Collaborative Filtering and Recommender SystemsPresentation OutlineCollaborative Filtering definedRecommender systems definedHobo symbols from http://www.slackaction.com/signroll.htmHobo symbols as RS?Compare to todayGlance, Arregui & Dardenne (1997)Slide 9Konstan, Miller, et al. (1997)Slide 11Slide 12Proctor & McKinlay (1997)Slide 14Slide 15Slide 16ConclusionsReferencesQuestionsCollaborative Filtering and Recommender SystemsBrian LewisINF 385Q Knowledge Management SystemsNovember 10, 20052Presentation OutlineCollaborative filtering and recommender systems definedNovel exampleReadings - overview & key conceptsGlance, Arregui & Dardenne (1997)Konstan, Miller, et al. (1997)Proctor & McKinlay (1997)ConclusionsReferences3Collaborative Filtering defined"Based on the premise that people looking for information should be able to make use of what others have already found and evaluated." (Maltz & Ehrlich, 1995)"Technique for dealing with overload in information environments" (Procter & McKinlay, 1997)4Recommender systems definedSystems that evaluate quality based on the preferences of others with a similar point of view5Hobo symbols from http://www.slackaction.com/signroll.htm6Hobo symbols as RS?Specific to a communityImplicit and explicit signsFiltered through encodingCold-start problem?7Compare to todayRecommendDon't recommend8Glance, Arregui & Dardenne (1997)Knowledge PumpDesigned for use with an electronic repositoryDocument management and recommendationCommunity-centered collaborative filteringCharacteristicsSocial filteringContent-based filtering9Glance, Arregui & Dardenne (1997)User-item matrix of ratings10Konstan, Miller, et al. (1997)GroupLensPilot study - Usenet newsRating systemIntegrate into an existing system/existing usersUse existing applications - open architectureCharacteristicsHigh volume / high turnoverHigh noise information resourceSparse set of ratingsPredictive utility cost/benefit11Konstan, Miller, et al. (1997)Predictive utilityRisk - costs of misses andfalse positivesBenefit - values of hits and correct rejectionsUsenet has high predictive utilityHigh volumeValue of correct rejection is highRisk of a miss is low12Konstan, Miller, et al. (1997)ChallengesRatings sparsity"first-rater" problemPartition articles into clustersCapture implicit ratingsFilter botsPerformance challengesSystem architectureComposite users13Proctor & McKinlay (1997)Social Affordances and Implicit RatingsHow implicit approaches might be improvedSources of rating and recommendation dataContext of ratings and recommendationsReal and virtual groupsPrivacy and accessibility14Proctor & McKinlay (1997)CharacteristicsExplicit ratings systemsReader ratings based approach is expensiveHow do you deal with trust issues?Implicit ratings systemsFree to usersHow do you capture context?15Proctor & McKinlay (1997)Social Affordances"…making the potential for social (inter)action visible."How can activities be made visible? (explicitly)Web bookmarksSharable annotationsHow can activities be made visible? (implicitly)Copy browsing behavior of experts (virtual groups)Documents context in a group of documents (discourse analysis)Temporal coherence16Proctor & McKinlay (1997)Extracting implicit ratings from web behaviorVirtual group proxiesProxy cache analysisNominal ratingFrequencySequential accountabilityDistributional accountabilitySourcesTopical coherenceTemporal coherencePrivacy Issues17ConclusionsMany different issuesDiverse domains / communitiesDiverse content needsContext dependentNature of informationPredictive utilityVery creative solutions to draw from18ReferencesGlance, N., Arregui, D., & Dardenne, M. (1997). Knowledge Pump: Community-centered collaborative filtering. 5th DELOS workshop on filtering and collaborative filtering, Budapest, Hungary.Konstan, J., Miller, B., Maltz, D., Herlocker, J., Gordon, L. and Riedl, J. (1997), Applying collaborative filtering to usenet news, Communication of the ACM, 40(3), 77-87.Maltz, D. and Ehrlick, K. (1995). Pointing the way: active collaborative filtering. CHI '95, ACM Press.Procter, R. and A. McKinley (1997). Social affordances and implicit ratings for social filtering on the Web. DELOS workshop on collaborative filtering, Budapest, Hungary.19QuestionsQuestions live
View Full Document