwi bandika amategeko hamwe n'ubukererwe buke mugihe ukivuga
Iteraniro-nyaryo no guhamagara transcript aho ibisubizo byigice bigomba kugaragara ubudahwema
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
Shaka uruhushya rusobanutse rwo gufata amajwi, gukoroniza, no gukoresha.
Ikizamini cyiza mubiganiro bitandukanye hamwe nuburyo bwimbere.
Sobanura igihe umuntu agomba gusuzuma cyangwa kwemeza ibisubizo.
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 RNN-Transducer Models 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
Ubuyobozi bukurikira
Inzira-ebyiri RNN Gutandukana
Ibibazo bikunze kubazwa
What is RNN-Transducer Models?
RNN-Transducer (RNN-T) ni uburyo bworoshye bwo kuvuga imvugo yubaka ikosora intege nke za CTC - kuba idashobora kwerekana urugero rushingiye ku bimenyetso bisohoka. Ifasha byinshi mubikoresho-bizima 'bizima' imvugo ukoresha buri munsi.
Ni izihe mbogamizi za CTC RNN-Transducer ikemura?
RNN-T yongeyeho umuyoboro wo guhanura ibintu bisabwa mbere yikimenyetso, ukuraho CTC yibwira ko buri gisohoka cyigenga ukurikije amajwi.
Nibihe bintu bitatu byingenzi bigize RNN-Transducer?
RNN-T ihuza kodegisi y'amajwi hamwe numuyoboro uteganijwe wo guhanura, uhujwe numuyoboro muto uhuriweho utanga amanota akurikira.
Ni uruhe ruhare umuyoboro uhanura ugira muri RNN-T?
Urusobekerane rwimikorere rukora nkururimi rwimbere hejuru yibisohoka byerekana ibimenyetso, biha RNN-T imyandikire ifatika nijambo.
Kuki RNN-T ishyigikira umuvuduko muke gutemba muburyo busanzwe?
RNN-T igenda ya monotonic umwanya-ku-token, bityo irashobora gusohora ibisohoka nkuko amajwi ageze hamwe nubukererwe aho gutegereza clip yuzuye.
Muri decode ya RNN-T, gusohora ikimenyetso cyambaye iki?
Muri lattice ya RNN-T, ubusa butera imbere umwanya (kwimukira kumurongo ukurikira wamajwi), mugihe ibitari ubusa bitera imbere label axis.