Rukunin Rubutun Rubutun FastText
FastText hanya ce ta 2016 Facebook AI wacce ke wakiltar kowace kalma a matsayin jakar halayen n-grams, don haka yana iya gina vectors har ma da kalmomin da bai taɓa gani ba yayin horo.
Dubawa
This subword approach excels at morphologically rich languages, typos, and rare words where Word2Vec and GloVe fail.
Zurfafa nutsewa
FastText, wanda Facebook AI Research ya haɓaka (Bojanowski, Grave, Joulin, Mikolov) a cikin 2016, ya ƙaddamar da samfurin Skip-Gram ta hanyar karya kowace kalma zuwa cikin n-grams. Kalmar "inda" tare da n-grams na tsawon 3 ya zama <wh, wh, her, ere, re> tare da cikakkiyar alamar kalma, inda maƙallan kusurwa ke alamar iyakoki. Ma'anar kalma ita ce jimlar vector ta n-gram. Wannan yana nufin FastText zai iya tsara vector don kalmar da ba ta cikin ƙamus kamar "rashin imani" daga guntun kalmomin da aka saba da su, kuma yana ɗaukar ilimin halittar jiki da aka raba, don haka "gudu," "mai gudu," da "gudu" suna da alaƙa ta halitta. Wannan aikin kuma yana aika da sauri, daidaitaccen mai rarraba rubutu na layi ("fastText" yanayin kulawa) da ake amfani dashi don ayyuka kamar tantance harshe da yin alama a ma'auni mai girma.
Fahimtar Fasaha
Kowane hali n-gram an haɗe shi a cikin tebur mai ƙayyadaddun girman girman kuma an sanya nasa vector; wakilcin kalma shine jimlar abubuwan da ke tattare da su na n-gram, waɗanda aka horar da su da maƙasudin Skip-gram mara kyau kamar Word2Vec. Wannan raba sigogin ƙananan kalmomi a cikin kalmomi shine dalilin da yasa ilimin halittar jiki ke canzawa kuma dalilin da yasa kalmomin da ba a gani ba har yanzu suna samun ma'ana. Mai rarrabawa da ake kulawa yana amfani da samfurin jakunkuna iri ɗaya tare da Softmax na matsayi, yana mai da shi da sauri sosai akan CPUs.
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 Rubutun Subword na FastText
Ra'ayin ƙananan kalmomi na FastText ya tabbatar da tushe: masu canji na zamani suna amfani da dabaru masu alaƙa kamar Byte-Pair Encoding da WordPiece tokenization don sarrafa kowane shigarwa ba tare da tsayayyen ƙamus ba. Facebook ya fitar da pretrained FastText vectors don yaruka 157, yana kiyaye shi zuwa tushe don harsuna da yawa da ƙananan albarkatun NLP inda manyan samfuran ba su da amfani. Kamar yadda ƙananan na'urori da ƙirar gefen ke samun mahimmanci, ƙaramin sawun FastText da saurin CPU suna kiyaye shi dacewa don ƙirar rubutu.
Aiwatar da Gaskiyar Duniya
Ƙirƙirar ƙira don kuskuren rubutu ko kalmomin da ba a taɓa ganin su ba kamar "realy" ko sabbin sunayen samfur
Bude tushen Facebook wanda aka riga aka horar da vectors wanda ke rufe yaruka 157 don bincika harsuna da yawa da sanya alama.
Ganewar harshe mai saurin gaske da rarrabuwar kawuna/masu magana akan CPU ba tare da GPU ba
Karɓar harsuna masu wadatar dabi'a kamar Finnish ko Baturke inda kalmomi ke ɗaukar nau'i iri-iri
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 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
Jagora na gaba
Matryoshka Wakilin Abubuwan Haɗawa
Tambayoyin da ake yawan yi
What is FastText Subword Embeddings?
FastText hanya ce ta 2016 Facebook AI wacce ke wakiltar kowace kalma a matsayin jakar halayen n-grams, don haka yana iya gina vectors har ma da kalmomin da bai taɓa gani ba yayin horo. Wannan tsarin ƙaramar kalma ta yi fice a cikin harsuna masu wadatar halitta, rubutu, da kalmomin da ba kasafai ba inda Word2Vec da GloVe suka kasa.
Ta yaya FastText ke wakiltar kalma ɗaya?
FastText yana lalata kowace kalma zuwa nau'ikan n-grams kuma yana tattara ra'ayoyin su don samar da wakilcin kalmar.
Menene babban iyakance Word2Vec da GloVe FastText ya ci nasara?
Saboda FastText yana tsara vectors daga ƙananan kalmomi, yana iya gina wakilci don kalmomin da bai taɓa gani ba a horo.
Don kalmar "inda" mai haruffa 3 n-grams, wanda shine ingantaccen n-gram (tare da alamomin iyaka)?
"inda" yana samar da trigrams kamar <wh, wh, her, ere, re>, da cikakkiyar alamar kalma; "whe" yana daya daga cikinsu.
Wanne maƙasudin horarwa ne tsarin shigar da FastText ke ginawa akai?
FastText yana tsawaita Skip-gram tare da maƙasudin samfur mara kyau, yana ƙara ƙananan kalmomin n-gram.
Me yasa FastText ke da amfani musamman ga harsuna kamar Finnish ko Baturke?
Harsuna masu wadatar dabi'a suna samar da nau'ikan kalmomi da yawa; Raba ƙananan kalmomi vectors damar FastText ya danganta su ta halitta.