Yenza Umbhalo Okhiqizwe yi-AI Ube Nobuntu nge-Pipeline ye-text-Humanizer Evulekile

Yini i-text-Humanizer?

i-text-Humanizer iphrojekthi ye-Python yamahhala, enelayisensi ye-MIT, yokuguqula umbhalo okhiqizwe yi-AI ube ukubhala okunezakhi zemisho zemvelo, amaphethini olimi ahlukahlukene, kanye nokwehlukahlukana kwesitayela okukhulu. Isebenzisa inqubo yokubhala kabusha nokuhumusha yezilimi eziningi, eklanyelwe ukugcina incazelo yombhalo wokuqala.

Iphrojekthi isekela amakhodi ezilimi en, ja, zh, ko, de, fr, kanye ne-es. I-README yayo iphinde ithi okuqukethwe okubhalwe kabusha kungadlula izitholi eziningi ze-AI. Nokho, imiphumela yokuthola ayiqinisekisiwe, futhi ithuluzi kufanele lisetshenziswe ngokuzibophezela, hhayi ukuze kuhlanekezelwe ubunikazi bombhali noma kugwenywe izidingo zokudalula.

Izici ezibalulekile

• Ukulondolozwa kwencazelo: Kubhala kabusha okuqukethwe kuzanywa ukugcina incazelo yakhona yasekuqaleni.
• Ukusetshenziswa kolimi lwemvelo: Kwethula izakhiwo zemisho, izinkulumo, kanye nokugeleza kolwazi okuhlukahlukene.
• Ukucutshungulwa kwezilimi eziningi: Kwamukela futhi kukhiqize okuqukethwe ngezilimi eziningana ezisekelwayo.
• Ukuguquguquka kwesitayela: Kwenziwe ukuze kulungiswe ukubhalwa kabusha kuye ngamathoni nezitayela zokubhala ezahlukene.
• Isigaba sokuzikhethela se-DeepL: Kwengeza enye injini yokuhumusha uma ukhiye we-DeepL API ulungisiwe.
• Ukufinyelela komthombo ovulekile: Kusatshalaliswa ngaphansi kweLayisensi ye-MIT.

Indlela inqubo yokubhala kabusha esebenza ngayo

Esikhundleni sokusebenzisa ukudlulisa okukodwa kokuchaza kabusha, i-text-Humanizer idlulisa okuqukethwe ezigabeni eziningana zamamodeli olimi nezokuhumusha.

1. Ukubhala kabusha nge-DeepSeek: I-DeepSeek idala inguqulo enencazelo efanayo, kodwa enamagama, isakhiwo semisho, kanye nokugeleza kolwazi okushintshiwe. Inqubo iphinde isebenzise isiShayina njengendlela emaphakathi yokwethula ukuhlukahluka kwesakhiwo.
2. Ukuhumushela esiTurkey: I-Google Translate iguqula umphumela ophakathi uwuyise esiTurkey, ingeze ukuhlukahluka kokwakheka kwemisho phakathi kwezilimi.
3. Ukuhumusha okukhethwayo kwesiJapane: Uma ukhiye we-DeepL API utholakala, i-DeepL ihumushela umbhalo wesiTurkey esiJapane. Lokhu kwethula injini yesibili yokuhumusha kanye nokuhlukahluka okwengeziwe kolimi.
4. Ukwakhiwa kabusha kokugcina: I-DeepSeek ihumusha futhi yakhe kabusha umphumela ngolimi lokuqala olulungisiwe, isuse izinsalela zokuhumusha futhi ibuyisele ukugeleza kwemisho okufundeka kahle.

Ngenxa yokuthi kuhileleke izinguquko eziningi, buyekeza umbhalo wokugcina ngokucophelela. Amagama abantu, amagama obuchwepheshe, izinombolo, izingcaphuno, kanye nomehluko ocashile kungadinga ukuqinisekiswa ngesandla.

Izidingo zangaphambi kokufakwa

Ngaphambi kokufaka iphrojekthi, lungiselela okulandelayo:

• Python 3.6 noma eyakamuva.
• Git yokukopisha ikhosombe.
• Ukhiye we-DeepSeek API.
• Ukhiye we-DeepL API ongakhethwa uma ufuna ukunika amandla isigaba sokuhumusha esengeziwe.

Ukufakwa

1. Kopisha ikhosombe

git clone https://github.com/SpaceDudem/text-humanizer.git
cd text-humanizer

2. Faka okuncikile kwe-Python

pip install -r requirements.txt

Uma indawo yakho isebenzisa umyalo ohlukile we-Python 3, sebenzisa isifaki samaphakheji esihambisanayo, njenge-pip3.

3. Dala ifayela lokumisa

Ku-Windows, kopisha ukumisa okuyisibonelo usebenzisa:

copy .\config\config.example.toml .\config.toml

Ku-Linux noma ku-macOS, sebenzisa:

cp ./config/config.example.toml ./config.toml

Lokhu kudala ifayela le-config.toml elisezingeni eliyimpande, elisetshenziswa uhlelo.

Lungisa i-text-Humanizer

Vula i-config.toml bese usetha izinketho ezidingwa umhlinzeki wakho we-API nolimi oluqondiswe kulo.

• target_language: Ulimi olusetshenziselwa okokufaka nomphumela. Amakhodi atholakalayo abalwe yiphrojekthi yi-en, ja, zh, ko, de, fr, kanye ne-es.
• deepseek_api_key: Ukhiye wakho odingekayo we-DeepSeek API.
• deepl_api_key: Ukhiye we-DeepL API ongakhethwa wesigaba sokuhumusha sisuka esiTurkey siya esiJapane.
• base_url: Ukushintshwa kwe-URL eyisisekelo ye-API. Yiyeke ingenalutho ukuze kusetshenziswe okuzenzakalelayo komhlinzeki.
• model: Igama elifunekayo lemodeli. Yiyeke ingenalutho ukuze kusetshenziswe okuzenzakalelayo komhlinzeki.
• temperature: Ilawula ukuhlukahluka komphumela we-LLM. Iphrojekthi incoma 1.3.

Ukumisa okucatshangwayo kungabukeka kanje:

target_language = "en"
deepseek_api_key = "YOUR_DEEPSEEK_API_KEY"
deepl_api_key = ""
base_url = ""
model = ""
temperature = 1.3

Sebenzisa ukwakheka okunembile osekuvele kukhona ku-config.example.toml. Faka esikhundleni ukhiye oyisibonelo ngemininingwane yakho yangempela, futhi ungayifaki i-config.toml ekomitini.

Qalisa i-humanizer

Qalisa uhlelo usuka enkombeni yekhosombe:

python main.py

Uma kudingeka, sebenzisa python3 main.py esikhundleni salokho. Faka umbhalo ngesixhumi esibonakalayo esivezwa uhlelo olusebenzayo, bese uvumela inqubo emisiwe ukuthi yenze izigaba zayo zokubhala kabusha nokuhumusha.

Isibonelo sokusetshenziswa okuyisisekelo

Enqubweni yesiNgisi, setha i-target_language ibe yi-en, faka ukhiye ovumelekile we-DeepSeek, bese usebenzisa i-main.py. Ungabe usucubungula umusho oqinile onjengalo:

Ukuqaliswa kwalolu hlelo kunikeza abasebenzisi ikhono lokubhala kabusha okuqukethwe ngendlela yemvelo kakhudlwana.

Inqubo kufanele ibuyise inguqulo enencazelo ecishe ifane, kodwa enamagama nokugeleza komusho okushintshiwe. Umphumela oqondile awunqunywa kusengaphambili futhi uncike kwimodeli ekhethiwe, ku-temperature, ezigabeni zokuhumusha, nasekuziphatheni kwe-API.

Amathiphu okucushwa okuthuthukile

Vula isigaba sokuzikhethela se-DeepL

Faka inani elivumelekile le-deepl_api_key ukuze ufake isinyathelo sokuhumusha sisuke esiTurkey siye esiJapane. Yishiye ingenalutho uma ukhetha iphayiphi elingenayo i-DeepL. Qhathanisa zombili izindlela embhalweni omele isimo, ngoba esinye isigaba sokuhumusha singandisa ukuhlukahluka kolimi futhi sidale nemininingwane eminingi okufanele iqinisekiswe.

Lungisa ukuhlukahluka kokuphumayo

I-README incoma temperature engu-1.3. Lesi silungiselelo sihlose ukukhuthaza amagama ahlukahlukene. Uma uzama amazinga ahlukene okushisa, shintsha into eyodwa ngesikhathi bese uqhathanisa ukuthi umphumela usawagcina yini amagama, amaqiniso, namagama aqondene nesizinda.

Khetha imodeli noma isiphetho sangokwezifiso

Setha i-model uma udinga i-slug ethile yemodeli esekelwa umhlinzeki. Setha i-base_url kuphela uma udinga ukubhala phezu kwesiphetho se-API esizenzakalelayo. Kungenjalo, shiya womabili amanani engenalutho ukuze kusetshenziswe okuzenzakalelayo komhlinzeki.

Hlola okuqukethwe okude noma okukhethekile ngezigaba

Ezintweni zobuchwepheshe, zomthetho, zezemfundo, noma ezigcwele idatha, buyekeza izigaba ezincane ngazinye. Ukubhalwa kabusha kwezigaba eziningi kungathinta amagama anembayo noma ubudlelwano, ngisho nalapho inhloso kuwukugcina incazelo injalo. Qhathanisa isigaba ngasinye sokugcina nomthombo waso ngaphambi kokushicilela.

Lungisa izinkinga zokucushwa

• Qinisekisa ukuthi imiyalo isebenza isuka kufolda ye-text-humanizer oyiklone.
• Hlola ukuthi i-config.toml ikhona endaweni elindelekile yempande.
• Qinisekisa ukuthi i-deepseek_api_key iqukethe ukhiye ovumelekile.
• Qinisekisa ukuthi i-target_language isebenzisa eyodwa yamakhodi ezilimi asekelwayo.
• Uma isigaba sokuzikhethela sehluleka, hlola i-deepl_api_key noma ushiye yona ingenalutho okwesikhashana.
• Uma usebenzisa izilungiselelo ze-API zangokwezifiso, qinisekisa amanani e-base_url ne-model.

Isiphetho

i-text-Humanizer ihlanganisa ukubhalwa kabusha yi-DeepSeek nokuhumusha phakathi kwezilimi ukuze ikhiqize umbhalo ohluke kakhulu futhi ozwakala ungokwemvelo, ngesikhathi izama ukugcina incazelo yokuqala. Ngemva kokufaka okuncikile nokulungisa okhiye be-API, ungaqalisa ukuhamba komsebenzi nge-python main.py. Thatha okukhiphayo njengohlaka lokuqala, qinisekisa imininingwane ebalulekile, futhi usebenzise iphrojekthi ngokuhambisana nemithetho esebenzayo yobunikazi nokudalula.