Xidhiidhka ugu sarreeya
Xidhiidhka ugu sarreeya (MMR) waa hab dib-u-qiimeyn ah oo dheellitiraya sida ay natiijadu u khusayso sida ay uga duwan tahay natiijooyinka hore loo doortay.
Dulmar
It matters because pure relevance ranking often returns near-duplicate passages that waste space in a RAG context window.
quusid qoto dheer
Marka nidaamka goobiddu uu keeno dukumeentiyo kaliya iyadoo la eegayo waydiinta, natiijooyinka ugu sarreeya ayaa ah kuwo aan badnayn - shan tuduc ayaa dhammaantood sheegaya wax isku mid ah. MMR, oo ay soo bandhigeen Carbonell iyo Goldstein sanadkii 1998, ayaa tan hagaajiya iyadoo mid mid u dooranaysa natiijooyinka. Tallaabo kasta waxay dooranaysaa musharraxa kordhinaya isku-dhafka miisaanka leh: lambda waxay waqtiyada ku habboon tahay weydiinta, laga jaray (1 laga jaray lambda) waxay la mid tahay wax kasta oo hore loo doortay. Lambda ku dhow 1 waxay jecel yihiin ku habboonaanta saafiga ah; ku dhow 0 waxay door bidaa kala duwanaanshaha. Jiilka la kordhiyey ee dib u soo celinta, MMR waxa ay caan ku tahay in la keeno qaybo kala duwan si qaabka luqaddu u arko caddaymo dhammaystiran oo aan ahayn xaqiiqo la mid ah oo lagu soo celceliyo, hagaajinta caymiska iyada oo aan la ballaarin macnaha guud.
Aragtida Farsamada
MMR waa hunguri weyn, algorithm-ku celcelin. Isku ekaanshaha iyo isku midka ah dukumeentiga labadaba waxaa badanaa lagu xisaabiyaa sida isku midka ah cosine ee u dhexeeya isku dhafka vectors. Habka dhibcaha waa: MMR = argmax ka badan dukumentiyada soo haray ee [lambda * sim (doc, query) - (1 - lambda) * max sim (doc, Selected) ]. Sababtoo ah waxay dib-u-qiimeyn ku sameyneysaa godadka soo koraya ee la doortay wareeg kasta, waa ku-tiirsanaan waxayna ku socotaa qiyaas ahaan O(k*n) isbarbardhigga isbarbardhigga k ee murashaxiinta n.
Saamaynta Istiraatijiyadeed
Xawaaraha iyo miisaanka
Socodka shaqada luqaddu si dhakhso leh ayay u socon kartaa iyada oo aan la hurayn joogteynta.
Helitaanka iyo gaarsiinta
Waxay balaadhisaa gelitaanka luqadaha iyo qaababka isgaarsiinta.
Go'aamo cad
Kooxuhu waxay waqti badan ku qaadan karaan xukunka halka otomaatiggu uu qabanayo ku celcelinta.
Mustaqbalka ugu sarreeya ee ku habboonaanta
MMR waxa ay ahaanaysaa mid khafiifa oo ah macaamiisha vector-database sida LangChain iyo Chroma, halkaas oo lagu bixiyo qaab dib u soo celin hal-line ah. Nidaamyada mustaqbalka waxay si sii kordheysa ula mideeyaan ujeedooyinka kala duwanaanta ee la bartay, xulashada kutlada ku saleysan, iyo cod-bixiyeyaal isdhaafsan oo u xukumaya cusub si macno ahaan ka badan masaafada kosine. Marka ay daaqadaha macnaha guud koraan, xooga lasaarku waxa uu ka wareegayaa kaydinta booska ilaa dajinta cadaymo dhammaystiran oo dhab ah, iyada oo la ilaalinayo kala-duwanaanshaha-ogsoonaanshaha doorashada sida MMR ee khuseeya xitaa marka awoodda cayriin ay badan tahay.
Dhaqangelinta Adduunka-dhabta ah
RAG chatbot waxay isticmaashaa dib u soo celinta MMR si ay 5 qaybood ee ugu sareysa ay daboolaan dhinacyo kala duwan oo siyaasad ah halkii ay ka ahaan lahaayeen shan cutub oo isku mid ah.
Aaladda soo koobidda cilmi-baadhistu waxay khusaysaa MMR si ay u soo qaadato marinnada yareeya isku-dhafka, soona saaraya soo koobid ballaadhan, soo noqnoqonaysa.
Isku duwaha wararka ayaa ku darajo maqaallada MMR si uu u muujiyo daboolka kala duwan ee dhacdada halkii uu toban xarumood ku celin lahaa hal sheeko siligga ah.
LangChain's vector retriever ayaa soo bandhigay search_type='mmr' oo wata fetch_k iyo lambda_mult si loo kala duwo dukumentiyada la soo celiyay.
Khatarta & Dariiqyada Ilaalada
Xaqiiqooyinka dhalanteed waxay si deggan u geli karaan warbixinnada, taageerada socodka, ama natiijooyinka cilmi-baarista.
Dareenka degdega ahi wuxuu abuuri karaa natiijooyin aan iswaafaqayn codsiyada la midka ah.
Xogta qoraalka xasaasiga ah ayaa laga yaabaa in la kashifo haddii kontaroolada gelitaanka ay daciif yihiin.
Qorshe Hawleedka Dhaqangelinta
Qeex qaabka wax soo saarka, codka, iyo heerarka tayada ka hor inta aan la baahin.
Jawaabaha salka ku haya ilo lagu kalsoon yahay mar kasta oo saxnidu ay muhiim tahay.
Hayso isbaarada dib u eegista bini aadamka ee wax soo saarka sare.
Lasoco qaababka guuldarada oo dib u leyli dardargelinta ama socodka shaqada si joogto ah.
Sii wad Sahaminta
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 Maximum Marginal Relevance 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
Hagaha xiga
Cadaadis Aan Ugu Sarreynin
Su'aalaha soo noqnoqda
What is Maximum Marginal Relevance?
Xidhiidhka ugu sarreeya (MMR) waa hab dib-u-qiimeyn ah oo dheellitiraya sida ay natiijadu u khusayso sida ay uga duwan tahay natiijooyinka hore loo doortay. Waa muhiim sababta oo ah darajada khusaysa saafiga ah waxay inta badan soo celisaa marino-koobi ah oo meel ku luminaya daaqada RAG.
Waa maxay dhibaatada aasaasiga ah ee MMR loogu talagalay in lagu xalliyo?
MMR waxa ay wax ka qabataa u janjeera ku habboonaanta-keliya darajada si ay u soo bandhigto natiijooyin badan oo sheegaya asal ahaan wax la mid ah, iyada oo abaalmarinaysa wax cusub.
Qaabka MMR, waa maxay qiimaha lambda ee ku dhow 1 xoogga saaraya?
ku habboonaanta miisaanka Lambda; ku dhow 1 ereyga khuseeya ayaa xukuma, halka ku dhow 0 kala duwanaanshaha (isku midka hoose) ereygu uu xukumo.
Sidee MMR u doorataa natiijooyinkeeda?
MMR waa hab ku celcelin hunguri weyn leh: mid kasta oo la soo doorto waxa uu kordhiyaa ku habboonaanta marka laga reebo u ekaanshaha ugu badan ee shayada hore loo doortay.
Waa maxay cabbirka isku midka ah ee inta badan lagu isticmaalo gudaha MMR ee gundhigga qoraalka?
Isku ekaanshaha cosine-ka ee udhaxeeya is-dhex-gelinta vectors waa doorashada caadiga ah ee khusaynta weydiinta iyo isbarbardhigga dukumeentiga dhexdooda ee MMR.
Maxay MMR si gaar ah faa'iido ugu leedahay jiilka la kordhiyay (RAG)?
Iyadoo la kala saarayo qaybo la soo saaray, MMR waxay ka caawisaa moodalku inuu arko caddaymo kala duwan halkii uu ka ahaan lahaa isla xaqiiqada lagu soo celceliyo, hagaajinta caymiska.