Ururimi AI

Amagambo-BERT

Interuro-BERT (SBERT) ihuza BERT kugirango itange icyerekezo kimwe-kirekire cyerekeranye ninteruro yose, ibisobanuro rero birashobora kugereranywa na cosine yihuta.

2 min somaIbiherutse kuvugururwa

Incamake

It made semantic search and clustering over millions of sentences practical, turning a job that took BERT hours into milliseconds.

Kwibira cyane

Ikibaya cya BERT gishobora kugereranya interuro ebyiri zisa, ariko mugaburira byombi hamwe binyuze murusobe, bikaba bitinda cyane mubipimo: kugereranya interuro 10,000 hamwe byombi bisaba pasiporo zigera kuri miliyoni 50. Interuro-BERT, yatangijwe muri 2019 na Reimers na Gurevych, ikosora ibi ukoresheje umuyoboro wa siamese (impanga): iminara ibiri ya BERT ifite uburemere busangiwe buri kodegisi imwe interuro yigenga, hanyuma intambwe yo guhuriza hamwe (mubisanzwe bivuze guhuriza hamwe ibimenyetso byerekana ibimenyetso) itanga vector imwe kumurongo. Icyitegererezo cyateguwe neza kuburyo interuro isa ninteruro yegeranye hamwe mumwanya wa vector. Noneho buri nteruro yashizwemo inshuro imwe mugushiramo gukoreshwa, kandi ibisa nabyo biba ibicuruzwa bihendutse bihendutse, bigafasha gushakisha, kugabanywa, no guhuriza hamwe murwego runini.

Ubushishozi

SBERT isanzwe itozwa hamwe nubwubatsi bwa siamese hamwe nintego itandukanye cyangwa itatu. Ururimi Kamere Yerekana amakuru arasanzwe: ibice bibiri bikururwa hamwe, kwivuguruza gusunitswe. Iminara yombi isangiye uburemere, bityo kodegisi irasa. Hagati yo guhuriza hamwe hejuru yicyerekezo cya nyuma muri rusange irusha imbaraga ikoresheje ikimenyetso cya [CLS] cyonyine, gitanga ibyashizwemo aho cosine isa neza ikurikirana neza hafi.

Ingaruka z'Ingamba

Umuvuduko n'igipimo

Ururimi rwakazi rushobora kugenda byihuse nta gutamba guhuzagurika.

Kugera no kugera

Yagura uburyo bwindimi nuburyo bwo gutumanaho.

Ibyemezo bisobanutse

Amakipe arashobora kumara umwanya munini murubanza mugihe automatike ikora gusubiramo.

Ejo hazaza h'interuro-BERT

SBERT-yuburyo bwa bi-encoders ubu bushimangira ibisekuruza byongerewe imbaraga, bigaburira imiterere bijyanye nururimi runini. Umwanya ugenda werekeza ku nini nini yerekana amabwiriza yashizwemo, gushyiramo indimi nyinshi na multimodal, hamwe na Matryoshka uhagarariye ibipimo bishobora kugabanywa ku muvuduko. Imiyoboro ya Hybrid ihuza bi-kodegisi yihuse hamwe na buhoro buhoro cross-encoder yongeye gutondekanya, igahuza igipimo cya SBERT hamwe nibisobanuro birambuye kubakandida ba mbere.

Gushyira mu bikorwa Isi

Imashini ishakisha semantike yashizemo ikibazo ninyandiko zose, hanyuma usubize ibice byegeranye aho kwishingikiriza ijambo ryibanze.

Sisitemu yo kugarura-yongerewe sisitemu ikoresha SBERT yashyizwemo kugirango ibone ibice bijyanye kugirango ushireho ibisubizo bya chatbot.

Ibikoresho-byunganira abakiriya cluster itike yinjira mugushiramo ibisa nitsinda ryigana cyangwa ibibazo bifitanye isano byikora.

Guhindura interuro isomero rya Python ritanga icyitegererezo cya SBERT cyo gucukura amabuye y'agaciro no kugabanya inyandiko hafi-imwe.

Ingaruka & Kurinda

Ibintu bifatika bishobora kwinjiza bucece raporo, gushyigikira imigendekere, cyangwa ibisubizo byubushakashatsi.

Kwihuta byihuse birashobora gukora ibisubizo bidahuye mubisabwa bisa.

Ibyanditswe byumvikana birashobora kugaragara niba kugenzura kugenzura ari ntege.

Igishushanyo mbonera

1

Sobanura imiterere isohoka, amajwi, hamwe nubuziranenge mbere yo gutangira.

2

Ibisubizo byibanze hamwe nisoko yizewe igihe cyose ukuri kwingirakamaro.

3

Komeza kugenzura abantu kugenzura ibisubizo byinshi.

4

Kurikirana uburyo bwo kunanirwa no kongera imyitozo cyangwa akazi gahoraho.

Komeza Ubushakashatsi

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.

Tangira ikibazo

Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation

Ubuyobozi bukurikira

HyDE Hypothetical Document Embeddings

Ibibazo bikunze kubazwa

What is Sentence-BERT Embeddings?

Interuro-BERT (SBERT) ihuza BERT kugirango itange icyerekezo kimwe-kirekire cyerekeranye ninteruro yose, ibisobanuro rero birashobora kugereranywa na cosine yihuta. Cyakoze gushakisha ibisobanuro no guhuriza hamwe miriyoni zinteruro zifatika, bihindura akazi kafashe amasaha BERT muri milisegonda.

Ni ikihe kibazo cyibanze hamwe na BERT isanzwe interuro-BERT ikemura?

Ikibaya cya BERT kigomba gutunganya interuro ebyiri, bityo rero ingano nini yo kugereranya ntishobora kubarwa; SBERT ikora buri nteruro inshuro imwe muri vector ikoreshwa.

Ni ubuhe buryo bwubaka urusobe Sentence-BERT ikoresha?

SBERT ikoresha imiterere ya siamese (impanga) aho kodegisi ebyiri za BERT zisangira uburemere kandi buri kimwe kigashyiramo interuro imwe yigenga.

Ni ubuhe buryo bwo guhuriza hamwe bukunze gusabwa gushira SBERT?

Kugereranya guhuriza hejuru yikimenyetso cya nyuma cyerekanwe muri rusange kurenza ukoresheje ikimenyetso cya [CLS] cyonyine kugirango ushiremo interuro.

Iyo interuro zimaze gushyirwamo, ni gute ubusanzwe busa?

Ingingo yose ya SBERT nuko guhuza bigabanuka kuri cosine ihendutse / akadomo-gicuruzwa ugereranije hagati ya vectors.

Ni ubuhe bwoko bwa dataset bukoreshwa muguhuza neza SBERT?

Amakuru ya NLI arakoreshwa cyane: ibice bibiri bikururwa hamwe kandi kwivuguruza gusunikwa, gushiraho umwanya ushimishije.