Imiphumela ibukeka kanjani
Izinga eliphakeme lokunemba: iMandarin ecacile iphuma ihlanzekile; cabanga ukuhlela okuningi kumandla endawo noma kumazwi athinta iCanton.
I-Mandarin, ne-Canton caveat ebalulekile
Le khasi lifakwe ku-Mandarin (zh) — 普通话/国语 njengoba kukhulunywa eChina, eTaiwan, eSingapore, naseMalaysia. I-Standard Mandarin iguqula ngokuphezulu, kufaka phakathi i-Mandarin ebhalwe phezulu evela kulabo abakhuluma ngesi-Southern.
I-Cantone AYILUNGI ukuba yi-Mandarin ngoba idinga ukuphawula kwezwi — iyisilimi esihlukile esinamakhodi ayo (yue), i-vocabulary yayo, nefomu layo elibhalweyo. Ukufaka i-Cantone okuthunyelwe ngalesi khasi kuzodala umbhalo we-garbage-to-mediocre Mandarin. Sizozama ukungangenisi ikhasi le-Cantone ngoba umphumela wokusebenza awuhlangabezani ne-bar yethu; uma ukhuluma i-Cantone, cabanga ukubheka i-Cantone ekhethekile.
I-Output elula, ukuphawula, nokuxoxwa okuhlanganisiwe
Ama-transcripts akhiqizwa ngezinto ezilula ngokuzenzakalela (i-convention yokuqeqesha enamandla) nge-Chinese full-width punctuation (,。?) efakwe — i-Chinese ibhalwe ngaphandle kwezinhlamvu futhi i-output ilandela lokho. Abakhuluma i-Taiwan Mandarin kufanele baqaphele ukuthi i-transcript izobe ilula, hhayi ejwayelekile; ukushintsha iskripthi kuyinto encane-cofa-kanye ezindaweni eziningi. Ama-English words and brand names inside Mandarin speech are transcribed inline in Latin script.
Imibuzo ebuzwa kaningi
Ingabe lokhu kusebenza nge-Cantone?
Akuna - futhi kubalulekile ukwazi ukuthi kungani: i-Canton (yue) iyisilimi esihlukile ku-Mandarin (zh), hhayi isilimi esisetshenziswa kakhulu. Le khasi lisebenzisa i-Mandarin. Umsindo we-Canton uzophuma kahle. Imodeli engezansi inekhono elithile le-Canton ngaphansi kwekhodi yayo ehlukile ye-yue, kodwa asiyicabangeli iyinto engcono kakhulu ukuthutha ikhasi.
Izimpawu ezilula noma ezijwayelekile kwi-output?
Ilula, njengemithetho — lokhu kuyisimo esiyinhloko emodelini ye-Mandarin yokuqeqesha idatha, ngisho nalabo abakhuluma isiTaiwan. Uma ufuna amaphawu ajwayelekile, sebenzisa i-TXT ekhishwayo nganoma yisiphi isiguquli esilula→sijwayelekile; ukuguqulwa kunzima kakhulu kuleyo ndlela.
Ingabe ukudluliswa kweMandarin kulungile?
Izinga eliphezulu — iMandarin ecacile (ukusabalalisa, izinhlanganiso, izifundo) ibhala ngokunethezeka, kufaka phakathi ukuchaza okuphathelene nethoni, ngoba imodeli yesilimi ikhetha amagama kusuka kusimo. Ama-accents anamandla asendawoni (iMandarin eSichuan-flavored, iMandarin eTaiwan-accented) asebenza; ama-languages asendawoni akhona (iShanghainese, iHokkien) angeke.
Umsebenzi we-punctuation ne-spacing usebenza kanjani encwadini?
Isi-Chinese sibhalwe ngaphandle kwezithuba, futhi isingeniso silandela imithetho, nge-full-width punctuation (,。!?、) efakwe emazingeni ajwayelekile. I-segment/SRT cue ihlukanise izwe emazingeni ajwayelekile, ngakho-ke izihloko zifunda ngokujwayelekile ngaphezu kokungqubuzana phakathi kwegama.
Umsindo we-Mandarin uguqulwe ube yi-English text — kungenzeka?
Lekhasi liguqula iMandarin libe yimibhalo yaseChina; lokhu kuyingxenye ephakeme kakhulu yokuqonda. E-English, thumela ngaphandle i-TXT futhi uguqule i-machine — zh→en ukuguqulela ku-text clean kunamandla, futhi imiphumela ye-step-two idlula ukuguqulela kwezwi elilodwa. Ukuguqulwa kwe-transcript efakwe ku-intended.
Yini mayelana namagama ase-English axhunywe ku-Mandarin?
Okujwayelekile kakhulu etekhnoloji nebhizinisi Mandarin, futhi kuphathwa kahle: Izilimi zase-English, ama-acronyms, nama-brand names abhalwe ngokuzenzakalela ku-Latin script (isibonelo, 我们用 Kubernetes 部署). I-Long English-only passages transcribe better via the auto-detect uploader on the home page.