Okuqukethwe[Fihla][Bonisa]
Sinekhono lemvelo lokubona nokuhlukanisa amagama abantu ngabanye, izindawo, izindawo, amanani, nokuningi noma nini lapho sizwa noma siwafunda. Abantu bayakwazi ukuhlukanisa amagama ngokwezigaba, ukukhomba, nokuqonda amagama ngokushesha.
Isibonelo, ungakwazi ukuhlukanisa into bese usheshe uthole okungenani izimfanelo ezintathu kuya kwezine uma uzwa igama elithi "Steve Jobs,"
- Umuntu: "Steve Jobs"
- Inhlangano: “Apple”
- Indawo: "California"
Njengoba amakhompiyutha engenalo leli khono azalwa nalo, kumelwe siwasize ekuboneni amagama noma umbhalo nokukuhlukanisa. I-Entity Entity Recognition (NER) isetshenziswa kulesi simo.
Kulesi sihloko, sizohlola i-NER (Eqanjwe Ukuqashelwa Kwebhizinisi) ngokuningiliziwe, okuhlanganisa ukubaluleka kwayo, izinzuzo, ama-NER API aphezulu, nokunye okuningi.
Yini i-NER (Eqanjwe Ukuqashelwa Kwebhizinisi)?
Indlela yokucubungula ulimi lwemvelo (NLP) eyaziwa ngokuthi ukuqashelwa kwebhizinisi (NER), kwesinye isikhathi eyaziwa ngokuthi ukuhlonza ibhizinisi noma ukukhishwa kwebhizinisi, ibona ngokuzenzakalela amabhizinisi aqanjwe amagama embhalweni futhi iwahlukanise ngezigaba ezinqunywe kusengaphambili.
Amabhizinisi afaka amagama abantu ngabanye, amaqembu, izindawo, izinsuku, amanani, amanani amadola, amaphesenti, nokuningi. Ngokuqashelwa kwebhizinisi okunegama, ungakwazi ukukusebenzisa ukuze uqoqe idatha ebalulekile yesizindalwazi noma ukuze ukhiphe ulwazi olubalulekile ukuze uqonde ukuthi idokhumenti imayelana nani.
I-NER iyitshe legumbi lapho uhlelo lwe-AI luncike kulo ukuze kuhlaziywe umbhalo ukuze uthole i-semantics ehlobene nomuzwa, ngisho noma i-NLP imele intuthuko enkulu enqubweni yokuhlaziya umbhalo.
Yini ukubaluleka kwe-NER?
Isisekelo sendlela yokuhlaziya umbhalo yi-NER. Imodeli ye-ML kufanele iqale inikezwe izigidi zamasampuli anezigaba ezichazwe ngaphambilini ngaphambi kokuthi iqonde isiNgisi.
I-API ithuthuka ngokuhamba kwesikhathi ekuboneni lezi zingxenye emibhalweni eyifunda okokuqala ngqa. Amandla enjini yokuhlaziya umbhalo akhuphuka ngamandla e-NER namandla.
Njengoba kubonakala lapha, imisebenzi eminingi ye-ML iqaliswa yi-NER.
Ukusesha kweSemantic
Usesho lwe-semantic manje seluyatholakala ku-Google. Ungafaka umbuzo, futhi izozama ngakho konke okusemandleni ukuphendula ngempendulo. Ukuze uthole ulwazi, umsebenzisi alufunayo, abasizi bedijithali abafana ne-Alexa, i-Siri, i-chatbots, nabanye basebenzisa uhlobo lokusesha lwe-semantic.
Lo msebenzi ungashaywa noma ungaphuthelwa, kodwa kunenani elikhulayo lokusetshenziswa kwawo, futhi ukusebenza kwawo kukhuphuka ngokushesha.
Idatha Analytics
Lona umusho ojwayelekile wokusebenzisa ama-algorithms ukudala ukuhlaziya kusuka kudatha engahlelekile. Ihlanganisa izindlela zokubonisa le datha nenqubo yokuthola nokuqoqa idatha efanelekile.
Lokhu kungase kuthathe uhlobo lwencazelo yezibalo eqondile yemiphumela noma ukumelwa okubonakalayo kwedatha. Ukuhlaziywa kokuthakaselayo kanye nokubandakanyeka esihlokweni esithile kungenziwa kusetshenziswa ulwazi oluvela ku-YouTube, kuhlanganisa nalapho ababukeli bechofoza ividiyo ethile.
Izilinganiso zenkanyezi yomkhiqizo zingahlaziywa kusetshenziswa i-data scraping kusuka kumasayithi we-e-commerce ukuze kuhlinzekwe ngesikolo sonke sokuthi umkhiqizo wenza kahle kangakanani.
Ukuhlaziywa Kwengqondo
Ngokuqhubekayo nokuhlola i-NER, ukuhlaziywa kwemizwa ingahlukanisa phakathi kokubuyekezwa okuhle nokubi ngisho nalapho lungekho ulwazi oluvela ezilinganisweni zenkanyezi.
Kuyaqapheleka ukuthi amagama anjengokuthi “ukweqile,” “ukufiphaza,” nelithi “siphukuphuku” anezincazelo ezingezinhle, kuyilapho amagama anjengokuthi “okuwusizo,” “okusheshayo,” nelithi “lula” ayasebenza. Igama elithi “lula” lingahunyushwa kabi emdlalweni wekhompyutha.
Ama-algorithms ayinkimbinkimbi angakwazi futhi ukubona ubudlelwano phakathi kwezinto.
Izibalo zombhalo
Ngokufanayo nokuhlaziywa kwedatha, ukuhlaziywa kombhalo kukhipha ulwazi kuyunithi yezinhlamvu zombhalo ezingahlelekile futhi kusebenzisa i-NER kuye kuqanda kudatha ebalulekile.
Ingasetshenziselwa ukuhlanganisa idatha ekukhulunyweni komkhiqizo, intengo emaphakathi, noma imigomo amakhasimende ayisebenzisa kakhulu ukuchaza uhlobo oluthile.
Ukuhlaziywa kokuqukethwe kwevidiyo
Amasistimu ayinkimbinkimbi kakhulu yilawo akhipha idatha kulwazi lwevidiyo kusetshenziswa ukubona ubuso, ukuhlaziya umsindo, nokubonwa kwesithombe.
Ngokusebenzisa ukuhlaziya okuqukethwe kwevidiyo, ungathola amavidiyo e-YouTube "unboxing", ukuboniswa kwegeyimu ye-Twitch, ukuvumelanisa izindebe zomsindo wakho ku-Reels, nokuningi.
Ukuze ugweme ukulahlekelwa ulwazi olubalulekile mayelana nokuthi abantu baxhuma kanjani kumkhiqizo noma isevisi yakho njengoba umthamo wevidiyo eku-inthanethi ukhula, izindlela ezisheshayo neziqanjiwe zokuhlaziya okuqukethwe kwevidiyo okusekelwe ku-NER zibalulekile.
Uhlelo lokusebenza lomhlaba wangempela lwe-NER
Ukuqashelwa kwebhizinisi okuqanjwe igama (NER) kukhomba izici ezibalulekile embhalweni njengamagama abantu, izindawo, izinhlobo, amanani emali, nokuningi.
Ukukhipha amabhizinisi amakhulu embhalweni kusiza ekuhleleni idatha engahlelekile futhi kutholwe ulwazi olubalulekile, olubalulekile lapho usebenza namasethi amakhulu edatha.
Nazi ezinye izibonelo zomhlaba wangempela ezihehayo zokuqashelwa kwebhizinisi:
Ihlaziya Impendulo Yekhasimende
Izibuyekezo eziku-inthanethi ziwumthombo omuhle kakhulu wempendulo yabathengi njengoba zingakunikeza ulwazi oluningiliziwe mayelana nokuthi amakhasimende athandani nazizondayo mayelana nezimpahla zakho kanye nokuthi yiziphi izindawo zenkampani yakho ezidinga ukuthuthukiswa.
Konke lokhu okufakwayo kweklayenti kungahlelwa kusetshenziswa amasistimu e-NER, angaphinda ahlonze izinkinga eziphinda zivele.
Isibonelo, ngokusebenzisa i-NER ukuhlonza izindawo ezivame ukucashunwa ekubuyekezweni kwamakhasimende okungekuhle, unganquma ukugxila egatsheni elithile lehhovisi.
Isincomo sokuqukethwe
Uhlu lwama-athikili axhunywe kuleyo oyifundayo lungatholakala kumawebhusayithi afana ne-BBC ne-CNN uma ufunda into lapho.
Lawa mawebhusayithi enza izincomo zamawebhusayithi engeziwe anikeza ulwazi mayelana namabhizinisi alukhiphe kokuqukethwe okufundayo usebenzisa i-NER.
Hlela Amathikithi Ekusekelweni Kwamakhasimende
Ungasebenzisa ama-algorithms okuqaphela ibhizinisi ukuze uphendule izicelo zeklayenti ngokushesha okukhulu uma uphethe ukukhuphuka kwenani lamathikithi osekelo avela kumakhasimende.
Yenza ngokuzenzakalelayo imisebenzi yokunakekela amakhasimende edla isikhathi, njengokuhlukanisa izikhalazo zamakhasimende kanye nemibuzo, ukuze uzigcinele imali, ukhuphule injabulo yekhasimende, futhi ukhuphule amanani okuxazulula.
Ukukhishwa kwebhizinisi kungasetshenziswa futhi ukuze kukhishwe idatha efanelekile, njengamagama emikhiqizo noma izinombolo zomkhiqizo, ukuze kwenziwe kube lula ukuhambisa amathikithi kumenzeli olungile noma ithimba lokuxazulula leyo nkinga.
I-algorithm yosesho
Uke wazibuza ukuthi amawebhusayithi anezigidi zolwazi angakhiqiza kanjani imiphumela ehambisana nosesho lwakho? Cabangela iwebhusayithi i-Wikipedia.
I-Wikipedia ibonisa ikhasi eliqukethe izinto ezichazwe ngaphambilini igama lokusesha elingahlobana nazo uma usesha “imisebenzi,” esikhundleni sokubuyisela zonke iziqephu zendatshana ezinegama elithi “imisebenzi” kuzo.
Ngakho-ke, i-Wikipedia inikeza isixhumanisi esihlokweni esichaza "umsebenzi," isigaba sabantu ababizwa ngokuthi Imisebenzi, kanye nenye indawo yemidiya enjengamamuvi, amageyimu evidiyo, nezinye izinhlobo zokuzijabulisa lapho igama elithi “imisebenzi” livela khona.
Uzobona enye ingxenye yezindawo eziqukethe igama lokusesha.
Ukunakekela ama-CV
Ekufuneni umfakisicelo ofanelekile, abaqashi bachitha ingxenye enkulu yosuku lwabo bebuyekeza kabusha. Yonke i-CV inolwazi olufanayo, kodwa zonke zethulwa futhi zihlelwa ngendlela ehlukile, okuyisibonelo esijwayelekile sedatha engahlelekile.
Ulwazi olubaluleke kakhulu mayelana namakhandidethi lungakhishwa ngokushesha ngamathimba aqashayo asebenzisa izikhiqizi zebhizinisi, okuhlanganisa idatha yomuntu siqu (njengegama, ikheli, inombolo yocingo, usuku lokuzalwa, ne-imeyili) kanye nolwazi mayelana nemfundo nolwazi lwabo (njengezitifiketi, iziqu. , amagama ezinkampani, amakhono, njll).
E-commerce
Mayelana ne-algorithm yokusesha umkhiqizo, abathengisi abaku-inthanethi abanamakhulu noma izinkulungwane zezimpahla bazozuza ku-NER.
Ngaphandle kwe-NER, ukusesha “kwamabhuzu esikhumba amnyama” kuzobuyisela imiphumela ehlanganisa kokubili isikhumba nezicathulo ebezingemnyama. Uma kunjalo, amawebhusayithi e-e-commerce asengozini yokulahlekelwa amakhasimende.
IEsimeni sethu, i-NER izohlukanisa igama lokusesha njengohlobo lomkhiqizo wamabhuzu esikhumba kanye nomnyama njengombala.
I-Best Entity Extraction APIs
I-Google Cloud NLP
Ngamathuluzi aseqeqeshiwe, i-Google Cloud NLP inikeza i-Natural Language API. Noma, i-AutoML Natural Language API ivumelana nezimo ezinhlotsheni eziningi zokukhishwa kombhalo nokuhlaziywa uma ufuna ukufundisa amathuluzi akho ngamagama omkhakha wakho.
Ama-API asebenzisana kalula ne-Gmail, i-Google AmaSpredishithi, nezinye izinhlelo zokusebenza ze-Google, kodwa ukuwasebenzisa nezinhlelo zezinkampani zangaphandle kungadinga ikhodi eyinkimbinkimbi.
Inketho yebhizinisi ekahle ukuxhuma izinhlelo zokusebenza ze-Google kanye Nesitoreji Samafu njengamasevisi aphethwe nama-API.
IBM Watson
I-IBM Watson iyinkundla enamafu amaningi esebenza ngokushesha ngendlela emangalisayo futhi ehlinzeka ngamakhono akhelwe ngaphambili, njengenkulumo-nombhalo, okuyisofthiwe emangalisayo ekwazi ukuhlaziya ngokuzenzakalelayo izingcingo ezilalelwayo nezingcingo.
Ngokusetshenziswa kwedatha ye-CSV, i-AI yokufunda ejulile ye-Watson Natural Language Understanding ingakha amamodeli okukhipha ukuze kukhishwe amabhizinisi noma amagama angukhiye.
Futhi ngokuzijwayeza, ungakha amamodeli ayinkimbinkimbi kakhulu. Yonke imisebenzi yayo ifinyeleleka ngama-API, nakuba ulwazi olubanzi lokubhala amakhodi luyadingeka.
Isebenza kahle emabhizinisini amakhulu adinga ukuhlola amadathasethi amakhulu futhi abe nezinsiza zangaphakathi zobuchwepheshe.
I-Cortical.io
Ukusebenzisa i-Semantic Folding, umbono ovela ku-neurology, i-Cortical.io inikeza ukukhishwa kombhalo kanye nezixazululo ze-NLU.
Lokhu kwenzelwa ukukhiqiza “izigxivizo zeminwe ze-semantic,” ezibonisa kokubili incazelo yombhalo kuwo wonke amagama awo kanye naqondile. Ukuze ubonise ubudlelwano phakathi kwamaqoqo amagama, izigxivizo zeminwe ze-semantic zibonisa idatha yombhalo.
Amadokhumenti e-Cortical.io's interactive API ahlanganisa ukusebenza kwesixazululo ngasinye sokuhlaziywa kombhalo, futhi kulula ukuyithola usebenzisa i-Java, Python, ne-Javascript API.
Ithuluzi le-Contract Intelligence elivela ku-Cortical.io ladalelwa ukuhlaziya okusemthethweni ukuze kwenziwe ukusesha kwe-semantic, ukuguqula amadokhumenti askeniwe, nokusiza futhi kuthuthukiswe ngesichasiselo.
Ilungele amabhizinisi afuna ama-API asebenziseka kalula angadingi ulwazi lwe-AI, ikakhulukazi emkhakheni wezomthetho.
Imfene Funda
Zonke izilimi eziyinhloko zekhompyutha zisekelwa i-MonkeyLearn's APIs futhi zisetha nje imigqa embalwa yekhodi ukuze kukhiqizwe ifayela le-JSON eliqukethe amabhizinisi akho akhishiwe. Kuma-extractors nabahlaziyi bombhalo abanokuqeqeshwa kwangaphambilini, isixhumi esibonakalayo sisebenziseka kalula.
Noma, ngezinyathelo ezimbalwa nje ezilula, ungakha i-extractor eyingqayizivele. Ukuze unciphise isikhathi futhi uthuthukise ukunemba, ukucubungula kolimi lwemvelo okuthuthukisiwe (NLP) ngokujulile ukufunda imishini ikuvumela ukuthi uhlole umbhalo ngendlela umuntu ebengenza ngayo.
Ukwengeza, ama-SaaS API aqinisekisa ukuthi ukusetha ukuxhumana ngamathuluzi afana ne-Google AmaSpredishithi, i-Excel, i-Zapier, i-Zendesk, namanye akudingi iminyaka yolwazi lwesayensi yekhompyutha.
Okutholakalayo esipheqululini sakho yigama elikhipha igama, isikhi senkampani, nesikhi sendawo. Ukuze uthole ulwazi lokuthi ungazakha kanjani eyakho, bheka isihloko sebhulogi esiqanjwe igama sebhizinisi.
Ilungele amabhizinisi abo bonke osayizi abathintekayo kwezobuchwepheshe, ukudayisa, kanye ne-e-commerce adinga ama-API alula ukuwasebenzisa ezinhlotsheni ezihlukahlukene zokukhipha umbhalo nokuhlaziywa kombhalo.
Ukuqonda kwe-Amazon
Ukuze wenze kube lula ukuxhuma nokusebenzisa amathuluzi e-Amazon Comprehend akhelwe ngaphambilini ngokushesha, aqeqeshwa emakhulwini ezinkambu ezahlukene.
Awekho amaseva angaphakathi adingekayo ngoba le yisevisi egadiwe. Ikakhulukazi uma njengamanje usebenzisa ifu le-Amazon ukuya ezingeni elithile, ama-API abo ahlangana kalula nezinhlelo zokusebenza ezikhona ngaphambilini. Futhi ngokuqeqeshwa okwengeziwe kancane, ukunemba kokukhipha kungaphakanyiswa.
Enye yezindlela ezithembeke kakhulu zokuhlaziya umbhalo zokuthola idatha kumarekhodi ezokwelapha nezivivinyo zomtholampilo i-Comprehend's Medical Named Entity and Relationship Extraction (NERe), engakhipha imininingwane yemithi, izimo, imiphumela yokuhlolwa, nezinqubo.
Uma uqhathanisa idatha yesiguli ukuze ihlolwe futhi icushwe kahle ukuxilongwa, kungaba yinzuzo kakhulu. Inketho engcono kakhulu yamabhizinisi afuna isevisi ephethwe ngamathuluzi aqeqeshwe ngaphambilini.
Aylien
Ukuze unikeze ukufinyelela okulula ekuhlaziyeni okuqinile kombhalo wokufunda komshini, i-AYLIEN inikeza ama-plug-in amathathu e-API ngezilimi eziyisikhombisa ezidumile zokuhlela.
I-API Yabo Yezindaba inikeza ukusesha kwesikhathi sangempela kanye nokuthathwa kwebhizinisi emashumini ezinkulungwane zemithombo yezindaba evela emhlabeni wonke.
Ukukhishwa kwebhizinisi neminye imisebenzi eminingana yokuhlaziya umbhalo kungenziwa kusetshenziswa i-Text Analysis API kumadokhumenti, social media izinkundla, izinhlolovo zabathengi, nokuningi.
Okokugcina, usebenzisa i-Text Analysis Platform, ungazakhela eyakho i-extractors futhi uqonde kakhudlwana esipheqululini sakho (TAP). Isebenza kahle ezinkampanini ezidinga ukuhlanganisa ngokuyinhloko ama-API angaguquki ngokushesha.
I-SpaCy
I-SpaCy iyiphakheji ye-Python Natural Language Processing (NLP) engumthombo ovulekile, mahhala, futhi enezici eziningi ezakhelwe ngaphakathi.
Iya ngokuya ijwayeleka Idatha ye-NLP ukucubungula nokuhlaziya. Idatha yombhalo engahlelekile idalwe ngezinga elikhulu, ngakho-ke kubalulekile ukuyihlaziya nokukhipha imininingwane kuyo.
Ukuze ufeze lokho, kufanele uveze amaqiniso ngendlela amakhompyutha angawaqonda. Ungakwenza nge-NLP. Iyashesha kakhulu, inesikhathi se-lag esingu-30ms kuphela, kodwa ngokubucayi, ayihloselwe ukusetshenziswa namakhasi e-HTTPS.
Lena inketho enhle yokuskena amaseva akho noma i-intranethi ngoba isebenza endaweni, kodwa akulona ithuluzi lokufunda yonke i-inthanethi.
Isiphetho
Ukuqashelwa kwebhizinisi okuqanjwe igama (NER) kuwuhlelo amabhizinisi angalusebenzisa ukuze alebule ulwazi olubalulekile ezicelweni zokusekelwa kwamakhasimende, athole amabhizinisi okubhekiselwa kuwo kumpendulo yekhasimende, futhi akhiphe ngokushesha idatha ebalulekile njengemininingwane yokuxhumana, izindawo, nezinsuku, phakathi kwezinye izinto.
Indlela ejwayeleke kakhulu yokuqanjwa ukuqashelwa kwebhizinisi iwukusebenzisa ama-API okukhishwa kwebhizinisi (noma ngabe ahlinzekwa ngamalabhulali omthombo ovulekile noma imikhiqizo ye-SaaS).
Kodwa-ke, ukukhetha enye indlela engcono kakhulu kuzoncika esikhathini sakho, izimali, kanye nesethi yamakhono. Kunoma yiluphi uhlobo lwebhizinisi, ukukhishwa kwebhizinisi kanye nobuchwepheshe bokuhlaziya umbhalo obuyinkimbinkimbi kakhulu kungaba nenzuzo ngokusobala.
Uma amathuluzi okufunda omshini efundiswa kahle, anembile futhi awashayi indiva idatha, akulondolozela isikhathi nemali. Ungalungiselela lezi zixazululo ukuthi zisebenze ngokuqhubekayo nangokuzenzakalelayo ngokuhlanganisa ama-API.
Vele ukhethe inkambo yesenzo elungele inkampani yakho.
shiya impendulo