Teerera Huya Utaure
Teerera, Teerera uye Spell (LAS) inzvimbo ye2015 neural network inonyora mutauro zvakananga mumabhii, isina duramazwi rakagadzirwa nemaoko kana modhi yemutauro wakasiyana.
Pfupiso
It showed that a single end-to-end model could do speech recognition.
Kudzika Kwakadzika
Teerera, Teerera uye Spell, yakaunzwa ne Google vatsvakurudzi Chan, Jaitly, Le, uye Vinyals muna 2015, akanga ari mumwe wevakatanga kuziva kutaura-kuguma-kumagumo. Iine zvikamu zviviri: 'Mutereri,' pyramidal bidirectional LSTM inokodha iyoodhiyo ichidzoreredza chiyero chenguva, uye 'Speller,' inotarisisa-yakavakirwa LSTM decoder inoburitsa mavara rimwe panguva. Iyo yekutarisisa michina inoita kuti Speller atarise pane yakakodzera chidimbu cheaudio kune yega yega inobuda tsamba. Kusiyana nemapaipi ekare eHMM-DNN, LAS haidi duramazwi refoni, hapana kurongeka kwekumanikidzwa, uye hapana yakadzidziswa yakadzidziswa modhi yemutauro; inodzidza zviperengo, miganhu yemazwi, uye acoustics zvakabatana kubva kune akanyorwa odhiyo. Yakafuridzira zvakananga kutevedzana-kune-kutevedzana kwemazuva ano uye kutarisa-kwakavakirwa ASR masisitimu.
Technical Insight
LAS inosanganisa encoder-decoder ine hanya. Iyo pyramidal LSTM encoder inopedza hafu yekugadzirisa nguva pane imwe neimwe yezvikamu zvitatu, ichicheka kureba acoustic kutevedzana kuita hurefu hunogoneka kuti kutarisisa kugoneke. Panhanho yega yega yekudhirodha Speller inokokorodza huremu hwekutarisisa pamusoro pese encoder nyika, inoasanganisa kuita vheti yemamiriro, uye anofanotaura hunhu hunotevera. Kudzidzira kunowedzera mukana wekutevedzana kwemavara; yakarongwa-sampling trick inoderedza chitima/bvunzo kusawirirana.
Strategic Impact
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.
Ramangwana Rekuteerera Pinda uye Zviperengo
LAS ikozvino yave nhoroondo, asi DNA yayo inomhanya kuburikidza neyazvino ASR system. Yayo yekutarisisa-yakavakirwa encoder-decoder pfungwa yakashanduka kuita Transformer uye Conformer inoziva, nepo nzira dzine hukama seRNN-Transducer simba pa-chishandiso dhikisheni. Masisitimu emangwana anoenderera mberi neino yekupedzisira-kusvika-kumagumo trajectory, kusanganisa kuzivikanwa neshanduro uye kunzwisisa mune imwe chete mitauro yakawanda, uye kusundira kune yekutepfenyura, yakaderera-latency transcription iyo LAS, isiri-yekuyerera, yaisakwanisa kupa pakutanga.
Real-World Implementation
Kunyora Chirungu chinotaurwa zvakananga mumabhii pasina duramanzwi
Kushanda sehwaro hwepfungwa hwekutarisisa-kwakavakirwa izwi diction uye captioning masisitimu
Kuratidza kudzidziswa kwekupedzisira-kusvika-kumagumo kwezvidzidzo zvekutaura-kuzivikanwa kosi uye mabhenji
Mamodheru anokurudzira kutevedzana-kune-nhevedzano akazoshandiswa mukududzira matauriro
Njodzi & Guardrails
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.
Implementation Roadmap
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
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.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
Gaidhi rinotevera
Music Auto-Tagging
Mibvunzo inowanzo bvunzwa
What is Listen Attend and Spell?
Teerera, Teerera uye Spell (LAS) inzvimbo ye2015 neural network inonyora mutauro zvakananga mumabhii, isina duramazwi rakagadzirwa nemaoko kana modhi yemutauro wakasiyana. Yakaratidza kuti imwe chete-yekupedzisira-yekupedzisira modhi inogona kuita kutaurwa kwekutaura.
Ndezvipi zvikamu zviviri zvakakosha zvemuenzaniso weLAS?
LAS ine 'Mutereri' encoder inogadzirisa odhiyo uye 'Speller' attention-based decoder inoburitsa mavara.
Chii chinonzi Speller muLAS inobuda?
Iyo Speller idecoder inogadzira iyo transcript hunhu nehunhu, kudzidza zviperengo zvakananga.
Sei LAS encoder ichinzi 'pyramidal'?
Imwe neimwe nhanho yepiramidi BLSTM inopedza kureba kwekutevedzana, kupfupisa iyo refu yekuteedzera kutevedzana kuti kutarisisa kugoneke.
Ndechipi chikamu chekare chinodiwa neLAS zvakanyanya HAZVINODA?
Kusiyana nemapaipi ekare eHMM-DNN, LAS inodzidza zvakananga kubva kune odhiyo-kune-mavara maviri maviri pasina duramazwi refoni kana kumanikidza kurongeka.
Ndeipi michina inoita kuti Speller atarise pane yakakodzera chikamu cheodhiyo kune yega yega hunhu?
Meshini yekutarisisa inounganidza huremu pamusoro peiyo encoder nyika, ichigadzira mamiriro vector iyo decoder inoshandisa padanho rega rega.