Rabewar Dan lokaci Mai Haɗin kai
Connectionist Temporal Classification (CTC) shine aikin asara da hanyar yanke hukunci wanda ke barin cibiyoyin sadarwa na jijiyoyi su juya jerin sauti mai tsawo zuwa rubutu ba tare da kowa ya daidaita kowane sauti da hannu zuwa kowane harafi ba.
Dubawa
It made end-to-end speech recognition practical by solving the brutal alignment problem.
Zurfafa nutsewa
Magana ba ta da kyau: kalmar 'sannu' na iya wuce firam ɗin sauti 40, kuma babu wanda ya yi wa lakabin ainihin firam ɗin 'h'. CTC, wanda Alex Graves ya gabatar a cikin 2006, ya koma gefe. Cibiyar sadarwa tana fitar da yuwuwar akan haruffa (da alamar 'blank' ta musamman) ga kowane firam. Daga nan CTC ta fayyace ingantacciyar jeri a matsayin kowace hanya ta firam-by-frame wacce ta ruguje zuwa rubutun da aka yi niyya bayan ka'idoji guda biyu: hade haruffan da aka maimaita, sannan a share fage. Saboda taswirar hanyoyi da yawa zuwa rubutu iri ɗaya, CTC ta ƙididdige yuwuwar dukkansu ta amfani da algorithm mai ƙarfi-tsara (algorithm na gaba-baya) kuma yana horar da hanyar sadarwa don haɓaka wannan jimillar. Alamar blank ita ce dabarar wayo wacce ke barin samfurin ya ce 'babu wani sabon abu a nan' kuma ya raba maimaitawa na gaske kamar na biyu-L a cikin 'sannu'.
Fahimtar Fasaha
Babban zato na CTC shine yancin kai na sharadi: idan aka ba da sautin, ana hasashen fitowar kowane firam ɗin da kansa, ba tare da wani samfurin harshe da aka gasa a ciki ba. Wannan yana sa taƙaitawar gaba-baya za ta iya faɗowa amma yana nufin CTC tana ƙoƙarin samar da saɓo, abubuwan da aka fi so (mafi yawa, tare da ƙayyadaddun halaye) da fa'ida daga samfurin harshe na waje a lokacin yanke hukunci. Binciken bim tare da fused LM, wanda galibi ake kira prefix-beam decoding, da matuƙar inganta daidaito akan ƙirƙirawar argmax.
Dabarun Tasiri
Shiga ku isa
Yana inganta samun dama ta hanyar rubutu, ba da labari, da mu'amalar murya.
Kudin da kasafin kuɗi
Ƙungiyoyin kafofin watsa labaru na iya jigilar sauti mai gogewa cikin sauri tare da ƙaramin kasafin kuɗi.
Gudu da sikelin
Tsarin fuskantar abokin ciniki na iya aiwatar da hulɗar magana a mafi girman ma'auni.
Makomar rarrabuwa na ɗan lokaci na Connectionist
CTC ya kasance dokin aiki, musamman inda yawo da ƙarancin latency, kuma ana ƙara amfani da shi azaman asarar taimako tare da hankali ko manufofin transducer a cikin ƙirar 'CTC/hankali' matasan. Yi tsammanin CTC ta dage a matsayin reshe mai sauri, mai sauƙi mai sauƙi a cikin tsarin maganganun multitask, kuma a matsayin injin daidaitawa a bayan kayan aikin daidaita-tsaye waɗanda ke buga kalmomi. Ƙididdiga masu kulawa da kai kamar wav2vec 2.0 yawanci ana daidaita su tare da shugaban CTC.
Aiwatar da Gaskiyar Duniya
Kyakkyawan daidaitawa wav2vec 2.0 tare da shugaban CTC don gina ƙirar buɗaɗɗen magana-zuwa rubutu a cikin ƙananan kayan aiki
Samar da tambarin matakin kalma- da lambar waya don ƙaraoke ta hanyar daidaitawar CTC ta tilastawa
Taken ainihin lokaci akan na'urar inda samfurin CTC mai yawo ke rubutowa tare da ƙaramin latency
Gane rubutun hannu, inda CTC ke karanta layin lanƙwasa ba tare da an riga an raba haruffa ɗaya ba
Hatsari & Tsare-tsare
Rashin amfani da murya da haɗarin kwaikwaya yana ƙaruwa lokacin da aka rasa izini.
Daidaituwa na iya faɗuwa cikin lafuzza, yaruka, ko mahalli masu hayaniya.
Ana iya kuskuren sauti na roba don ingantacciyar magana ba tare da bayyananniyar lakabi ba.
Taswirar Hanya
Sami tabbataccen izini don ɗaukar murya, cloning, da sake amfani.
Gwajin ingantattun masu magana daban-daban da yanayin baya.
Ƙayyade lokacin da dole ne ɗan adam ya duba ko ya amince da abubuwan da aka fitar.
Yi lakabin sauti na roba da kuma adana bayanan da aka tabbatar don yin lissafi.
Ci gaba da Bincike
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 Connectionist Temporal Classification 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
Jagora na gaba
Rarraba Salon Kiɗa
Tambayoyin da ake yawan yi
What is Connectionist Temporal Classification?
Connectionist Temporal Classification (CTC) shine aikin asara da hanyar yanke hukunci wanda ke barin cibiyoyin sadarwa na jijiyoyi su juya jerin sauti mai tsawo zuwa rubutu ba tare da kowa ya daidaita kowane sauti da hannu zuwa kowane harafi ba. Ya sanya fahimtar magana ta ƙarshe zuwa-ƙarshe mai amfani ta hanyar warware matsalar daidaitawa.
Wace babbar matsala ce aka ƙera CTC don warwarewa don gane magana?
CTC yana barin hanyar sadarwa ta horar da nau'i-nau'i (audio, rubutu) nau'i-nau'i ba tare da kowa ya yi wa lakabin firam ɗin da ya dace da wane hali ba, yana magance matsalar jeri.
Menene alamar 'blank' ta musamman a cikin CTC da ake amfani da ita?
Alamar blank tana barin ƙirar ta fitar da 'babu wani sabon abu' akan firam kuma, mai mahimmanci, ya raba maimaitawar gaskiya kamar sau biyu-L a cikin 'sannu' daga maimaitawar rugujewa.
Ta yaya CTC ke ƙididdige asarar don rubutun da aka yi niyya?
CTC tana amfani da algorithm na gaba-baya mai ƙarfi-tsara don taƙaita yuwuwar akan kowane jeri da taswira zuwa manufa bayan haɗa maimaitawa da share fage.
Menene ka'idojin rugujewar guda biyu da CTC ke amfani da su don juya hanyar firam zuwa rubutu na ƙarshe?
Ana ƙaddamar da ɗanyen hanyar kowane-frame ta hanyar fara haɗa haruffan kwafin haruffa sannan kuma cire duk alamun da ba komai.
Menene sassauƙan zato daidaitaccen CTC yayi game da abubuwan da yake samarwa?
CTC tana ɗaukar yancin kai na kowane-frame, wanda ke sa taƙaitaccen bayani amma yana nufin babu ginanniyar ƙirar harshe.