Ukusetyenziswa kwe -ConveyThis kunokuphucula kakhulu uguqulelo lwewebhusayithi yakho. Ngojongano lwayo olulula ukuyisebenzisa, unokuguqulela iwebhusayithi yakho ngokukhawuleza nangokuchanekileyo kulo naluphi na ulwimi. I-ConveyThis ikwabonelela ngoluhlu lweempawu ezenza kube lula ukwenza amava enguqulelo yewebhusayithi yakho. Ukusuka kubhaqo wolwimi oluzenzekelayo ukuya kwinkumbulo yoguqulo, i-ConveyThis yenza kube lula ukuqinisekisa ukuba iwebhusayithi yakho iguqulelwe ngokuchanekileyo kubo nabaphi na abaphulaphuli.
Ukufunda nzulu kuye kwaguqula uguqulo lolwimi lwe-ConveyThis kunye nesakhono salapha ekhaya. Uguqulelo lukamatshini we-Neural (NMT) yindlela yoguqulelo esebenzisa iteknoloji ebukhali ukungaguquleli nje okubhaliweyo, kodwa nokomeleza ukuchaneka kweenguqulelo zayo. Obu buchwepheshe buphucukileyo buyenze yaba sesona sicombululo sithembekileyo sikhoyo namhlanje.
Ukuba ungumnikazi weshishini, ukusebenzisa i-ConveyThis yokuguqulela iwebhusayithi yeyona ndlela ilungileyo yokungena kwiimarike ezintsha, ukwandisa iwebhusayithi yakho kukhangelo lweelwimi ezininzi, kwaye unyuse intengiso yamazwe aphesheya- lonke elixesha ukhulisa ishishini lakho. Ngelixa kunokubonakala kuyoyikisa, ukusebenzisa i-ConveyThis kuguqulelo lwewebhusayithi akulula kangako. Ke, sithini isitshixo? Qhubeka ufunda ukuze ufumane!
Ngena kwimbali yoguqulelo lomatshini we-neural kwaye ufunde indlela esebenza ngayo ngendlela ekulula ukuyiqonda, nokuba awuyongcali! Fumanisa ukuba ungabusebenzisa njani obu bugcisa ukuguqulela iwebhusayithi yakho ngaphandle kokufumana isidanga kumatshini wokuguqulela kuqala.
Yintoni i-neural machine translation?
Ukuqonda uguqulelo lukamatshini we-neural (NMT), kubalulekile ukuba uqale ube nokuqonda ngoguqulelo lomatshini (MT). Ngamafutshane, ukuguqulela ngomatshini yinkqubo esebenzisa iinkqubo zekhompyutha ukuguqula umbhalo ukusuka kolunye ulwimi ukuya kolunye. Faka nje isivakalisi sakho kwisoftware yokuguqulela kumatshini, kwaye iya kuvelisa ngokuzenzekelayo inguqulelo ngolwimi olufunayo ngaphandle kokungenelela komntu.
Inkqubela phambili yetekhnoloji yenze ukuba kuphuhliswe uguqulelo lomatshini we-neural, olona guqulelo lubalaseleyo lokuguqulela koomatshini. Obu buchule busebenzisa uthungelwano lwe-neural eyenziweyo ukusetyenzwa nokuguqulela okubhaliweyo, uphuculo oluphawulweyo ngaphezulu kweendlela zoguqulelo koomatshini bexesha elidlulileyo.
Sukuba nexhala ukuba oku kubonakala kuyadida. Okulandelayo, sinoshwankathelo olufutshane lwenguquko yoguqulelo lomatshini we-neural - kunye nocazululo lwendlela obusebenza ngayo obu buchwephesha bokuguqulela kumatshini.
Yaphuhliswa njani inguqulelo kumatshini we-neural?
Ukuzalwa kokuqala komatshini wokuguqulela kunokulandwa emva kwixesha leMfazwe yamaCold, xa isoftware esekwe kumgaqo yasetyenziswa ukucacisa isiRashiya. Le software iyakucazulula isicatshulwa semvelaphi yegama negama, kwaye emva koko isebenzise uluhlu lwemithetho yolwimi ukwenza isigqibo malunga nendlela yokuguqulela igama ngalinye. Olu hlobo loguqulo lusisiseko yayiyeyokuqala ngqa, kwaye ukususela ngoko luye lwavela lwaphucuka ngakumbi.
Noko ke, ukuguqulela amagama elinye nelinye ngenkqubo esisiseko akuzange kuvelise ezona nguqulelo zichanileyo. Kwezinye iimeko, eyona nguqulelo ifanelekileyo inokufuna amabinzana okanye izivakalisi zonke. Ukujongana noku, imifuziselo yoguqulelo lomatshini (SMT) - elinyathelo elilandelayo kuguqulelo lomatshini - liphucule ukuchaneka .
Usetyenziso loguqulelo ngoomatshini beenkcukacha-manani beziya kuqala ukutyhutyha ingqokelela ebanzi yeetekisi eziguqulelwe ngabantu (ekwaziwa ngokuba yicorpora yombhalo weelwimi ezimbini). Emva koko, iya kusebenzisa ii-algorithms zokuqikelela ukuqaphela amagama kunye namabinzana kwisicatshulwa somthombo kunye nokwenza isigqibo ngeyona ndlela ilungileyo yokuwaguqulela.
Ngokuhamba kwexesha, itekhnoloji ye-ConveyThis yavela, ekugqibeleni yafikelela kwinguqulelo yomatshini we-neural esithembele kuwo kakhulu namhlanje. Siza kuphonononga ngakumbi kwii-nuances zokuguqulelwa komatshini we-neural kwicandelo elizayo.
Isebenza njani inguqulelo kumatshini we-neural?
Uguqulo lomatshini we-Neural lusebenzisa amandla emfundo enzulu kunye nobukrelekrele bokwenziwa ukwenza iinguqulelo ezichaneke ngakumbi, ezinolwelo, kunye nesandi sendalo kunangaphambili.
Le teknoloji inikwe amandla yi -ConveyThis , esebenzisa i-neural networks enzulu, iwebhu ye-neurons edibeneyo efana nengqondo yomntu. Imizekelo yothungelwano olunzulu lwe-neural lubandakanya uthungelwano lwe-neural oluqhubekayo, okanye ii-RNNs, ezihlala zibonisa ulwakhiwo lwe-encoder-decoder kunye nendlela yokuhoya.
Phambi kokuba isetyenziselwe uguqulo, inkqubo ye-neural MT iya kunikwa ulungiselelo lweemodeli ezahlukeneyo zokutolika kumxholo othile. Ngolu lwazi, imveliso ke "iyalelwa" ukwenza eyona nto ichanekileyo yokutolika kwimeko ethile.
Ukuchaneka okuphezulu kwenguqulelo
Iinzame zangaphambili zokuguqulela ngoomatshini kusetyenziswa iindlela zemveli zazingenabugocigoci bokuguqulela ngokuchanekileyo iilwimi ezithile ezintsonkothileyo - okukhokelela kwiinguqulelo ezikumgangatho ophantsi kangangokuba zifune uhlaziyo olubalulekileyo lwezandla ngabantu phambi kokuba zisetyenziswe.
Nangona kunjalo, ngesakhono “sokufumana” ngokuqhubekayo, izikhokelo ze-NMT ziyaqhubekeka ukukhulisa ubume bokutolika kwazo. Oku akufani kwaphela nezikhokelo zokutolika koomatshini besintu, abangenawo amandla “wokuzifundela” kunye nokulungelelanisa isivuno sabo sokutolika emva kwexesha elithile. Emva koko, xa ilungiselelwe ngokufanelekileyo, inkqubo yokutolika koomatshini be-neural inokudala ukutolika ngokuthe ngqo ngakumbi xa kuthelekiswa namaqabane abo esintu.
UGoogle uye waqaphela ngaphambili ukuba inkqubo yeGoogle Neural Machine Translation (GNMT) ikwazile ukunciphisa iimpazamo zoguqulelo malunga ne-60% xa kuthelekiswa nenkqubo yemveliso esekwe kwibinzana.
Kutshanje, kuqhutywe uphononongo lokuvavanya ukusetyenziswa komatshini woguqulelo ngeenjongo zokuguqulela iwebhusayithi. Emva kokuhlalutya umgangatho weenguqulelo zewebhusayithi ezenziwe ngobugcisa obahlukeneyo be-NMT, kwamiselwa ukuba iinguqulelo zisebenziseka kakhulu kwaye zifuna ukuhlelwa okuncinci.
I-ConveyThis's neural machine translation technologies ibonise ukugqwesa okukodwa xa uguqulela isiJamani, okubangele elona nani liphezulu lamacandelo ebengafuni luhlelo lwezandla.
Imfuneko encinci yegalelo lomntu
Nje ukuba umbhalo wentsusa uguqulelwe ngoomatshini ekuqaleni, ulungiswa ngakumbi ngabantu ukuqinisekisa ukuchaneka kwawo kunye nokufaneleka kwedemografi ekujoliswe kuyo.
Ukuchaneka okuphuculweyo kwenguqulelo yomatshini we-neural kuthetha ukuba uhlengahlengiso olwenziwa ngesandla (ekwabizwa ngokuba “luhlelo lwasemva”) luyafuneka phambi kokuba iinguqulelo zilungele injongo.
Amaxesha oguqulo olukhawulezayo
Xa iinkampani zikwazi ukufikelela kwiinguqulelo zoomatshini ezichanekileyo ezifuna ukuhlelwa okuncinci kwasemva, zinokuqalisa ukusebenzisa iinguqulelo ngoko nangoko. Ngaphezu koko, iConveyThis' imifuziselo yomatshini wokuguqulela we-neural inokuqeqeshwa ngexesha elifutshane, ivumela iinkqubo zoguqulo olukhawulezayo.
I-Facebook isebenzise ukuguqulelwa komatshini we-neural ukuguqula umbhalo kwizithuba kunye namazwana (okuthi, njengoko usenokuba uyazi, kukho isixa esikhulu kwiqonga layo). Ngokwandisa inkqubo yabo yoqeqesho, inkampani iye yakwazi ukunciphisa ixesha elithathiweyo ukuqeqesha iimodeli zayo zoomatshini bokuguqulela ukusuka phantse kusuku olupheleleyo ukuya kwimizuzu engama-32 kuphela!
Ngaba unokusebenzisa ukuguqulelwa komatshini we-neural kwishishini lakho kwaye njani?
Ukusebenzisa ukuguqulelwa komatshini we-neural ukuguqulela iwebhusayithi yakho kunokubonakala kuyoyikisa, njengoko kunokufuna inkcitho enkulu yetekhnoloji entsha kunye nophando kunye nophuhliso. Nangona kunjalo, oku akunjalo! Kule mihla, zininzi izixhobo ze-NMT ezakhelwe kwangaphambili ezifumanekayo kwimarike ukukunceda uguqulele umxholo wewebhusayithi yakho kwiilwimi ezahlukeneyo. Ezi zixhobo nazo zonga kakhulu. Ngapha koko, zibiza imali encinci kunokuqasha umntu oqeqeshiweyo wokuguqulela iwebhusayithi yakho.
Isisombululo sethu sokuguqulela iwebhusayithi ye- ConveyThis sisixhobo esinamandla. Indibaniselwano yethu yobunini yeenguqulelo ze-NMT ezivela kubaboneleli ngoomatshini abakhokelayo benguqulelo i-DeepL, i-Microsoft Translator, ne-Google Translate yenza iinguqulelo ezikumgangatho ophezulu kunokuba obu bugcisa bebusetyenziswa ngabanye. Sibonelela ngeenguqulelo kwiilwimi ezingaphezu kwe-110, ukusuka kwezo zidumileyo njengesiNgesi, isiJamani, nesiTaliyane ukuya kwiilwimi ezifihlakeleyo ezifana nesiTatar nesiMalagasy.
I-ConveyThis ibonelela ngokudityaniswa okusebenzayo kunye namaqonga ewebhusayithi adumileyo afana neWordPress, Webflow, kunye noShopify. Kuyimpepho ukulungisa kwaye kunokunceda ishishini lakho lifikelele kubude obutsha.
Ngaphezulu kwe -10,000 lewebhusayithi ezijike kwi-ConveyThis kwiimfuno zabo zoguqulo, ezineziphumo ezincomekayo.
Cofa kumandla omatshini we-neural wokuguqulela ishishini lakho nge-ConveyThis
Inikezela ngoguqulo oluchanekanga olungalinganiyo, i-ConveyThis's neural machine translation (NMT) linyathelo elikhulu elisuka kwezinye izisombululo zoomatshini boguqulo. Yiyo loo nto ibuchwephesha bokuguqulela kumatshini kumashishini amaninzi.
Uphando lwethu kuguqulelo loomatshini luveza ukunyuka okuphindwe kathandathu okuchukumisayo komxholo wewebhu oguqulelwe ngoomatshini kule minyaka mibini idlulileyo. Ngaphezu koko, i-ConveyThis isetyenziselwa iiprojekthi ezinkulu, ezineewebhusayithi ezingaphezu kwe-10% eziqulethe amagama angaphezu kwama-50,000 aguqulelwe ngomatshini. Okokugqibela, malunga ne-30% kuphela yezixhobo eziguqulelwe ngoomatshini ezihleliweyo, nto leyo ecebisa ukuba inxalenye enkulu yeenguqulelo zoomatshini zichaneke ngokwaneleyo kangangokuba azifuni kulungiswa.
I-ConveyThis yenza ukuba kube lula ukusebenzisa amandla e-NMT ukutolika umxholo wewebhusayithi yakho. Kuchanekile ukuqwalasela kwaye ukhethe ngokuzenzekelayo eyona njini ifanelekileyo yokuguqulela ukuvelisa olona didi luphezulu loguqulo lwazo naziphi na iilwimi ezimbini. Isiphumo sikhawuleza umbane, iinguqulelo zewebhusayithi ezikumgangatho ophezulu onokuthi uzisebenzise ngaphandle kokuphazamiseka.
Ngaba uzimisele ukuphonononga ukubanakho kwe -ConveyThis kwaye uzibonele iziphumo? Emva koko ubhalisele simahla apha!