龙星计划课程信息检索CourseOverviewBackground.ppt

《龙星计划课程信息检索CourseOverviewBackground.ppt》由会员分享，可在线阅读，更多相关《龙星计划课程信息检索CourseOverviewBackground.ppt（65页珍藏版）》请在三一办公上搜索。

1、2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,1,龙星计划课程:信息检索 Course Overview&Background,ChengXiang Zhai(翟成祥)Department of Computer ScienceGraduate School of Library&Information ScienceInstitute for Genomic Biology,StatisticsUniversity of Illinois,Urbana-Champaignhttp:

2、/www-faculty.cs.uiuc.edu/czhai,czhaics.uiuc.edu,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,2,Outline,Course overviewEssential backgroundProbability&statisticsBasic concepts in information theoryNatural language processing,2008 ChengXiang Zhai Dragon Star Lecture a

3、t Beijing University,June 21-30,2008,3,Course Overview,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,4,Course Objectives,Introduce the field of information retrieval(IR)Foundation:Basic concepts,principles,methods,etcTrends:Frontier topics Prepare students to do rese

4、arch in IR and/or related fieldsResearch methodology(general and IR-specific)Research proposal writingResearch project(to be finished after the lecture period),2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,5,Prerequisites,Proficiency in programming(C+is needed for as

5、signments)Knowledge of basic probability&statistics(would be necessary for understanding algorithms deeply)Big plus:knowledge of related areasMachine learningNatural language processingData mining,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,6,Course Management,Teac

6、hing staffInstructor:ChengXiang Zhai(UIUC)Teaching assistants:Hongfei Yan(Peking Univ)Bo Peng(Peking Univ)Course website:http:/group discussion:http:/First post the questions on the group discussion forum;if questions are unanswered,bring them to the office hours(first office hour:June 23,2:30-4:30p

7、m),2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,7,Format&Requirements,Lecture-based:Morning lectures:Foundation&TrendsAfternoon lectures:IR research methodologyReadings are usually available online 2 Assignments(based on morning lectures)Coding(C+),experimenting wit

8、h data,analyzing results,open explorations(5 hours each)Final exam(based on morning lectures):1:30-4:30pm,June 30.Practice questions will be available,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,8,Format&Requirements(cont.),Course project(Mini-TREC)Work in teamsPha

9、se I:create test collections(3 hours,done within lecture period)Phase II:develop algorithms and submit results(done in the summer)Research project proposal(based on afternoon lectures)Work in teams2-page outline done within lecture periodfull proposal(5 pages)due later,2008 ChengXiang Zhai Dragon St

10、ar Lecture at Beijing University,June 21-30,2008,9,Coverage of Topics:IR vs.TIM,Text Information Management(TIM),Information Retrieval(IR),Multimedia,etc,IR and TIM will be used interchangeably,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,10,What is Text Info.Manage

11、ment?,TIM is concerned with technologies for managing and exploiting text information effectively and efficientlyImportance of managing text informationThe most natural way of encoding knowledgeThink about scientific literatureThe most common type of informationHow much textual information do you pr

12、oduce and consume every day?The most basic form of informationIt can be used to describe other media of informationThe most useful form of information!,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,11,Text Management Applications,Access,Mining,Organization,Select inf

13、ormation,Create Knowledge,Add Structure/Annotations,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,12,Examples of Text Management Applications,SearchWeb search engines(Google,Yahoo,)Library systemsRecommendationNews filterLiterature/movie recommenderCategorizationAuto

14、matically sorting emailsMining/ExtractionDiscovering major complaints from email in customer serviceBusiness intelligenceBioinformaticsMany others,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,13,Elements of Text Info Management Technologies,Search,Text,Filtering,Cat

15、egorization,Summarization,Clustering,Natural Language Content Analysis,Extraction,Mining,Visualization,RetrievalApplications,MiningApplications,InformationAccess,KnowledgeAcquisition,InformationOrganization,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,14,Text Manage

16、ment and Other Areas,TM Algorithms,User,Text,StorageCompression,Probabilistic inferenceMachine learning,Natural language processing,Human-computer interaction,TM Applications,Software engineeringWeb,Computer science,InformationScience,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,Ju

17、ne 21-30,2008,15,Related Areas,InformationRetrieval,Databases,Library&InfoScience,Machine LearningPattern RecognitionData Mining,NaturalLanguageProcessing,ApplicationsWeb,Bioinformatics,StatisticsOptimization,Software engineeringComputer systems,Models,Algorithms,Applications,Systems,2008 ChengXiang

18、 Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,16,Publications/Societies(Incomplete),ACM SIGIR,VLDB,PODS,ICDE,ASIS,Learning/Mining,NLP,Applications,Statistics,Software/systems,COLING,EMNLP,ANLP,HLT,ICML,NIPS,UAI,RECOMB,PSB,JCDL,Info.Science,Info Retrieval,ACM CIKM,Databases,ACM SIGM

19、OD,ACL,ICML,AAAI,ACM SIGKDD,ISMB,WWW,SOSP,OSDI,TREC,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,17,Schedule:available at http:/,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,18,2008 ChengXiang Zhai Dragon Star Lecture at Beijing Uni

20、versity,June 21-30,2008,19,Essential Backgroud 1:Probability&Statistics,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,20,Prob/Statistics&Text Management,Probability&statistics provide a principled way to quantify the uncertainties associated with natural languageAllo

21、w us to answer questions like:Given that we observe“baseball”three times and“game”once in a news article,how likely is it about“sports”?(text categorization,information retrieval)Given that a user is interested in sports news,how likely would the user use“baseball”in a query?(information retrieval),

22、2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,21,Basic Concepts in Probability,Random experiment:an experiment with uncertain outcome(e.g.,tossing a coin,picking a word from text)Sample space:all possible outcomes,e.g.,Tossing 2 fair coins,S=HH,HT,TH,TTEvent:ES,E hap

23、pens iff outcome is in E,e.g.,E=HH(all heads)E=HH,TT(same face)Impossible event(),certain event(S)Probability of Event:1P(E)0,s.t.P(S)=1(outcome always in S)P(A B)=P(A)+P(B)if(AB)=(e.g.,A=same face,B=different face),2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,22,Ba

24、sic Concepts of Prob.(cont.),Conditional Probability:P(B|A)=P(AB)/P(A)P(AB)=P(A)P(B|A)=P(B)P(A|B)So,P(A|B)=P(B|A)P(A)/P(B)(Bayes Rule)For independent events,P(AB)=P(A)P(B),so P(A|B)=P(A)Total probability:If A1,An form a partition of S,thenP(B)=P(BS)=P(BA1)+P(B An)(why?)So,P(Ai|B)=P(B|Ai)P(Ai)/P(B)=P

25、(B|Ai)P(Ai)/P(B|A1)P(A1)+P(B|An)P(An)This allows us to compute P(Ai|B)based on P(B|Ai),2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,23,Interpretation of Bayes Rule,Hypothesis space:H=H1,HnEvidence:E,If we want to pick the most likely hypothesis H*,we can drop P(E),P

26、osterior probability of Hi,Prior probability of Hi,Likelihood of data/evidenceif Hi is true,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,24,Random Variable,X:S(“measure”of outcome)E.g.,number of heads,all same face?,Events can be defined according to XE(X=a)=si|X(si

27、)=aE(Xa)=si|X(si)aSo,probabilities can be defined on XP(X=a)=P(E(X=a)P(aX)=P(E(aX)Discrete vs.continuous random variable(think of“partitioning the sample space”),2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,25,An Example:Doc Classification,X1:sport 1 0 1 1,Topic the

28、 computer game baseball,X2:sport 1 1 1 1,X3:computer 1 1 0 0,X4:computer 1 1 1 0,X5:other 0 0 1 1,For 3 topics,four words,n=?,EventsEsport=xi|topic(xi)=“sport”Ebaseball=xi|baseball(xi)=1Ebaseball,computer=xi|baseball(xi)=1&computer(xi)=0,Sample Space S=x1,xn,Conditional Probabilities:P(Esport|Ebaseb

29、all),P(Ebaseball|Esport),P(Esport|Ebaseball,computer),.,An inference problem:Suppose we observe that“baseball”is mentioned,how likely the topic is about“sport”?,But,P(B=1|T=“sport”)=?,P(T=“sport”)=?,P(T=“sport”|B=1)P(B=1|T=“sport”)P(T=“sport”),Thinking in terms of random variablesTopic:T“sport”,“com

30、puter”,“other”,“Baseball”:B 0,1,P(T=“sport”|B=1),P(B=1|T=“sport”),.,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,26,Getting to Statistics.,P(B=1|T=“sport”)=?(parameter estimation)If we see the results of a huge number of random experiments,then But,what if we only s

31、ee a small sample(e.g.,2)?Is this estimate still reliable?In general,statistics has to do with drawing conclusions on the whole population based on observations of a sample(data),2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,27,Parameter Estimation,General setting:Gi

32、ven a(hypothesized&probabilistic)model that governs the random experimentThe model gives a probability of any data p(D|)that depends on the parameter Now,given actual sample data X=x1,xn,what can we say about the value of?Intuitively,take your best guess of-“best”means“best explaining/fitting the da

33、ta”Generally an optimization problem,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,28,Maximum Likelihood vs.Bayesian,Maximum likelihood estimation“Best”means“data likelihood reaches maximum”Problem:small sampleBayesian estimation“Best”means being consistent with our“

34、prior”knowledge and explaining data wellProblem:how to define prior?,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,29,Illustration of Bayesian Estimation,Posterior:p(|X)p(X|)p(),2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,30,Maximum

35、 Likelihood Estimate,Data:a document d with counts c(w1),c(wN),and length|d|Model:multinomial distribution M with parameters p(wi)Likelihood:p(d|M)Maximum likelihood estimator:M=argmax M p(d|M),Well tune p(wi)to maximize l(d|M),Use Lagrange multiplier approach,Set partial derivatives to zero,ML esti

36、mate,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,31,What You Should Know,Probability concepts:sample space,event,random variable,conditional prob.multinomial distribution,etcBayes formula and its interpretationStatistics:Know how to compute maximum likelihood estim

37、ate,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,32,Essential Background 2:Basic Concepts in Information Theory,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,33,Information Theory,Developed by Shannon in the 40sMaximizing the amount

38、of information that can be transmitted over an imperfect communication channelData compression(entropy)Transmission rate(channel capacity),2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,34,Basic Concepts in Information Theory,Entropy:Measuring uncertainty of a random

39、variableKullback-Leibler divergence:comparing two distributionsMutual Information:measuring the correlation of two random variables,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,35,Entropy:Motivation,Feature selection:If we use only a few words to classify docs,what

40、kind of words should we use?P(Topic|“computer”=1)vs p(Topic|“the”=1):which is more random?Text compression:Some documents(less random)can be compressed more than others(more random)Can we quantify the“compressibility”?In general,given a random variable X following distribution p(X),How do we measure

41、 the“randomness”of X?How do we design optimal coding for X?,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,36,Entropy:Definition,Entropy H(X)measures the uncertainty/randomness of random variable X,Example:,P(Head),H(X),1.0,2008 ChengXiang Zhai Dragon Star Lecture at

42、Beijing University,June 21-30,2008,37,Entropy:Properties,Minimum value of H(X):0What kind of X has the minimum entropy?Maximum value of H(X):log M,where M is the number of possible values for XWhat kind of X has the maximum entropy?Related to coding,2008 ChengXiang Zhai Dragon Star Lecture at Beijin

43、g University,June 21-30,2008,38,Interpretations of H(X),Measures the“amount of information”in XThink of each value of X as a“message”Think of X as a random experiment(20 questions)Minimum average number of bits to compress values of XThe more random X is,the harder to compress,A fair coin has the ma

44、ximum information,and is hardest to compressA biased coin has some information,and can be compressed to 1 bit on averageA completely biased coin has no information,and needs only 0 bit,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,39,Conditional Entropy,The condition

45、al entropy of a random variable Y given another X,expresses how much extra information one still needs to supply on average to communicate Y given that the other party knows XH(Topic|“computer”)vs.H(Topic|“the”)?,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,40,Cross

46、 Entropy H(p,q),What if we encode X with a code optimized for a wrong distribution q?Expected#of bits=?,Intuitively,H(p,q)H(p),and mathematically,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,41,Kullback-Leibler Divergence D(p|q),What if we encode X with a code optim

47、ized for a wrong distribution q?How many bits would we waste?,Properties:-D(p|q)0-D(p|q)D(q|p)-D(p|q)=0 iff p=q,KL-divergence is often used to measure the distance between two distributions,Interpretation:Fix p,D(p|q)and H(p,q)vary in the same wayIf p is an empirical distribution,minimize D(p|q)or H

48、(p,q)is equivalent to maximizing likelihood,Relative entropy,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,42,Cross Entropy,KL-Div,and Likelihood,Likelihood:,log Likelihood:,Criterion for selecting a good model Perplexity(p),2008 ChengXiang Zhai Dragon Star Lecture a

49、t Beijing University,June 21-30,2008,43,Mutual Information I(X;Y),Comparing two distributions:p(x,y)vs p(x)p(y),Properties:I(X;Y)0;I(X;Y)=I(Y;X);I(X;Y)=0 iff X&Y are independent,Interpretations:-Measures how much reduction in uncertainty of X given info.about Y-Measures correlation between X and Y-R

50、elated to the“channel capacity”in information theory,Examples:I(Topic;“computer”)vs.I(Topic;“the”)?I(“computer”,“program”)vs(“computer”,“baseball”)?,2008 ChengXiang Zhai Dragon Star Lecture at Beijing University,June 21-30,2008,44,What You Should Know,Information theory concepts:entropy,cross entrop