Frase-BERT
Sentence-BERT (SBERT) dafay méngale BERT ak benn vecteur bu am guddaay bu takku ngir benn frase lëmm, suko defee ñu mëna méngale luñuy tekki ak nuru cosine bu gaaw.
Résumé
It made semantic search and clustering over millions of sentences practical, turning a job that took BERT hours into milliseconds.
Plongeur bu xóot
BERT bu leer mën na méngale ñaari frase ngir nuru, waaye du lenn ludul boole ñoom ñaar ci reso bi, te loolu dafa yeex lool ci escalier bi: méngale 10,000 frase ñaar-ñaar dina soxla lu tollu ci 50 milioŋ ci paas yu jëm kanam. Sentence-BERT, bi Reimers ak Gurevych dugal ci 2019, dafa defar lii ci jëfandikoo reso siamese (seex): ñaari tour BERT ak poids yu ñu bokk bu nekk encode benn frase ci boppam, ginaaw ga benn jéego pooling (dafay faral di tekki pooling ci kaw token embeddings) dafay joxe benn vecteur ci frase bu nekk. Dañu defar model bi bu baax suko defee frase yu nuru ci wàllu semantik ñu jege seen biir ci espace vecteur bi. Leegi frase bu nekk dañu koy kode benn yoon ci benn embedding buñu mëna jëfandikoowaat, ba noppi nuru nekk na produit dot bu yomb, may seetlu, deduplication, ak clustering ci escale bu mag.
Gis-gis xarala
SBERT dañu koy faral di tàggat ci architecture siamese ak objectif bu wuute wala bu am ñatti mbir. Done yiñ jëlee ci làkk wiñ nàmp barina luñu koy gis: dañuy boole ñaari entailment, ñuy tàqale lu wuute. Ñaari tour yi bokk nañu poids, kon encodage bi dafay symétrique. Moyenne pooling ci kaw vecteur token yu mujj yi dañuy gëna mëna jëfandikoo token [CLS] kese, di defar ay embeddings fu nuru cosine di topp jegewaale semantik bi.
njeextalu pexe
Gaawaay ak yaatuwaay
Liggéeyukaay yi ci làkk yi mën nañu gëna gaaw te duñu yàq deggoo gi.
Dugg ak yegg
Dafay yaatal jëfandikoo gi ci làkk yi ak ci anam yi ñuy jokkoo.
dogal yu gëna leer
Ekip yi mën nañu gëna yàgg ci àtte ci jamono ji otomatisation di liggéey ci baamtu.
Ëlëgu santaane-BERT
SBERT-style bi-encoders leegi dafay jàppale jikkoom-yokkum, di dundal contexte bi war ci modeli làkk yu yaatu. Barab bi mingi jubal ci xeeti samp yu gëna mag yuñ defaree ay tegtal, samp yu bari làkk ak anam yu bari, ak misaali Matryoshka yuñ mëna dagg dimension yi ngir gaawaay. Tuyo hybrid yi dañuy boole bi-encoder yu gaaw yi ak cross-encoder yu gëna néew, boole ci eskaalu SBERT ak gëna dëggu ci kaw kàndidaa yi.
Doxal ci àdduna dëgg
Motëri seetlu semantik yi dañuy dugal ab laaj ak bépp këyit, ba noppi ñu dello vecteur yi gëna jege ci barabu wéeru ci baatu-caabi yiy jaxasoo.
Sistemu defar yiñ yokk ci seet dañuy jëfandikoo SBERT ngir seet ay pàcc yu am solo ngir gëna dëgëral tontu chatbot yi.
Jumtukaayi jàppale kiliyaan yi dañuy boole këyit yiy dugg ci boole ci saasi lu nuru ci ñaari yoon wala jafe-jafe yu jëm ci loolu.
Bibliothèque Python biy soppi frase yi dafay joxe ay model SBERT yuñ tàggat bu njëkk ngir mine paraphrase ak deduplicate mbind yu nuru lool.
Risk yi ak balustrade yi
Lépp lu jaarul yoon mën na dugg ci rapoor yi, jàppale ci liggéey bi, wala ci njariñu gëstu bi.
Sensibilite bu gaaw mën na jur njariñ yu wuute ci laajte yu noonu mel.
Done yu am solo mën nañu feeñ sudee seytu jëfandikoo gi néew doole.
Roadmap ngir samp gi
Mandargal formaa génne gi, melokaan bi, ak standard kalite yi laata ngay dugal ko.
Tontu yu am solo ak balluwaay yu wóor saa yu dëggu bi di am solo.
Fexeel am barabu xool nit ñi ngir am njariñ yu am solo.
Toppal anami gacce yi ak di faral di tàggataat ay laaj wala def-liggéey.
Weyal di banneexu
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 Sentence-BERT 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
Gis bi ci topp
HyDE Document Hypothétique
Laaj yi ñuy faral di laaj
What is Sentence-BERT Embeddings?
Sentence-BERT (SBERT) dafay méngale BERT ak benn vecteur bu am guddaay bu takku ngir benn frase lëmm, suko defee ñu mëna méngale luñuy tekki ak nuru cosine bu gaaw. Defna seetlu semantik ak dajale lu ëpp ay milioŋ ciy santaane nekk lu am solo, mu soppi liggéey bu jël ay waxtu BERT ci ay milisegond.
Ban jafe-jafe bu mag la Sentence-BERT di saafara ci BERT bu leer bi?
BERT bu leer dafa wara liggéey ci ñaari frase yuñ boole, kon méngale ñaar ci escale bu yaatu mënul am ci wàllu ordinatër; SBERT dafay kode frase bu nekk benn yoon ci vecteur buñ mëna jëfandikoowaat.
Ban jumtukaayu reso la Sentence-BERT di jëfandikoo?
SBERT dafay jëfandikoo structure siamese (seex) fu ñaari encodeur BERT bokk poids te ku nekk ci ñoom dugal benn frase ci boppam.
Ban pexem pooling lañu gëna digal ñu boole ci SBERT?
Moyen pooling ci kaw vecteur token yu mujj yi dafay gëna am njariñ ci jëfandikoo token [CLS] kese ngir samp ay frase.
Buñu boole ay frase, naka lañuy nattee seen nuru?
Li SBERT bëgga wax mooy nuru gi dafay wàññi cosine/dot-produit bu yomb diggante vecteur yuñ njëkka xayma.
Ban xeetu done lañuy gëna jëfandikoo ngir defar SBERT?
Done NLI dañu leen di jëfandikoo bu baax: ñaari entailment dañu leen boole, ba noppi ñu tàqale lu wuute, daal di jëmmal barab bu am solo.