Learning, Expertise amp; Individual Differences专业知识的学习amp；个体差异.ppt

《Learning, Expertise amp; Individual Differences专业知识的学习amp；个体差异.ppt》由会员分享，可在线阅读，更多相关《Learning, Expertise amp; Individual Differences专业知识的学习amp；个体差异.ppt（56页珍藏版）》请在三一办公上搜索。

1、A Framework for Personalization:,Susan DumaisMicrosoft Rhttp:/,Delos-NSF Workshop:June 18-20,2001,When do you want to go Where Everybody Knows Your Name(and mailing address,and preferences,and last 50 web pages visited)?,June 18,2001,Delos-NSF Workshop,A Working Definition,Outcome(t)=f(Action(t),Per

2、sonalHistory(t-n)Examples,Relevance feedbackContent-based filteringCollaborative filteringCaching,history lists,auto completion,MRUImplicit queries,Rememberance Agent,Watson,KenjinMyYahoo!,MyAOL,MyMSN,MyLibrary,etc.AltaVistas MySearch,iLOR,June 18,2001,Delos-NSF Workshop,A Demonstration:What do you

3、see?,June 18,2001,Delos-NSF Workshop,Many Kinds of Individual Differences,Task“info need”Short-term,relevance feedbackLong-term,content-based filteringPreferences,e.g.,CFExpertise,domain and application Cognitive aptitudesVerbal,spatial,reasoning skills,etc.DemographicsAge,major,gender,location,etc.

4、Cognitive styles,personality and affect,June 18,2001,Delos-NSF Workshop,Individual Differences Are,LargeSystematicSystems can often be modified to accommodateE.g.,robust systemsE.g.,personalization,June 18,2001,Delos-NSF Workshop,How Big Are Individual Diffs?,E.g.,Web searching(Chen Intermediate web

5、/search experience30 search tasks(e.g.,Home page for“Seattle Weekly”)Average RT(seconds)=52.3 secondsIndividual subjects average RT:69,30,76,48,29,68,69,49,75,62,64,69,26,89,50,44,54,35,39,30,71,56,28,59,36,67,93,37,39,49,28,89,37,36,31,47,66,62,51,30,40,38,31,70,37,36,36,88,41,50,84,68,42,58,34,25,

6、23,22,41,62,35,41,41,60,36,56,78,144,43,58,58,45,38,115,June 18,2001,Delos-NSF Workshop,Characterizing Indiv Diffs,HistogramMax:Min144,22=6.5:1Q3:Q166,36=1.8:1SD/X.42,June 18,2001,Delos-NSF Workshop,Example Individual Diffs,June 18,2001,Delos-NSF Workshop,Individual Diffs Correlated w/Performance in

7、 HCI/IR Tasks,Experience both application and domainReasoning(Egan et al.;Card et al.;Greene et al.)Spatial abilities(Egan Konvalina et al.)Personality and affect Gender,June 18,2001,Delos-NSF Workshop,Framework for Identifying and Accommodating Indiv Diffs,Assay which user characteristics predict d

8、ifferences in performance;many studies stop hereIsolate isolate the source of variation to a specific sub-task or design componentAccommodate do something about itOften harder than you think E.g.,Spatial ability and hierarchy navigation E.g.,ExpertiseEvaluate!,June 18,2001,Delos-NSF Workshop,Greene

9、et al.No IFs,ANDs,or ORs:A Study of Database Querying,Task:Find all employees who either work in the toy department or are managed by Grant,and also come from the city London.SQL fixed syntax,logical operators,parenthesesE.g.,SELECT NameFROMEmployeeWHERE(Department=ToyORManager=Grant)ANDCity=LondonT

10、EBI just need attribute names and values;recognize alternatives from system-generated tableE.g.,Name,Department=Toy,Manager=Grant,City=London,June 18,2001,Delos-NSF Workshop,Greene et al.(Assay),Assessed individual characteristics:Age,spatial memory,reasoning,integrative processing,reading comprehen

11、sion&vocabularyFound large effects of:Integrative processing(on accuracy,for SQL interface)Age(on time,for SQL interface),June 18,2001,Delos-NSF Workshop,Greene et al.,June 18,2001,Delos-NSF Workshop,Greene et al.,June 18,2001,Delos-NSF Workshop,Greene et al.(Isolate),Examined two possible sources o

12、f difficultiesInterpreting the querySpecifying the query in a formal notation or query language,June 18,2001,Delos-NSF Workshop,Example TEBI Table,June 18,2001,Delos-NSF Workshop,Greene et al.,June 18,2001,Delos-NSF Workshop,Greene et al.,June 18,2001,Delos-NSF Workshop,Greene et al.(Accommodate),SQ

13、L hard,especially for some usersTEBI new query specification languageImproved performance overallReduced many dependencies on reasoning skills and age“Robust interface”,June 18,2001,Delos-NSF Workshop,Dumais and Schmitt,June 18,2001,Delos-NSF Workshop,Dumais and Schmitt,June 18,2001,Delos-NSF Worksh

14、op,Dumais and Schmitt,June 18,2001,Delos-NSF Workshop,How to Accommodate?,Robust interfaces:A new design improves the performance for allE.g.,Greene et al.s TEBI interfaceE.g.,Dumais&Schmitts LikeThese interfaceTraining:Personalization:Different interfaces/systems for different peopleGroup level-E.g

15、.,Grundy prototypes,I3R sterotypes,Expert/NoviceIndividual levelTask(Info Need)level,June 18,2001,Delos-NSF Workshop,Personalization Framework,Characteristics for personalizationExpertise,Task,Preferences,Cog Aptitudes,Demographics,Cog Styles,Etc.Assay:How specified/modeled?Implicit,Explicit,Interac

16、tionStability over time?Long-term,short-termAccommodate:What to do about it?Many ways of accommodatingEvaluationBenefits of correct assessment and accommodationCosts of mis-assessment,June 18,2001,Delos-NSF Workshop,Content-Based Filtering,Match new content to standing info needAssay:Explicit or Imp

17、licit profile specification?Ongoing feedback?How rapidly does profile it change?Accommodate:Match profile against stream of new docsReduce number of docs to viewReturn more relevant docsBenefits/Costs,June 18,2001,Delos-NSF Workshop,June 18,2001,Delos-NSF Workshop,ASI Examples,Collaborative Filterin

18、gImplicit/Background QueryLumiereTemporal Query Patterns,June 18,2001,Delos-NSF Workshop,Collaborative filtering algorithmsBayesian networkCorrelation+Vector similarityBayesian clusteringPopularityTest collectionsEach MovieNielsenM,Example:MSRweb Recommender,Predicted Individual scores Ranked score,

19、June 18,2001,Delos-NSF Workshop,Example:MSRweb Recommender,June 18,2001,Delos-NSF Workshop,Identify content at users focus of attention Formulate query,provide related information as part of normal work flow Background,implicit queries,Example:Background Query,June 18,2001,Delos-NSF Workshop,Data Mo

20、untain with Implicit Query results(highlighted pages to left of selected page),June 18,2001,Delos-NSF Workshop,Implicit Query Results,Filing strategiesNumber of categories,June 18,2001,Delos-NSF Workshop,Implicit Query Results,June 18,2001,Delos-NSF Workshop,17 subjects(9 IQ1,8 IQ1&2),Implicit Query

21、 Results(Delayed Retrieval,6 months),June 18,2001,Delos-NSF Workshop,Example:Lumiere,Inferring users goals under uncertainty,June 18,2001,Delos-NSF Workshop,Example:Lumiere,June 18,2001,Delos-NSF Workshop,Example:Lumiere,June 18,2001,Delos-NSF Workshop,Example:Web Queries,161858lion lions 163041lion

22、 facts 163919picher of lions164040lion picher 165002lion pictures165100 pictures of lions165211 pictures of big cats165311lion photos 170013video in lion 172131pictureof a lioness172207picture of a lioness172241lion pictures 172334lion pictures cat 172443lions172450lions,150052lion152004lions152036l

23、ions lion 152219lion facts153747roaring153848lions roaring160232africa lion160642lions,tigers,leopards and cheetahs161042lions,tigers,leopards and cheetahs cats 161144wild cats of africa 161414africa cat161602africa lions161308africa wild cats161823 mane161840lion,user=A1D6F19DB06BD694date=970916exc

24、ite log,June 18,2001,Delos-NSF Workshop,June 18,2001,Delos-NSF Workshop,Query Dynamics&User Goals,Queries are not independentConsider:Search goals(e.g.,current events,weather)Refinement actions(e.g.,specialize,new)Temporal dynamicsBayes net to predict next action,or next search goalHand-tagged sampl

25、e of Excite log,June 18,2001,Delos-NSF Workshop,Temporal dynamics results,June 18,2001,Delos-NSF Workshop,June 18,2001,Delos-NSF Workshop,Potential Applications,Calculate probability of next search action based on time since last querySuggest appropriate queriesInvoke targeted help for this type of

26、queryPredict informational goal of userRequires knowledge of refinement classEnhance searchHighlight links related to users goalPerform targeted advertising,June 18,2001,Delos-NSF Workshop,Real-World Examples,Implicit storage of history of interactionCachingHistoryAuto CompletionDynamic MenusExplici

27、t storageFavoritesMySearch,iLORRecommendationsMyBlah,June 18,2001,Delos-NSF Workshop,June 18,2001,Delos-NSF Workshop,June 18,2001,Delos-NSF Workshop,June 18,2001,Delos-NSF Workshop,June 18,2001,Delos-NSF Workshop,June 18,2001,Delos-NSF Workshop,June 18,2001,Delos-NSF Workshop,June 18,2001,Delos-NSF

28、Workshop,June 18,2001,Delos-NSF Workshop,Personalization Success,Effectively Assay and Accommodate:Easy to specify relevant informationExplicitly:profile changes slowlyImplicitly:capture automatically,esp short timeWe know what to do about itAlgorithmic and application levelsAnd,the user can see the

29、 benefitAnd,there are few big failures,June 18,2001,Delos-NSF Workshop,Personalization Opportunities,Geo-codingQuery historyQuery plus usage contextKeeping found things found,June 18,2001,Delos-NSF Workshop,Open Issues,Evaluation difficult for personalized systemsComponents,easierEnd-to-end applications,harderQuestionnairesPre-Post assessmentAlgorithmic issues in situPrivacy,security,June 18,2001,Delos-NSF Workshop,The End,