RecommendationsProblemsContent-BasedCollaborativeOtherHybridsCompromise Driven RetrievalConstraint SatisfactionCompletenessConclusionsSummaryRecommendationsCompromise Driven RetrievalConclusionsRecommender Systems as IDSSChad Hogg2006-11-13Chad Hogg Recommender Systems as IDSSRecommendationsCompromise Driven RetrievalConclusionsOutline1RecommendationsProblemsContent-BasedCollaborativeOtherHybrids2Compromise Driven RetrievalConstraint SatisfactionCompleteness3ConclusionsSummaryChad Hogg Recommender Systems as IDSSRecommendationsCompromise Driven RetrievalConclusionsProblemsContent-BasedCollaborativeOtherHybridsOutline1RecommendationsProblemsContent-BasedCollaborativeOtherHybrids2Compromise Driven RetrievalConstraint SatisfactionCompleteness3ConclusionsSummaryChad Hogg Recommender Systems as IDSSRecommendationsCompromise Driven RetrievalConclusionsProblemsContent-BasedCollaborativeOtherHybridsInformation OverloadThere is too much stuff for anyone to read / watch / buy /experience all of it.We would like to spend our time and money wisely, onthings of interest.How do we know what we won’t like without trying it?Chad Hogg Recommender Systems as IDSSRecommendationsCompromise Driven RetrievalConclusionsProblemsContent-BasedCollaborativeOtherHybridsInformation OverloadThere is too much stuff for anyone to read / watch / buy /experience all of it.We would like to spend our time and money wisely, onthings of interest.How do we know what we won’t like without trying it?Chad Hogg Recommender Systems as IDSSRecommendationsCompromise Driven RetrievalConclusionsProblemsContent-BasedCollaborativeOtherHybridsInformation OverloadThere is too much stuff for anyone to read / watch / buy /experience all of it.We would like to spend our time and money wisely, onthings of interest.How do we know what we won’t like without trying it?Chad Hogg Recommender Systems as IDSSRecommendationsCompromise Driven RetrievalConclusionsProblemsContent-BasedCollaborativeOtherHybridsMaking RecommendationsRecommender systems make predictions of what peoplewill enjoy.Typically, input is ratings of some items by a user.Output is a list of unrated items that may be of interest tothe user.Chad Hogg Recommender Systems as IDSSRecommendationsCompromise Driven RetrievalConclusionsProblemsContent-BasedCollaborativeOtherHybridsMaking RecommendationsRecommender systems make predictions of what peoplewill enjoy.Typically, input is ratings of some items by a user.Output is a list of unrated items that may be of interest tothe user.Chad Hogg Recommender Systems as IDSSRecommendationsCompromise Driven RetrievalConclusionsProblemsContent-BasedCollaborativeOtherHybridsMaking RecommendationsRecommender systems make predictions of what peoplewill enjoy.Typically, input is ratings of some items by a user.Output is a list of unrated items that may be of interest tothe user.Chad Hogg Recommender Systems as IDSSRecommendationsCompromise Driven RetrievalConclusionsProblemsContent-BasedCollaborativeOtherHybridsOutline1RecommendationsProblemsContent-BasedCollaborativeOtherHybrids2Compromise Driven RetrievalConstraint SatisfactionCompleteness3ConclusionsSummaryChad Hogg Recommender Systems as IDSSRecommendationsCompromise Driven RetrievalConclusionsProblemsContent-BasedCollaborativeOtherHybridsContent DataEarly recommender systems used information about rateditems.Inter-item similarity may be computed based on features.Each feature may be of a different type and have a localsimilarity metric.Items similar to those ranked highly by user will berecommended.Chad Hogg Recommender Systems as IDSSRecommendationsCompromise Driven RetrievalConclusionsProblemsContent-BasedCollaborativeOtherHybridsContent DataEarly recommender systems used information about rateditems.Inter-item similarity may be computed based on features.Each feature may be of a different type and have a localsimilarity metric.Items similar to those ranked highly by user will berecommended.Chad Hogg Recommender Systems as IDSSRecommendationsCompromise Driven RetrievalConclusionsProblemsContent-BasedCollaborativeOtherHybridsContent DataEarly recommender systems used information about rateditems.Inter-item similarity may be computed based on features.Each feature may be of a different type and have a localsimilarity metric.Items similar to those ranked highly by user will berecommended.Chad Hogg Recommender Systems as IDSSRecommendationsCompromise Driven RetrievalConclusionsProblemsContent-BasedCollaborativeOtherHybridsContent DataEarly recommender systems used information about rateditems.Inter-item similarity may be computed based on features.Each feature may be of a different type and have a localsimilarity metric.Items similar to those ranked highly by user will berecommended.Chad Hogg Recommender Systems as IDSSRecommendationsCompromise Driven RetrievalConclusionsProblemsContent-BasedCollaborativeOtherHybridsExample Content DataTitle Instructor Level Bldg Days TimeSys. Software Kay 100 PL MWF 8:00Databases Korth 200 PL MWF 14:00Graphics Huang 300 MG MWF 9:00Automata Munoz-Avila 300 MG MWF 14:00Pattern Rec. Baird 300 MG TR 11:00IDSS Munoz-Avila 300 LL MWF 9:00Chad Hogg Recommender Systems as IDSSRecommendationsCompromise Driven RetrievalConclusionsProblemsContent-BasedCollaborativeOtherHybridsExample Recommendation(Presume that courses are always taught by the sameprofessor, in the same room and at the same time.)Suppose Bob has previously taken Pattern Recognition,which he hated, and Automata, which he loved.IDSS would probably be a good choice for Bob, because ithas the same instructor, level, and days as another coursehe liked.Chad Hogg Recommender Systems as IDSSRecommendationsCompromise Driven RetrievalConclusionsProblemsContent-BasedCollaborativeOtherHybridsExample Recommendation(Presume that courses are always taught by the sameprofessor, in the same room and at the same time.)Suppose Bob has previously taken Pattern Recognition,which he hated, and Automata, which he loved.IDSS would probably be a good choice for Bob, because ithas the same instructor, level, and days as another coursehe liked.Chad Hogg Recommender Systems as IDSSRecommendationsCompromise Driven RetrievalConclusionsProblemsContent-BasedCollaborativeOtherHybridsExample Recommendation(Presume that courses are always taught by the sameprofessor, in the same room and at the same time.)Suppose Bob has previously taken Pattern Recognition,which he hated, and Automata, which he loved.IDSS would probably be a
View Full Document