ELMo Contextual Embeddings
ELMo (Ku-xidhnaanta Qaababka Luuqadda) waxay ahayd horumar 2018 ah oo erey kasta siisay matalaad u qaabaysan jumladdeeda, markaa 'bangi' ee 'bangiga webiga' wuu ka duwan yahay 'bangi' 'bangiga kaydka'. Waxay calaamadisay beddelka ka beddelka ereyada taagan oo loo beddelay NLP macnaha guud.
quusid qoto dheer
ELMo, oo ay soo bandhigtay Machadka Allen ee cilmi-baarayaasha AI (Peters et al., 2018), waxay soo saartaa ereyada ereyada iyadoo la adeegsanayo jumlad iyada oo loo marayo qaabka qoto dheer ee luqadda LSTM ee laba jiho oo lagu tababaray bilyan-ereyga corpus. Si ka duwan Word2Vec ama GloVe, kaas oo ku meeleeya hal halbeeg oo go'an eray kasta, ELMo waxay xisaabisaa vector cusub dhacdo kasta oo ku salaysan macnaha ku hareeraysan. Muhiimad ahaan, ELMo waxay isku daraysaa dhammaan lakabyada LSTM gudaha iyada oo loo marayo la bartay, miisaan hawleed gaar ah halkii la isticmaali lahaa oo keliya lakabka sare. Lakabyada hoose waxay u muuqdaan inay qabtaan syntax (qayb-hadal, qaab-dhismeedka) halka lakabyada sare ay qabtaan semantics iyo macnaha erayga. Ku darista ELMo moodooyinka jira waxay soo saartay guulo waaweyn oo dhan lix hawlood oo bartilmaameed ah, oo ay ku jiraan ka jawaabista su'aalaha, falanqaynta dareenka, iyo aqoonsiga cidda la magacaabay.
Aragtida Farsamada
ELMo waxay soo ururisay laba LSTMs: qaabka luuqada hore ee saadaaliya kalmada soo socota iyo mid gadaal u saadaalinaya kelmadii hore, mid walbana waxa uu ka sarreeyaa heer-beeleedka CNN (sidaas darteed waxa ay gacanta ku haysaa ereyada aan la arki karin). Hawsha hoose ee socota, ELMo waxay duminaysaa matalaada lakabka iyadoo la isticmaalayo softmax-caadi miisaan iyo miisaan, dhammaan la bartay intii lagu jiray hagaajinta. Tani waxay la macno tahay hawl kastaa waxay go'aamin kartaa inta syntactic iyo semantic signal ay ka rabto biLM hore loo tababaray ee la qaboojiyey.
Saamaynta Istiraatijiyadeed
Xawaaraha iyo miisaanka
Socodka shaqada luqaddu si dhakhso leh ayay u socon kartaa iyada oo aan la hurayn joogteynta.
Helitaanka iyo gaarsiinta
Waxay balaadhisaa gelitaanka luqadaha iyo qaababka isgaarsiinta.
Go'aamo cad
Kooxuhu waxay waqti badan ku qaadan karaan xukunka halka otomaatiggu uu qabanayo ku celcelinta.
Mustaqbalka Isku-xidhka Macnaha guud ee ELMo
Fikradda udub-dhexaadka ah ee ELMo, matalaadda macnaha guud ee ka-hortagga-qaabaynta luqadda, ayaa noqday aasaas, laakiin qaab-dhismeedkeeda LSTM ee soo noqnoqda ayaa si dhakhso ah u daboolay moodooyinka ku-salaysan Transformer sida BERT dabayaaqadii 2018, kuwaas oo akhriyay jumlado dhan isbar-bar iyo miisaan aad u wanaagsan. Maanta ELMo inta badan waa muhiimada taariikhiga ah iyo waxbarashada, in kasta oo la tacaalida jilaha-CNN iyo fikradaha lakabka miisaanku ay wali saamaynayaan ku-xidhka shaqada khaaska ah ee agabka hoose iyo luqadaha hodanka ah.
Dhaqangelinta Adduunka-dhabta ah
Hagaajinta nidaamyada aqoonsiga magac ee ay tahay inay sheegaan in 'Washington' loola jeedo qof, gobol, ama magaalo iyadoo lagu saleynayo erayada ku xeeran
Kobcinta falanqaynta dareenka iyada oo la qabanayo in 'jiro' ay ka dhigan tahay diidmo 'waxaan dareemayaa jirran' laakiin si togan u leh 'taasi waa jiran'
Kor u qaadida nidaamyada su'aalaha-jawaabista ee bartilmaameedka SQuAD iyadoo la quudinayo calaamooyinka xasaasiga ah ee akhristaha
Dareenka ereyada kala-duwan ee tarjumaadda mishiinka si erayada badan sida 'dhirta' ay u turjumaan si sax ah macnaha guud
Khatarta & Dariiqyada Ilaalada
Xaqiiqooyinka dhalanteed waxay si deggan u geli karaan warbixinnada, taageerada socodka, ama natiijooyinka cilmi-baarista.
Dareenka degdega ahi wuxuu abuuri karaa natiijooyin aan iswaafaqayn codsiyada la midka ah.
Xogta qoraalka xasaasiga ah ayaa laga yaabaa in la kashifo haddii kontaroolada gelitaanka ay daciif yihiin.
Qorshe Hawleedka Dhaqangelinta
Qeex qaabka wax soo saarka, codka, iyo heerarka tayada ka hor inta aan la baahin.
Jawaabaha salka ku haya ilo lagu kalsoon yahay mar kasta oo saxnidu ay muhiim tahay.
Hayso isbaarada dib u eegista bini aadamka ee wax soo saarka sare.
Lasoco qaababka guuldarada oo dib u leyli dardargelinta ama socodka shaqada si joogto ah.
Sii wad Sahaminta
Free newsletter
Get the daily AI briefing
Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.
One email each weekday. Unsubscribe in one click. We never sell or share your address.
Test yourself
Take the ELMo Contextual Embeddings quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
Hagaha xiga
Erayada Embending
Su'aalaha soo noqnoqda
What is ELMo Contextual Embeddings?
ELMo (Ku-xidhnaanta Qaababka Luuqadda) waxay ahayd horumar 2018 ah oo erey kasta siisay matalaad u qaabaysan jumladdeeda, markaa 'bangi' ee 'bangiga webiga' wuu ka duwan yahay 'bangi' 'bangiga kaydka'. Waxay calaamadisay beddelka ka beddelka ereyada taagan oo loo beddelay NLP macnaha guud.
Waa maxay farqiga ugu muhiimsan ee u dhexeeya ELMo iyo ku-xidhitaanadii hore sida Word2Vec?
ELMo waxay soo saartaa isku-xidhka macnaha guud: vector-ka kelmadu wuu beddelaa iyadoo lagu salaynayo jumlada ku xeeran, si ka duwan halbeegga go'an ee Word2Vec eray kasta.
Waa maxay qaab-dhismeedka neerfaha ee ay ELMO u isticmaasho akhrinta qoraalka?
ELMo waxay ku dhisantahay LSTM laba jiho ah oo qoto dheer oo loo tababaray qaab luuqadeed, habaynta isku xigxiga ee soo noqnoqda.
Sidee ELMo isugu geysaa lakabyada gudaha hawsha hoose?
ELMo waxay barataa hawl gaar ah oo hawl gaar ah oo softmax-caadi ah si ay isugu keento dhammaan lakabyada biLM, iyada oo u oggolaanaysa hawl kasta in ay xoogga saarto syntax ama semantics haddii loo baahdo.
Maxay ELMo u isticmaashaa sidii unugyadeeda wax gelinta si ay u qabato erayada aan la arkin?
ELMo waxay ka dhistaa ereyada CNN ee heerka dabeecadda ah, si ay u matasho ereyada aysan waligood arag inta lagu jiro tababarka.
Qiyaastii goorma ayaa ELMo la soo bandhigay yaase keenay?
ELMo waxaa lagu daabacay 2018 Peters et al. Machadka Allen ee AI (AI2).