Ngwa ngwa ngwa ngwa ngwa mgbakwunye
FastText bụ usoro 2016 Facebook AI nke na-anọchi anya okwu ọ bụla dị ka akpa nke agwa n-gram, ya mere ọ nwere ike wulite vectors ọbụna maka okwu ọ na-ahụtụbeghị n'oge ọzụzụ.
Nchịkọta
This subword approach excels at morphologically rich languages, typos, and rare words where Word2Vec and GloVe fail.
Ime miri emi
FastText, nke Facebook AI Research (Bojanowski, Grave, Joulin, Mikolov) mepụtara na 2016, na-agbatị ihe nlereanya Skip-Gram site n'imebi okwu ọ bụla n'ime n-gram agwa. Okwu "ebe" nwere n-gram nke ogologo 3 na-aghọ <wh, wh, her, ere, re> gbakwunyere akara okwu zuru oke, ebe akara nkuku na-aka akara oke okwu. Vector nke okwu bụ nchikota vector n-gram ya. Nke a pụtara FastText nwere ike ịdepụta vector maka okwu na-apụ apụ dị ka "enweghị ike ikweta" site na mkpụrụ okwu subword a ma ama, ọ na-ejidekwa morphology na-ekerịta, ya mere "na-agba ọsọ," "onye na-agba ọsọ," na "na-agba ọsọ" na-ejikọta ya. Otu oru ngo a na-ebufekwa ngwa ngwa nkesa ederede linear ziri ezi (ụdị "fastText" a na-elekọta) ejiri maka ọrụ dị ka njirimara asụsụ na mkpado n'ogo buru ibu.
Nghọta nka nka
A na-awụnye n-gram agwa ọ bụla n'ime tebụl ịwụ nwere oke ma kenye vector nke ya; Nnọchite anya okwu bụ nchikota nke ihe mejupụtara ya n-gram vectors, zụrụ ya n'otu ebumnobi Skip-gram na-adịghị mma dịka Word2Vec. Nkesa nke mkpụrụokwu subword n'ofe okwu bụ ihe kpatara morphology na-ebufe yana ihe kpatara na okwu ndị a na-adịghị ahụ anya ka ji bụrụ ihe ezi uche dị na ya. Nkewa nke a na-elekọta na-eji ụdị akpa-nke njirimara nwere softmax hierarchical, na-eme ka ọ dị oke ọsọ na CPUs.
Mmetụta atụmatụ
Ọsọ na ọnụ ọgụgụ
Usoro ọrụ asụsụ nwere ike ịga ngwa ngwa n'achụghị nkwụsi ike.
Nweta na iru
Ọ na-agbasawanye ohere n'ofe asụsụ na ụdị nzikọrịta ozi.
Mkpebi doro anya
Ndị otu nwere ike itinyekwu oge na ikpe ebe akpaaka na-ejikwa nkwughachi.
Ọdịnihu nke ntinye okwu Subword FastText
Echiche subword FastText gosipụtara na ntọala: ndị na-agbanwe ọgbara ọhụrụ na-eji usoro ndị metụtara ya dị ka Byte-Pair Encoding na WordPiece tokenization iji jikwaa ntinye ọ bụla na-enweghị okwu edobere. Facebook weputara vector FastText a zụrụ azụ maka asụsụ 157, na-edobe ya ka ọ bụrụ ntọala maka asụsụ ọtụtụ asụsụ na NLP dị ala ebe nnukwu ụdị na-agaghị ekwe omume. Dị ka obere na ngwaọrụ na ihu ụdị na-enweta mkpa, FastText obere akara ukwu na ọsọ CPU na-eme ka ọ dị mkpa maka nhazi ederede mmepụta.
Mmejuputa n'ezie n'ụwa
Na-emepụta vectors maka mkpụrụokwu ndị ezighi ezi ma ọ bụ nke a na-ahụtụbeghị mbụ dị ka "ezigbo" ma ọ bụ aha ngwaahịa ọhụrụ
Vectors a zụrụ azụ nke Facebook mepere asụsụ 157 maka ọchụchọ na mkpado ọtụtụ asụsụ.
Nchọpụta asụsụ dị elu yana nhazi spam/isiokwu na CPU na-enweghị GPU
Ijikwa asụsụ bara ọgaranya dị ka Finnish ma ọ bụ Turkish ebe okwu nwere ọtụtụ ụdị
Ihe ize ndụ & okporo ụzọ nche
Eziokwu ndị e chepụtara echepụta nwere ike jiri nwayọ tinye akụkọ, nkwado nkwado, ma ọ bụ nsonaazụ nyocha.
Mmetụta ngwa ngwa nwere ike ịmepụta nsonaazụ na-ekwekọghị ekwekọ n'ofe arịrịọ ndị yiri ya.
Enwere ike ikpughe data ederede nwere mmetụta ma ọ bụrụ na njikwa ohere adịghị ike.
Map mmejuputa
Kọwaa usoro mmepụta, ụda, na ụkpụrụ ịdịmma tupu ibugharị.
Weghachite nzaghachi site na isi mmalite ntụkwasị obi mgbe ọ bụla izi ezi dị mkpa.
Debe ebe nleba anya mmadụ maka mpụta dị elu.
Sochie ụkpụrụ ọdịda ma na-azụghachi mkpali ma ọ bụ usoro ọrụ mgbe niile.
Nọgide na-eme nchọpụta
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 FastText Subword 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
Ntuziaka na-esote
Ntinye ihe nnọchi anya Matryoshka
Ajụjụ a na-ajụkarị
What is FastText Subword Embeddings?
FastText bụ usoro 2016 Facebook AI nke na-anọchi anya okwu ọ bụla dị ka akpa nke agwa n-gram, ya mere ọ nwere ike wulite vectors ọbụna maka okwu ọ na-ahụtụbeghị n'oge ọzụzụ. Ụzọ obere okwu a kacha mma n'asụsụ bara ụba nke ọdịdị ọdịdị, typos, na okwu ndị na-adịghị ahụkebe ebe Word2Vec na GloVe dara.
Kedu ka FastText si anọchi anya otu mkpụrụokwu?
FastText na-emebi mkpụrụokwu ọ bụla ka ọ bụrụ n-gram agwa wee chịkọta vectors ha ka ọ bụrụ nnọchi okwu ahụ.
Kedu oke oke nke Word2Vec na GloVe FastText meriri?
N'ihi na FastText na-edepụta vectors site na mpempe okwu subword, ọ nwere ike wulite ihe nnọchianya maka okwu ọ na-ahụtụbeghị na ọzụzụ.
Maka okwu "ebee" nwere n-gram 3-agwa, nke bụ n-gram dị irè (nwere akara oke)?
"ebe" na-ewepụta trigram dị ka <wh, wh, her, ere, re>, gbakwunyere akara okwu zuru oke; "whe" bụ otu n'ime ha.
Kedu ebumnuche nkuzi nke FastText na-ewuli elu?
FastText na-agbatị Skip-gram na ebumnobi nlele na-adịghị mma, na-agbakwunye mkpụrụ okwu n-gram vectors.
Kedu ihe kpatara FastText ji baa uru karịsịa maka asụsụ dị ka Finnish ma ọ bụ Turkish?
Asụsụ ndị bara ọgaranya na-emepụta ọtụtụ ụdị okwu; Ịkekọrịta subword vector na-ahapụ FastText ka ọ kọwaa ha n'ụzọ nkịtị.