InoteveraGaidhi rinotevera
Conformer Architecture
AI Yenyaya
Chidzidzo cheAI yeRuzha
DeepSpeech Architecture
DeepSpeech imhando yekuziva yekutaura-yekupedzisira-kusvika-kumagumo yakaunzwa naBaidu muna 2014 iyo inoburitsa maredhiyo manyoro akananga kumavara achishandisa inodzokororwa neural network yakadzidziswa nekurasikirwa kweCTC.
Teerera kugwaro iriAI-yakarondedzerwa · ~2 min
Pa peji ino2 min verenga
Pfupiso
Yakabatsira kupayona kusimuka kubva kune yakaoma, yakagadzirwa nemaoko ASR mapaipi kuenda kune yakadzidza, inofambiswa nedata.
Kudzika Kwakadzika
Matauriro echinyakare akabatanidza maacoustic modhi akaparadzana, maduramazwi emutauro, uye mhando dzemitauro dzine zvikamu zvakarongwa nemaoko. DeepSpeech yakatsiva yakawanda yeiyo neine imwe neural network yakadzidziswa kupera kusvika kumagumo. Mavakirwo ayo anotora spectrogram kana MFCC maficha pamusoro pemafuremu mapfupi odhiyo uye anoadyisa kuburikidza akati wandei akanyatsobatanidzwa akaturikidzana, bidirectional inodzokororwa layer inobata mamiriro kubva zvakapfuura uye ramangwana, uye inobuda layer inoburitsa mukana wekugovera pamusoro pemavara panguva yega yega nhanho. Zvine hutsinye, inoshandisa Connectionist Temporal Classification (CTC), iyo inoita kuti network idzidze kurongeka pakati peodhiyo uye zvinyorwa pasina kuda mavara-chikamu. Mozilla yakazoburitsa yakakurumbira yakavhurika-sosi kuita (ine shanduro nyowani uchishandisa LSTM-yakavakirwa, inoyerera dhizaini), ichiita kuti nzira yacho iwanikwe zvakanyanya.
Kuitwa muNyika yeChokwadi
Offline, pa-mudziyo izwi rekuraira kucherechedzwa kwekuvanzika-yakatarisana maapplication uchishandisa Mozilla yakavhurika DeepSpeech.
Kugadzira zvinyorwa zvinyorwa zvepodcasts kana hurukuro pasina kuvimba nesevhisi yegore
Kudzidzisa izvo zvakakosha zvekupedzisira-kusvika-kumagumo ASR uye CTC kurasikirwa mumayunivhesiti-yekudzidza makosi.
Kuvaka tsika yezwi inopindirana yeIoT kana yakamisikidzwa michina panodiwa kuremerwa, inoyerera inoziva
Ramangwana reDeepSpeech Architecture
DeepSpeech pachayo yakasimudzwa zvakanyanya nekutarisisa- uye transformer-based architectures (Conformer, Whisper, wav2vec 2.0) inobata mamiriro akareba uye kuzvitarisira pane isina kunyorwa odhiyo. Asi pfungwa dzayo dzepakati, kupera-kusvika-kumagumo kudzidziswa uye CTC decoding, inoramba iri hwaro uye ichiri kuoneka mukati memazuva ano mahybrid masisitimu. Nhaka iyi ndeyepfungwa: yakaratidza kuti modhi imwe chete yakadzidzwa inogona kukwikwidza mapaipi akagadziridzwa, achigadzira nzira yemazuva ano makuru, emitauro yakawanda, anozvitarisira ega enheyo yekutaura.
Kuzvishandisa
Nei zvichikosha
- Svika uye svika. Inonatsiridza kusvikika kuburikidza nekunyora, kurondedzera, uye mazwi ekubatanidza.
- Mutengo uye bhajeti. Zvikwata zveMedia zvinogona kutumira odhiyo yakakwenenzverwa nekukurumidza nemabhajeti madiki.
- Kumhanya uye chiyero. Masisitimu anotarisana nevatengi anogona kugadzirisa kutaurirana kwekutaura pamwero mukuru.
Chenjera
- Kushandisa izwi zvisizvo uye njodzi dzekuedzesera dzinowedzera kana chibvumirano chisipo.
- Kururama kunogona kudonha mumitauro, mataurirwo, kana nharaunda dzine ruzha.
- Synthetic audio inogona kukanganisa kutaura kwechokwadi isina mavara akajeka.
Matanho anotevera
- Wana mvumo yakajeka yekutora inzwi, kugadzira, uye kushandisa zvakare.
- Yedza mhando pavatauri vakasiyana uye mamiriro ekumashure.
- Tsanangura apo munhu anofanira kuongorora kana kubvumidza zvabuda.
- Label synthetic odhiyo uye chengetedza marekodhi ekuzvidavirira.
Ramba Uchiongorora
Tsamba yemahara
Tora iyo yemazuva ese AI pfupiso
Nhatu dzakasimbiswa dzeAI nyaya mangwanani ega ega pakati pevhiki, dzakanyorwa neChirungu chakajeka. Yemahara zvachose, hapana ads.
Imwe email zuva rega rega revhiki. Bvisa kunyoresa kamwe chete. Hatimbotengese kana kugovera kero yako.
Zviedze iwe pachako
Take the DeepSpeech Architecture quiz
Mhinduro yepakarepo pamhinduro yega yega, uye chitupa chinogovaniswa chine ID inobatika kana wapedza kosi.
Tsigira dzidzo yemahara yeAI. AI Understanding ndeye 501(c)(3) isingabatsiri - hapana ads, hapana paywall, zvachose. Ita chipo
Mibvunzo inowanzo bvunzwa
Chii chinonzi DeepSpeech Architecture?
DeepSpeech imhando yekuziva yekutaura-yekupedzisira-kusvika-kumagumo yakaunzwa naBaidu muna 2014 iyo inoburitsa maredhiyo manyoro akananga kumavara achishandisa inodzokororwa neural network yakadzidziswa nekurasikirwa kweCTC. Yakabatsira kupayona kusimuka kubva kune yakaoma, yakagadzirwa nemaoko ASR mapaipi kuenda kune yakadzidza, inofambiswa nedata.
Ndeipi basa rekurasikirwa rinobvumira DeepSpeech kudzidzisa pasina furemu-level kurongeka pakati pekuteerera uye zvinyorwa?
CTC inosuma chiratidzo chisina kuvharwa uye sums pamusoro pezvese zvigadziriso zvinodonha kune izvo zvakanangwa zvinyorwa, zvichibvisa kudiwa kwemazita e-per-frame.
Chii chaive chikuru hunyanzvi hwekuvaka iyo DeepSpeech yakakurumbira?
DeepSpeech yakatsiva iyo yechinyakare-matanho pombi neimwe neural network yakadzidziswa kupera kusvika kumagumo, shanduko huru mukugadzira ASR.
Sei DeepSpeech ichishandisa bidirectional recurrent layer?
Iyo bidirectional RNN inobata kutevedzana mumativi ese, saka imwe neimwe inobuda inobatsira kubva kune yakatenderedza acoustic mamiriro, kuvandudza huchokwadi.
Ndeupi rudzi rwekupinza maficha ayo DeepSpeech inowanzoshanda pairi?
Iyo modhi inoshandisa furemu-level odhiyo maficha senge spectrograms kana MFCCs uye zvinobuda maitiro zvingangoitika kune yega yega nhanho.
Chii chinowanzowedzerwa panguva yedecode kuti uvandudze DeepSpeech's izwi sarudzo uye zviperengo?
Kubatanidza iyo acoustic inobuda neyekunze yemutauro modhi panguva yekunyora kunobatsira sisitimu kuburitsa mamwe mazwi anonzwisisika uye zviperengo.
Ramba uchidzidza
Mazano anoenderana
Mamwe madhairekitori akasarudzirwa nyaya iyi