Haɗin Kalma
Ƙirƙirar kalmomi suna juya kalmomi zuwa lissafin lambobi ta yadda kalmomin da aka yi amfani da su ta hanyoyi iri ɗaya su ƙare tare a cikin sararin lissafi.
Dubawa
They are the foundation that lets a computer treat language as something it can measure and compare.
Zurfafa nutsewa
Ƙirƙirar kalma tana wakiltar kowace kalma azaman vector - jerin lambobi masu tsawo, yawanci 100 zuwa 300 don ƙirar ƙira. Ana koyon waɗannan lambobi daga babban adadin rubutu ta hanyar lura da waɗanne kalmomi suke bayyana kusa da juna. Word2vec, wanda Tomas Mikolov da abokan aikinsa suka saki a Google a cikin 2013, sun yada ra'ayin tare da dabaru na horo guda biyu: skip-gram (annabta kewaye kalmomi daga kalmar manufa) da CBOW (yi hasashen manufa daga makwabta). Stanford's GloVe ya biyo baya a cikin 2014, yana gina ƙididdiga daga ƙididdiga na haɗin gwiwar kalmomin duniya. Shahararren sakamakon shi ne cewa vector math yana ɗaukar ma'ana: sarki cire namiji da mace ya sauka kusa da sarauniya. Manyan nau'ikan harshe na yau sun ci gaba, suna koyon abubuwan haɗawa don alamun da ke canzawa tare da mahallin.
Fahimtar Fasaha
Abubuwan da aka haɗa ana koyan su, ba da hannu ba. Yayin horarwa samfurin yana daidaita yanayin kowane kalma ta yadda kalmomin da ke bayyana a cikin mahallin iri ɗaya su matso kusa da juna, a auna ta hanyar kamanni na cosine (kusurwar da ke tsakanin vectors). Classic word2vec da GloVe suna ba kowace kalma kafaffen vector guda ɗaya ba tare da la'akari da jumla ba. Samfuran tasfoma na zamani a maimakon haka suna farawa daga alamar alama sannan su sake fasalinta ta hanyar Layer, don haka kalma ɗaya kamar 'banki' ta sami nau'i daban-daban a cikin 'bankin kogi' da 'bankin ajiya' - waɗannan ana kiran su abubuwan haɗawa da mahallin.
Dabarun Tasiri
Gudu da sikelin
Gudun aikin harshe na iya tafiya da sauri ba tare da sadaukar da daidaito ba.
Shiga ku isa
Yana faɗaɗa damar shiga cikin harsuna da salon sadarwa.
Shawarwari masu haske
Ƙungiyoyi za su iya ciyar da ƙarin lokaci akan hukunci yayin da aiki da kai ke sarrafa maimaitawa.
Makomar Shigar Kalma
Matsakaicin nau'i-nau'i-nau'i-nau'i-nau'i-nau'i-nau'i-nau'i a yanzu mafi yawan ra'ayi ne na koyarwa da tushe mai sauri; Tsarin samarwa suna amfani da abubuwan da aka haɗa da mahallin mahallin daga samfuran masu canji. Iyakar da ke girma ita ce haɗe-haɗe don jimla gabaɗayan, takardu, hotuna, da sauti da aka tattara cikin sarari guda ɗaya, wanda ke ba da ikon bincike na ma'ana da haɓaka-dawowa. Yi tsammanin abubuwan haɗawa don ci gaba da samun rahusa don ƙididdigewa, yaruka da yawa ta tsohuwa, da kuma tsakiyar yadda tsarin AI ke samun bayanai masu dacewa maimakon haddace shi a cikin ma'auninsu.
Aiwatar da Gaskiyar Duniya
Injunan bincike na Semantic waɗanda ke dawo da takaddun da suka dace da ma'anar tambaya, ba kawai madaidaicin ma'anar kalmar ba.
Tsarin shawarwarin da ke ba da shawarar samfura ko labarai iri ɗaya ta hanyar kwatanta abubuwan haɗa su.
Ƙarfafa ƙarfin dawo da ƙarfi (RAG), inda chatbot ke shigar da tambayar ku don cire mafi mahimmancin rubutun rubutu daga tushen ilimi.
Tari da rarrabuwa, kamar tara tikitin tallafi na kusa iri ɗaya ko labaran labarai ta kusancin vector.
Hatsari & Tsare-tsare
Abubuwan da aka ruɗe suna iya shigar da rahotanni cikin nutsuwa, kwararar tallafi, ko abubuwan bincike.
Hankali na gaggawa na iya ƙirƙirar sakamako mara daidaituwa a cikin buƙatun iri ɗaya.
Za a iya fallasa bayanan rubutu mai ma'ana idan ikon samun dama yana da rauni.
Taswirar Hanya
Ƙayyade tsarin fitarwa, sautin, da ma'auni masu inganci kafin fitowa.
Amsa a ƙasa tare da amintattun tushe a duk lokacin da daidaito ya shafi mahimmanci.
Ajiye wurin binciken ɗan adam don abubuwan da ake samu masu girma.
Bibiyar tsarin gazawar kuma sake horar da tsokaci ko tafiyar aiki akai-akai.
Ci gaba da Bincike
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 Word 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
Jagora na gaba
Rubutun Rubutu
Tambayoyin da ake yawan yi
What is Word Embeddings?
Ƙirƙirar kalmomi suna juya kalmomi zuwa lissafin lambobi ta yadda kalmomin da aka yi amfani da su ta hanyoyi iri ɗaya su ƙare tare a cikin sararin lissafi. Su ne ginshiƙan da ke ba wa kwamfuta damar ɗaukar harshe a matsayin abin da za ta iya aunawa da kwatantawa.
Menene saka kalma?
Ƙirƙirar taswirorin kalma zuwa jerin lambobi (vector) ta yadda kalmomin da aka yi amfani da su su kasance kusa da juna a cikin wannan sararin lamba.
Ta yaya samfura irin su word2vec ke koyon ma'anar kalma?
Word2vec yana koya daga mahallin: kalmomin da suka bayyana a cikin rubutu iri ɗaya suna samun nau'i iri ɗaya. Ba a buƙatar ma'anar ɗan adam.
Mene ne babban bambanci tsakanin word2vec/GloVe embeddings da abubuwan da aka saka a cikin ƙirar taswira ta zamani?
Abubuwan da aka haɗa na gargajiya suna sanya vector guda ɗaya ga kowace kalma ba tare da la'akari da jumla ba; na'urorin lantarki suna daidaita vector bisa ga mahallin da ke kewaye, don haka 'banki' ya bambanta ta hanyar amfani.
Wane aiki ya fi dogara kai tsaye kan kwatanta haɗawa da ɓarna?
Binciken Semantic yana haɗa tambaya da takaddun, sannan ya nemo mafi kusa, yana maido da sakamakon da ya dace da ma'ana maimakon ainihin kalmomi.
Kusan lambobi nawa (girma) ke amfani da kalmar kalma2vec ko GloVe na yau da kullun?
Abubuwan da aka saba amfani da su na yau da kullun akan tsari na nau'ikan 100 zuwa 300, sun isa su kama dangantaka mai kyau ba tare da rashin ƙarfi ba.