Ururimi AI

Kugarura Ibice Byinshi

Ibyinshi Byakuweho (DPR) isanga inyandiko ijyanye no kugereranya ibisobanuro byikibazo nibice nkibice byumubare, ntabwo bihuye namagambo.

2 min somaIbiherutse kuvugururwa

Incamake

It matters because it can retrieve correct answers even when the query and the document share zero vocabulary.

Kwibira cyane

DPR, yatangijwe na Facebook AI muri 2020, ikoresha kodegisi ebyiri zitandukanye za BERT: kodegisi yibibazo hamwe na kodegisi. Buri kimwe gihindura inyandiko muburyo burebure buringaniye (akenshi 768). Ibyingenzi nigicuruzwa cyadomo hagati yikibazo cyibibazo na vectori, bityo kugarura bihinduka byihuse-umuturanyi ushakisha hejuru yibice byashizwemo. Icyitegererezo cyatojwe gifite intego zinyuranye: gukurura icyerekezo cyiburyo hafi yikibazo no gusunika ibitari byiza, ukoresheje ibibi byongeweho wongeyeho ibibi byacukuwe muri BM25. Kumurongo ufunguye QA ibipimo nkibibazo bisanzwe, DPR yatsinze BM25 yiganje cyane ku ntera nini, byerekana ko guhuza ibisobanuro byizewe bishobora kurenza ijambo ryibanze gushakisha ibisubizo.

Ubushishozi

DPR ni bi-koderi: ikubiyemo ikibazo na buri gice cyigenga, bityo ibice byose byaciwe bibarwa rimwe bikabikwa mubipimo byerekana (urugero, FAISS). Mugihe cyibibazo uhisha gusa ikibazo, hanyuma ukore hafi yishakisha ryabaturanyi. Amahugurwa ashingiye kubibi-bindi - ibindi bice muri mini-batch imwe bitanga ingero mbi hafi yubusa, ibyo bigatuma umwe muburyo bwiza butanga ibigereranyo byinshi bitandukanye.

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.

Kazoza Kuzana Ibice Byinshi

Kugarura kwinshi ubu gushigikira imiyoboro myinshi yo kugarura-yongerewe imiyoboro igaburira imvugo nini y'ururimi. Ubushakashatsi bugenda bugana kuri sisitemu ya Hybrid ihuza amanota yuzuye namagambo, moderi yatinze-imikoranire nka ColBERT igumisha kuri buri kimenyetso kugirango ihuze neza, hamwe ninyandiko-yashizwemo ihuza imirimo myinshi. Tegereza kodegisi zihendutse, indimi nyinshi, hamwe nigihe kirekire-kodegisi, wongeyeho gufatanya-gufatanya-gutoza hamwe nabaguzi batanga.

Gushyira mu bikorwa Isi

Gufungura-domaine ikibazo gisubiza sisitemu ikurura ibice bya Wikipedia mbere yuko LLM yandika igisubizo

Gushakisha inyandiko zumushinga aho abakozi babaza ibibazo bisanzwe bakabona paragarafu zijyanye nubwo nta jambo ryibanze ryibanze

Bots-infashanyo yumukiriya kugarura ingingo yukuri ifasha-ikigo kuva mubibazo bisobanuwe

Gukuramo-byongerewe ibiganiro biganisha kubisubizo mubumenyi bwihariye kugirango ugabanye salusiyo

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 Dense Passage Retrieval 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

ColBERT Yatinze Imikoranire

Ibibazo bikunze kubazwa

What is Dense Passage Retrieval?

Ibyinshi Byakuweho (DPR) isanga inyandiko ijyanye no kugereranya ibisobanuro byikibazo nibice nkibice byumubare, ntabwo bihuye namagambo. Nibyingenzi kuko irashobora kubona ibisubizo nyabyo nubwo ikibazo hamwe ninyandiko bisangiye zeru amagambo.

Ni ikihe gitekerezo cyibanze cyihishe inyuma ya Dense Passage?

Ikarita ya DPR yerekana ibibazo n'ibice mubice byinshi kandi bipima isano bifitanye isano (ibicuruzwa byadomo), bifata ibisobanuro aho guhuza ijambo.

DPR ikoresha 'bi-encoder' yubatswe. Ibyo bivuze iki?

Bi-encoder ifite kodegisi yigenga kubibazo nibice, bityo ibice byinzira birashobora kubarwa no gutondekwa mbere yigihe.

Ni ukubera iki ibice bishobora kubarwa mbere muri DPR?

Kubera ko kodegisi ya parike idashingiye kubibazo, ibyanditswe byose birashobora kubanzirizwa no kubikwa, hasigara gusa ikibazo cyo gushishoza mugihe cyo gushakisha.

Ni ubuhe buryo bwo guhugura butuma DPR itanga ingero nyinshi mbi bihendutse?

Muri-batch ibibi bifata ibindi bice muri mini-batch nkibibi kubibazo runaka, bitanga ibice byinshi bitandukanye.

Ni ikihe gikoresho gikunze gukoreshwa kugirango DPR yegerejwe-umuturanyi ishakisha byihuse ku gipimo?

Indangantego za Vector nka FAISS zikora byihuse hafi-umuturanyi ushakisha hejuru ya miriyoni zashizwe imbere.