Isiqulatho[Fihla][Bonisa]
Sinawo amandla okuqonda nokuhlela amagama ngokwabantu ngabanye, iindawo, iindawo, amaxabiso, nangakumbi xa siwava okanye siwafunda. Abantu bayakwazi ukwahlula amagama ngokweendidi, ukuchonga nokuqonda amagama ngokukhawuleza.
Ngokomzekelo, unokuhlela into kwaye ngokukhawuleza uze neempawu ezintathu ukuya kwezine xa usiva igama elithi "Steve Jobs,"
- Umntu: "Steve Jobs"
- Umbutho: "Apple"
- Indawo: "California"
Ekubeni iikhomputha zingenabo obu buchule bendalo, kufuneka sibancede ekuqondeni amagama okanye isicatshulwa kunye nokuhlela. UQwalaselo lweZiko eliGayweyo (NER) lisetyenziswa kule meko.
Kweli nqaku, siza kuhlola i-NER (ebizwa ngokuba yiNgcaciso yeNkampani) ngokubanzi, kubandakanywa nokubaluleka kwayo, izibonelelo, ii-API eziphezulu ze-NER, kunye nokunye okuninzi.
Yintoni i-NER (iNgqwalasela yeQumrhu)?
Inkqubo yolwimi lwendalo (NLP) indlela eyaziwa ngokuba yi-entity recognition (NER), maxa wambi eyaziwa njengokuchongwa kwequmrhu okanye utsalo lwequmrhu, ibona ngokuzenzekelayo amaqumrhu amagama kwisicatshulwa ize iwadibanise ngokweendidi ezimiselwe kwangaphambili.
Amaqumrhu aquka amagama abantu, amaqela, iindawo, imihla, iimali, iimali zeedola, iipesenti, nokunye. Ngokuqatshelwa kwequmrhu, unokuyisebenzisa ukuqokelela idatha ebalulekileyo yesiseko sedatha okanye ukhuphe ulwazi olubalulekileyo ukuqonda ukuba uxwebhu lungantoni.
I-NER ilitye lembombo apho inkqubo ye-AI ixhomekeke khona ukuze ihlalutye isicatshulwa kwi-semantics ehambelanayo kunye neemvakalelo, nokuba i-NLP ibonisa ukuqhubela phambili okubalulekileyo kwinkqubo yohlalutyo lombhalo.
Yintoni intsingiselo ye-NER?
Isiseko sendlela yokuhlalutya isicatshulwa yiNER. Imodeli ye-ML kufuneka iqale inikwe izigidi zeesampulu ezineendidi ezichazwe kwangaphambili phambi kokuba iqonde isiNgesi.
I-API iphucula ngexesha lokuqaphela la macandelo kwiitekisi ezifunda okokuqala. Amandla enjini yohlalutyo lokubhaliweyo ayanda ngesakhono se-NER kunye namandla.
Njengoko kubonwa apha, imisebenzi emininzi yeML iqalwa yiNER.
Uphendlo lweSemantic
Uphendlo lweSemantic ngoku luyafumaneka kuGoogle. Ungangenisa umbuzo, kwaye iya kuzama kangangoko ukuphendula ngempendulo. Ukuze ufumane ulwazi, umsebenzisi ukhangela, abancedisi bedijithali abafana ne-Alexa, i-Siri, i-chatbots, kunye nabanye basebenzisa uhlobo lokukhangela kwe-semantic.
Lo msebenzi unokubethelwa okanye uphoswe, kodwa kukho inani elikhulayo lokusetyenziswa kuwo, kwaye ukusebenza kwawo kunyuka ngokukhawuleza.
Uhlalutyo lweedatha
Eli libinzana eliqhelekileyo lokusebenzisa i-algorithms ukwenza uhlalutyo olusuka kwidatha engacwangciswanga. Idibanisa iindlela zokubonisa le datha kunye nenkqubo yokufumana kunye nokuqokelela idatha efanelekileyo.
Oku kunokuthatha uhlobo lwengcaciso yeenkcukacha-manani ethe ngqo yeziphumo okanye umboniso obonwayo wedatha. Uhlalutyo lomdla kunye nokuzibandakanya ngesihloko esithile lunokwenziwa kusetyenziswa ulwazi oluvela kwimibono yeYouTube, kubandakanywa naxa ababukeli becofa ividiyo ethile.
Imilinganiselo yeenkwenkwezi yemveliso inokuhlalutywa ngokusebenzisa i-data scraping kwiindawo ze-e-commerce ukubonelela ngenqaku elipheleleyo lokuba imveliso iqhuba kakuhle kangakanani.
Uhlalutyo lwesivakalisi
Ukuphonononga ngakumbi i-NER, Uhlalutyo lweemvakalelo inokwahlula phakathi kokuphononongwa okuhle nokubi nangona kungekho lwazi lusuka kumanqaku eenkwenkwezi.
Iyaqonda ukuba amabinzana anjengelithi “ubaxiweyo,” “ububhanxa,” nelithi “isidenge” aneentsingiselo ezingakhiyo, ngoxa amabinzana afana nelithi “luncedo,” “ukukhawuleza,” nelithi “lula” esebenza. Igama elithi “lula” linokutolikwa kakubi kumdlalo wekhompyutha.
Ii-algorithms ezintsonkothileyo zinokuqaphela ubudlelwane phakathi kwezinto.
Uhlalutyo lombhalo
Ngokufana nohlalutyo lwedatha, uhlalutyo lombhalo lukhupha ulwazi oluvela kwiintambo zetekisi ezingaqingqwanga kwaye isebenzisa i-NER ukuya kwi-zero kwidatha ebalulekileyo.
Ingasetyenziselwa ukuqokelela idatha ekukhankanyiweyo kwemveliso, ixabiso eliphakathi, okanye imigaqo esetyenziswa rhoqo ngabathengi ukuchaza uphawu oluthile.
Uhlalutyo lwesiqulatho seVidiyo
Iinkqubo ezinzima kakhulu zezo zikhupha idatha kwiinkcukacha zevidiyo usebenzisa ukubonwa kobuso, uhlalutyo lomsindo, kunye nokubonwa kwemifanekiso.
Usebenzisa uhlalutyo lomxholo wevidiyo, unokufumana iividiyo zeYouTube "ezingenabhokisi", imiboniso yomdlalo weTwitch, ulungelelwaniso lomlomo wemathiriyeli yakho yeaudio kwiReels, kunye nokunye.
Ukuze ugweme ukuphulukana nolwazi olubalulekileyo malunga nendlela abantu abanxibelelana ngayo nemveliso okanye inkonzo yakho njengoko umthamo wezinto zevidiyo ezikwi-intanethi ukhula, iindlela ezikhawulezayo nezithe chatha zohlalutyo lomxholo wevidiyo esekwe kwi-NER zibalulekile.
Usetyenziso lwehlabathi lokwenyani lwe-NER
Ukuqaphela iqumrhu elinikwe igama (NER) ichonga imiba ebalulekileyo kumbhalo onje ngamagama abantu, iindawo, iibrendi, amaxabiso emali, nokunye.
Ukukhupha amaziko amakhulu kwisicatshulwa kunceda ekuhleleni idatha engacwangciswanga kunye nokufumanisa ulwazi olubalulekileyo, olubalulekileyo xa ujongene nedatha enkulu.
Nantsi eminye imizekelo enomdla yehlabathi yokwenyani yokuqondwa kwequmrhu:
Ukuhlalutya iNgxelo yoMthengi
Uphononongo lwe-Intanethi ngumthombo omangalisayo wengxelo yabathengi kuba banokukunika ulwazi oluneenkcukacha malunga nokuba abathengi bathanda ntoni kwaye bayithiye ngantoni impahla yakho kunye nokuba yeyiphi imimandla yenkampani yakho ekufuneka iphuculwe.
Lonke olu galelo lomxhasi lunokucwangciswa kusetyenziswa iinkqubo ze-NER, ezinokuphinda zichonge imiba eqhubekayo.
Umzekelo, ngokusebenzisa i-NER ukuchonga iindawo ezihlala zikhankanywa kuphononongo olungathandekiyo lwabathengi, unokwenza isigqibo sokugxila kwiofisi yesebe ethile.
Isindululo somxholo
Uluhlu lwamanqaku aqhagamshelwe kulowo ufundayo lunokufumaneka kwiiwebhusayithi ezinjenge-BBC kunye ne-CNN xa ufunda into apho.
Ezi webhusayithi zenza izindululo kwiiwebhusayithi ezongezelelweyo ezinikezela ngolwazi malunga nezinto ezizikhuphileyo kumxholo owufundayo usebenzisa i-NER.
Lungiselela amatikiti kwiNkxaso yoMthengi
Ungasebenzisa i-algorithms yokuqondwa kwequmrhu ukuphendula kwizicelo zabathengi ngokukhawuleza ukuba ulawula ukwanda kwenani lamatikiti enkxaso kubathengi.
Yenza imisebenzi yokhathalelo lwabathengi ethatha ixesha, njengokuhlela izikhalazo zabathengi kunye nemibuzo, ukuze uzigcinele imali, ukonyusa ulonwabo lwabathengi, kunye nokunyusa amanani esisombululo.
Utsalo lwequmrhu lusenokusetyenziswa ukukhupha idatha efanelekileyo, efana namagama emveliso okanye iinombolo zesiriyali, ukwenza kube lula ukuhambisa amatikiti ukuya kwiarhente elungileyo okanye iqela lokusombulula loo mba.
I-algorithm yokukhangela
Ngaba ukhe wazibuza ukuba iiwebhusayithi ezinezigidi zolwazi zinokuvelisa njani iziphumo ezihambelana nokukhangela kwakho? Cinga ngewebhusayithi iWikipedia.
IWikipedia ibonisa iphepha eliqulethe izinto ezichazwe kwangaphambili elinokuthi igama lokukhangela linxulumane xa ukhangela “imisebenzi,” endaweni yokubuyisela onke amanqaku anegama elithi “imisebenzi” kuwo.
Ke, iWikipedia ibonelela ngekhonkco kwinqaku elichaza "umsebenzi," icandelo labantu ababizwa ngokuba yiMisebenzi, kunye nenye indawo yeendaba ezinje ngeemuvi, Iimidlalo yevidiyo, kunye nezinye iindlela zokuzonwabisa apho igama elithi “imisebenzi” livela khona.
Uya kubona kwakhona elinye icandelo leendawo eziqulathe igama lokukhangela.
Ukunyamekela ukuqalisa kwakhona
Ukukhangela umfaki-sicelo ofanelekileyo, abaqeshwa bachitha inxalenye enkulu yosuku lwabo lokuphonononga kwakhona. Yonke i-CV inolwazi olufanayo, kodwa zonke zinikezelwe kwaye zicwangciswe ngokwahlukileyo, ngumzekelo oqhelekileyo wedatha engacwangciswanga.
Olona lwazi lufanelekileyo malunga nabaviwa lunokutsalwa ngokukhawuleza ngamaqela aqeshayo asebenzisa ii-extractors zequmrhu, kuquka nedatha yomntu (efana negama, idilesi, inombolo yomnxeba, umhla wokuzalwa, kunye ne-imeyile) kunye nolwazi malunga nemfundo kunye namava abo (ezifana neziqinisekiso, isidanga. , amagama enkampani, izakhono, njl njl).
E-yorhwebo
Ngokumalunga ne-algorithm yokukhangela imveliso, abathengisi be-intanethi abanamakhulu okanye amawaka eempahla baya kuxhamla kwi-NER.
Ngaphandle kwe-NER, ukhangelo “lweebhutsi zesikhumba ezimnyama” beluya kubuyisela iziphumo eziquka zombini ufele nezihlangu ebezingekho mnyama. Ukuba kunjalo, iiwebhusayithi ze-e-commerce zisengozini yokuphulukana nabathengi.
IKwimeko yethu, i-NER ingahlela igama lokukhangela njengodidi lwemveliso yeebhutsi zesikhumba kunye nomnyama njengombala.
Eyona Entity Extraction APIs
Google Cloud NLP
Kwizixhobo esele ziqeqeshiwe, i-Google Cloud NLP ibonelela nge-API yoLwimi lweNdalo. Okanye, i-AutoML yoLwimi lweNdalo ye-API ilungelelaniswa kwiindidi ezininzi zokutsalwa kombhalo kunye nohlalutyo ukuba ufuna ukufundisa izixhobo zakho kwisigama soshishino lwakho.
I-APIs isebenzisana ngokulula ne-Gmail, i-Google Sheets, kunye nezinye ii-apps zikaGoogle, kodwa ukuzisebenzisa ngeenkqubo zomntu wesithathu kunokufuna ikhowudi enzima ngakumbi.
Inketho yeshishini elifanelekileyo kukudibanisa izicelo zikaGoogle kunye noGcino lwamafu njengeenkonzo ezilawulwayo kunye neeAPI.
IBM Watson
I-IBM Watson liqonga elinamafu amaninzi elenza ngokumangalisayo ngokukhawuleza kwaye libonelela ngesakhono esakhelwe ngaphambili, njengentetho-kuya-kumbhalo, eyisoftware emangalisayo enokuhlaziya ngokuzenzekelayo ii-audio kunye neefowuni ezirekhodiweyo.
Ngokusetyenziswa kwedatha ye-CSV, i-Watson Natural Language Understanding yokufunda nzulu i-AI inokudala imizekelo yokukhupha ukukhupha amaqumrhu okanye amagama angundoqo.
Kwaye ngokuziqhelanisa, unokwenza iimodeli ezinobugocigoci ngakumbi. Yonke imisebenzi yayo ifumaneka ngee-APIs, nangona ulwazi olubanzi lwekhowudi luyafuneka.
Isebenza kakuhle kumashishini amakhulu afuna ukuphonononga iiseti zedatha enkulu kwaye abe nemithombo yobugcisa yangaphakathi.
Cortical.io
Ukusebenzisa i-Semantic Folding, ingcamango evela kwi-neurology, i-Cortical.io inikeza isicatshulwa sombhalo kunye nezisombululo ze-NLU.
Oku kwenziwa ukuvelisa "iminwe yeminwe ye-semantic," ebonisa intsingiselo yesicatshulwa ngokupheleleyo kunye namagama athile. Ukuze ubonise unxulumano phakathi kweqela lamagama, iiprinta zeminwe zesemantic zibonisa idatha yokubhaliweyo.
Amaxwebhu e-API asebenzisanayo e-Cortical.io aquka ukusebenza kwezisombululo zohlalutyo lombhalo ngamnye, kwaye kulula ukufikelela kuzo usebenzisa iJava, iPython, kunye neJavascript APIs.
Isixhobo soBukrelekrele bekhontrakthi evela kwi-Cortical.io yenzelwe ngokukodwa uhlalutyo lomthetho ukwenza uphendlo lwe-semantic, ukuguqula amaxwebhu askeniweyo, kunye noncedo kunye nokuphucula ngesichasiselo.
Ilungele amashishini akhangela ii-API ezilula ukuzisebenzisa ezingadingi lwazi lwe-AI, ngakumbi kwicandelo lezomthetho.
Inkawu Funda
Zonke iilwimi eziphambili zekhompyuter zixhaswa yiMonkeyLearn's APIs kwaye zisete nje iilayini ezimbalwa zekhowudi ukuvelisa ifayile ye-JSON equlethe izinto zakho ezikhutshiweyo. Kwi-extractors kunye nabahlalutyi besicatshulwa ngoqeqesho lwangaphambili, i-interface isebenziseka lula.
Okanye, ngamanyathelo ambalwa nje alula, unokwenza i-extractor ekhethekileyo. Ukunciphisa ixesha kunye nokuphucula ukuchaneka, ukuqhubela phambili kolwimi lwendalo (NLP) ngobunzulu yokufunda umatshini ikuvumela ukuba uvavanye okubhaliweyo njengokuba umntu ebenokwenza.
Ukongezelela, ii-API ze-SaaS ziqinisekisa ukuba ukuseta uxhulumaniso kunye nezixhobo ezifana ne-Google Sheets, i-Excel, i-Zapier, i-Zendesk, kunye nabanye ayifuni iminyaka yolwazi lwesayensi yekhompyutha.
Okwangoku ekhoyo kwibrawuza yakho igama le-extractor, i-extractor yenkampani, kunye ne-extractor yendawo. Ukufumana ulwazi malunga nendlela yokwakha eyakho, jonga inqaku lebhlog yokwamkelwa kwequmrhu.
Kufanelekile kumashishini abo bonke ubungakanani obubandakanyekayo kwithekhnoloji, ukuthengisa, kunye ne-e-commerce efuna i-APIs elula yokuphumeza kwiindidi ezahlukeneyo zokutsalwa kombhalo kunye nohlalutyo lombhalo.
Amazon Comprehend
Ukuze wenze kube lula ukuplaga kwaye usebenzise izixhobo ezakhelwe ngaphambili zeAmazon ngoko nangoko, baqeqeshwa kumakhulu amasimi ahlukeneyo.
Akukho ziseva zangaphakathi ezifunekayo kuba le yinkonzo ebekwe iliso. Ngokukodwa ukuba ngoku usebenzisa ilifu leAmazon ukuya kwinqanaba elithile, ii-APIs zabo zidibanisa ngokulula kunye nee -apps ezazikho ngaphambili. Kwaye ngoqeqesho oluncinci ngakumbi, ukuchaneka kokutsalwa kunokunyuswa.
Enye yeendlela ezithembekileyo zokuhlalutya isicatshulwa sokufumana idatha kwiirekhodi zonyango kunye nezilingo zeklinikhi yi-Comprehend's Medical Named Entity kunye ne-Relationship Extraction (NERe), enokuthi ikhuphe iinkcukacha kumayeza, iimeko, iziphumo zovavanyo kunye neenkqubo.
Xa uthelekisa idatha yesigulana ukuvavanya kunye nokuxilongwa kakuhle, kunokuba luncedo kakhulu. Olona khetho lungcono kumashishini afuna inkonzo elawulwayo ngezixhobo eziqeqeshwe kwangaphambili.
Aylien
Ukuze unikeze ukufikelela okulula kuhlalutyo lombhalo oluqinileyo lokufunda ngomatshini, i-AYLIEN inikezela ngeeplagi ze-API ezintathu kwiilwimi ezisixhenxe ezidumileyo.
I-API yeendaba zabo ibonelela ngexesha lokwenyani lokukhangela kunye nokutsalwa kwequmrhu kumashumi amawaka emithombo yeendaba evela kwihlabathi liphela.
Utsalo lwequmrhu kunye neminye imisebenzi yohlahlelo lwetekisi inokuqhutywa kusetyenziswa i-Text Analysis API kumaxwebhu, Imidiya yokuncokola amaqonga, uphando lwabathengi, kunye nokunye.
Ekugqibeleni, usebenzisa i-Text Analysis Platform, unokwenza ii-extractors zakho kwaye ngokuthe ngqo kwi-browser yakho (TAP). Isebenza kakuhle kwiinkampani ezifuna ukudibanisa ngokuyinhloko ii-API ezimiselweyo ngokukhawuleza.
I-SpaCy
I-SpaCy yiphakheji yePython Natural Language Processing (NLP) evulekile, ekhululekile, kwaye inetoni yezinto ezakhelwe ngaphakathi.
Iya ixhaphaka ngakumbi Idatha ye-NLP ukusetyenzwa nohlalutyo. Idatha yombhalo engacwangciswanga yenziwe ngokomlinganiselo omkhulu, ngoko ke kubalulekile ukuyihlalutya kwaye ukhuphe ulwazi kuyo.
Ukufezekisa oko, kufuneka ubonise iinyani ngendlela iikhompyuter ezinokuqondwa ngayo. Ungayenza nge-NLP. Ikhawuleza kakhulu, kunye nexesha le-lag le-30ms kuphela, kodwa ngokubaluleke kakhulu, ayenzelwanga ukusetyenziswa ngamaphepha e-HTTPS.
Olu lukhetho oluhle lokuskena iiseva zakho okanye i-intranet kuba isebenza ekuhlaleni, kodwa ayisosixhobo sokufunda yonke i-intanethi.
isiphelo
Ukuqatshelwa kwequmrhu elibizwa ngegama (NER) yinkqubo amashishini anokusebenzisa ukuleyibheliza ulwazi olufanelekileyo kwizicelo zenkxaso yabathengi, ukufumana amaqumrhu ekubhekiselwa kuwo kwingxelo yabathengi, kwaye ngokukhawuleza akhuphe idatha ebalulekileyo njengeenkcukacha zoqhagamshelwano, iindawo, kunye nemihla, phakathi kwezinye izinto.
Eyona ndlela ixhaphakileyo yokubizwa ngegama lokuqondwa kwequmrhu kukusebenzisa ii-APIs zokutsalwa kwequmrhu (nokuba zibonelelwa ngamathala eencwadi avulelekileyo okanye iimveliso ze-SaaS).
Nangona kunjalo, ukukhetha eyona ndlela ilungileyo kuya kuxhomekeka kwixesha lakho, imali kunye neseti yesakhono. Kulo naluphi na uhlobo lweshishini, ukutsalwa kwequmrhu kunye nobuchwephesha bokucazulula itekhnoloji engaphezulu kunokuba luncedo ngokucacileyo.
Xa izixhobo zokufunda ngoomatshini zifundiswa ngokuchanekileyo, zichanekile kwaye azihoyi nayiphi na idatha, zikongela ixesha kunye nemali. Ungaqwalasela ezi zisombululo ukuba ziqhube ngokuqhubekayo kwaye ngokuzenzekelayo ngokudibanisa ii-API.
Khetha ngokulula ikhondo lesenzo elilungele inkampani yakho.
Shiya iMpendulo