资源描述：
《A_Survey_of_Explanations_in_Recommender_Systems.pdf》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。
1、ASurveyofExplanationsinRecommenderSystemsNavaTintarev,JudithMasthoffUniversityofAberdeen,Scotland,U.K.ntintare,jmasthoff@csd.abdn.ac.ukAbstract2Whyexplanationsarethebestthingsinceslicedbread...Thispaperprovidesacomprehensivereviewofexplana-tionsinrecommendersyste
2、ms.Wehighlightsevenpossibleadvantagesofanexplanationfacility,anddescribehowex-istingmeasurescanbeusedtoevaluatethequalityofex-Table1.AimsAimDefinitionplanations.Sinceexplanationsarenotindependentoftherecommendationprocess,weconsiderhowthewaysrecom-Transparency(Tra
3、.)Explainhowthesystemworksmendationsarepresentedmayaffectexplanations.Next,weScrutability(Scr.)Allowuserstotellthesystemitlookatdifferentwaysofinteractingwithexplanations.Theiswrongpaperisillustratedwithexamplesofexplanationsthrough-TrustIncreaseusers’confidencein
4、out,wherepossiblefromexistingapplications.thesystemEffectiveness(Efk.)HelpusersmakegooddecisionsPersuasivenessConvinceuserstotryorbuy1Introduction(Pers.)Efficiency(Efc.)HelpusersmakedecisionsfasterThehistoryofexplanationsinintelligentsystemsbeganSatisfaction(Sat.)
5、Increasetheeaseofusabilityorwithexpertsystemswhichwerepredominantlybasedonenjoymentheuristics[7],butalsooncase-basedreasoning(CBR)[12],andmodelbasedapproaches[13].Inrecentyearstheirmorecommercialorentertainmentinclinedsuccessors-Amongotherthings,goodexplanationsc
6、ouldhelpin-recommendersystems-havebeguntoofferexplanationsasspireusertrustandloyalty,increasesatisfaction,makeitwell[5,18,24].Thesesystemsrepresentuserpreferencesquickerandeasierforuserstofindwhattheywant,andper-forthepurposeofsuggestingitemstopurchaseorexamine,su
7、adethemtotryorpurchasearecommendeditem.Table1i.e.recommendations.definessevenpossibleaimsofexplanationfacilitiesinrec-Intherecommendersystemscommunityitisincreas-ommendersystems.Someoftheseaimsaresimilartotheinglyrecognizedthataccuracymetricssuchasmeanreasonsforex
8、plainingreasoninginexpertsystems,c.f.[7].averageerror(MAE),precisionandrecall,canonlyInTable2wesummarizewhichoftheseaimsanumberpartiallyevaluatearecommendersys
