1. 2002.11.07 - SLIDE 1IS 202 – FALL 2002 Lecture 20: Lexical Relations & WordNet Prof. Ray Larson & Prof. Marc Davis UC Berkeley SIMS Tuesday and Thursday 10:30 am - 12:00 pm Fall 2002 http://www.sims.berkeley.edu/academics/courses/is202/f02/ SIMS 202: Information Organization and Retrieval

2. 2002.11.07 - SLIDE 2IS 202 – FALL 2002 Lecture Overview Review –Probabilistic Models of IR –Relevance Feedback Lexical Relations WordNet Can Lexical and Semantic Relations be Exploited to Improve IR? Credit for some of the slides in this lecture goes to Marti Hearst and Warren Sack

3. 2002.11.07 - SLIDE 3IS 202 – FALL 2002 Lecture Overview Review –Probabilistic Models of IR –Relevance Feedback Lexical Relations WordNet Can Lexical and Semantic Relations be Exploited to Improve IR? Credit for some of the slides in this lecture goes to Marti Hearst and Warren Sack

4. 2002.11.07 - SLIDE 4IS 202 – FALL 2002 Probability Ranking Principle If a reference retrieval system’s response to each request is a ranking of the documents in the collections in the order of decreasing probability of usefulness to the user who submitted the request, where the probabilities are estimated as accurately as possible on the basis of whatever data has been made available to the system for this purpose, then the overall effectiveness of the system to its users will be the best that is obtainable on the basis of that data. Stephen E. Robertson, J. Documentation 1977

5. 2002.11.07 - SLIDE 5IS 202 – FALL 2002 Probabilistic Models: Some Unifying Notation D = All present and future documents Q = All present and future queries (D i,Q j ) = A document query pair x = class of similar documents, y = class of similar queries, Relevance (R) is a relation:

6. 2002.11.07 - SLIDE 6IS 202 – FALL 2002 Probabilistic Models Model 1 -- Probabilistic Indexing, P(R|y,D i ) Model 2 -- Probabilistic Querying, P(R|Q j,x) Model 3 -- Merged Model, P(R| Q j, D i ) Model 0 -- P(R|y,x) Probabilities are estimated based on prior usage or relevance estimation

7. 2002.11.07 - SLIDE 7IS 202 – FALL 2002 Probabilistic Models Q D x y DiDi QjQj

8. 2002.11.07 - SLIDE 8IS 202 – FALL 2002 Logistic Regression Another approach to estimating probability of relevance Based on work by William Cooper, Fred Gey and Daniel Dabney Builds a regression model for relevance prediction based on a set of training data Uses less restrictive independence assumptions than Model 2 –Linked Dependence

9. 2002.11.07 - SLIDE 9IS 202 – FALL 2002 Logistic Regression 100 - 90 - 80 - 70 - 60 - 50 - 40 - 30 - 20 - 10 - 0 - 0 10 20 30 40 50 60 Term Frequency in Document Relevance

10. 2002.11.07 - SLIDE 10IS 202 – FALL 2002 Logistic Regression Probability of relevance is based on Logistic regression from a sample set of documents to determine values of the coefficients At retrieval the probability estimate is obtained by: For the 6 X attribute measures shown previously

11. 2002.11.07 - SLIDE 11IS 202 – FALL 2002 Relevance Feedback in an IR System Interest profiles & Queries Documents & data Rules of the game = Rules for subject indexing + Thesaurus (which consists of Lead-In Vocabulary and Indexing Language Storage Line Potentially Relevant Documents Comparison/ Matching Store1: Profiles/ Search requests Store2: Document representations Indexing (Descriptive and Subject) Formulating query in terms of descriptors Storage of profiles Storage of Documents Information Storage and Retrieval System Selected relevant docs

12. 2002.11.07 - SLIDE 12IS 202 – FALL 2002 Relevance Feedback Main Idea: –Modify existing query based on relevance judgements Extract terms from relevant documents and add them to the query And/or re-weight the terms already in the query –Two main approaches: Automatic (pseudo-relevance feedback) Users select relevant documents –Users/system select terms from an automatically-generated list

13. 2002.11.07 - SLIDE 13IS 202 – FALL 2002 Rocchio/Vector Illustration Retrieval Information 0.5 1.0 0 0.51.0 D1D1 D2D2 Q0Q0 Q’ Q” Q 0 = retrieval of information = (0.7,0.3) D 1 = information science = (0.2,0.8) D 2 = retrieval systems = (0.9,0.1) Q’ = ½*Q 0 + ½ * D 1 = (0.45,0.55) Q” = ½*Q 0 + ½ * D 2 = (0.80,0.20)

14. 2002.11.07 - SLIDE 14IS 202 – FALL 2002 Alternative Notions of Relevance Feedback Find people whose taste is “similar” to yours –Will you like what they like? Follow a users’ actions in the background –Can this be used to predict what the user will want to see next? Track what lots of people are doing –Does this implicitly indicate what they think is good and not good?

15. 2002.11.07 - SLIDE 15IS 202 – FALL 2002 Alternative Notions of Relevance Feedback Several different criteria to consider: –Implicit vs. Explicit judgements –Individual vs. Group judgements –Standing vs. Dynamic topics –Similarity of the items being judged vs. similarity of the judges themselves

16. 2002.11.07 - SLIDE 16IS 202 – FALL 2002 Lecture Overview Review –Probabilistic Models of IR –Relevance Feedback Lexical Relations WordNet Can Lexical and Semantic Relations be Exploited to Improve IR? Credit for some of the slides in this lecture goes to Marti Hearst and Warren Sack

17. 2002.11.07 - SLIDE 17IS 202 – FALL 2002 Syntax The syntax of a language is to be understood as a set of rules which accounts for the distribution of word forms throughout the sentences of a language These rules codify permissible combinations of classes of word forms

18. 2002.11.07 - SLIDE 18IS 202 – FALL 2002 Semantics Semantics is the study of linguistic meaning Two standard approaches to lexical semantics (cf., sentential semantics; and, logical semantics): –(1) compositional –(2) relational

19. 2002.11.07 - SLIDE 19IS 202 – FALL 2002 Lexical Semantics: Compositional Approach Compositional lexical semantics, introduced by Katz & Fodor (1963), analyzes the meaning of a word in much the same way a sentence is analyzed into semantic components. The semantic components of a word are not themselves considered to be words, but are abstract elements (semantic atoms) postulated in order to describe word meanings (semantic molecules) and to explain the semantic relations between words. For example, the representation of bachelor might be ANIMATE and HUMAN and MALE and ADULT and NEVER MARRIED. The representation of man might be ANIMATE and HUMAN and MALE and ADULT; because all the semantic components of man are included in the semantic components of bachelor, it can be inferred that bachelor  man. In addition, there are implicational rules between semantic components, e.g. HUMAN  ANIMATE, which also look very much like meaning postulates. –George Miller, “On Knowing a Word,” 1999

20. 2002.11.07 - SLIDE 20IS 202 – FALL 2002 Lexical Semantics: Relational Approach Relational lexical semantics was first introduced by Carnap (1956) in the form of meaning postulates, where each postulate stated a semantic relation between words. A meaning postulate might look something like dog  animal (if x is a dog then x is an animal) or, adding logical constants, bachelor  man and never married [if x is a bachelor then x is a man and not(x has married)] or tall  not short [if x is tall then not(x is short)]. The meaning of a word was given, roughly, by the set of all meaning postulates in which it occurs. –George Miller, “On Knowing a Word,” 1999

21. 2002.11.07 - SLIDE 21IS 202 – FALL 2002 Pragmatics Deals with the relation between signs or linguistic expressions and their users Deixis (literally “pointing out”) –E.g., “I’ll be back in an hour” depends upon the time of the utterance Conversational implicature –A: “Can you tell me the time?” –B: “Well, the milkman has come.” [I don’t know exactly, but perhaps you can deduce it from some extra information I give you.] Presupposition –“Are you still such a bad driver?” Speech acts –Constatives vs. performatives –E.g., “I second the motion.” Conversational structure –E.g., turn-taking rules

22. 2002.11.07 - SLIDE 22IS 202 – FALL 2002 Language Language only hints at meaning Most meaning of text lies within our minds and common understanding –“How much is that doggy in the window?” How much: social system of barter and trade (not the size of the dog) “doggy” implies childlike, plaintive, probably cannot do the purchasing on their own “in the window” implies behind a store window, not really inside a window, requires notion of window shopping

23. 2002.11.07 - SLIDE 23IS 202 – FALL 2002 Semantics: The Meaning of Symbols Semantics versus Syntax –add(3,4) –3 + 4 –(different syntax, same meaning) Meaning versus Representation –What a person’s name is versus who they are A rose by any other name... –What the computer program “looks like” versus what it actually does

24. 2002.11.07 - SLIDE 24IS 202 – FALL 2002 Semantics Semantics: Assigning meanings to symbols and expressions –Usually involves defining: Objects Properties of objects Relations between objects –More detailed versions include Events Time Places Measurements (quantities)

25. 2002.11.07 - SLIDE 25IS 202 – FALL 2002 The Role of Context The concept associated with the symbol “21” means different things in different contexts –Examples? The question “Is there any salt?” –Asked of a waiter at a restaurant –Asked of an environmental scientist at work

26. 2002.11.07 - SLIDE 26IS 202 – FALL 2002 What’s In a Sentence? “A sentence is not a verbal snapshot or movie of an event. In framing an utterance, you have to abstract away from everything you know, or can picture, about a situation, and present a schematic version which conveys the essentials. In terms of grammatical marking, there is not enough time in the speech situation for any language to allow for the marking of everything which could possibly be significant to the message.” Dan Slobin, in Language Acquisition: The state of the art, 1982

27. 2002.11.07 - SLIDE 27IS 202 – FALL 2002 Lexical Relations Conceptual relations link concepts –Goal of Artificial Intelligence Lexical relations link words –Goal of Linguistics

28. 2002.11.07 - SLIDE 28IS 202 – FALL 2002 Major Lexical Relations Synonymy Polysemy Metonymy Hyponymy/Hyperonymy Meronymy Antonymy

29. 2002.11.07 - SLIDE 29IS 202 – FALL 2002 Synonymy Different ways of expressing related concepts Examples –cat, feline, Siamese cat Overlaps with basic and subordinate levels Synonyms are almost never truly substitutable: –Used in different contexts –Have different implications This is a point of contention

30. 2002.11.07 - SLIDE 30IS 202 – FALL 2002 Polysemy Most words have more than one sense –Homonym: same word, different meaning bank (river) bank (financial) –Polysemy: different senses of same word That dog has floppy ears. She has a good ear for jazz. bank (financial) has several related senses –the building, the institution, the notion of where money is stored

31. 2002.11.07 - SLIDE 31IS 202 – FALL 2002 Metonymy Use one aspect of something to stand for the whole –The building stands for the institution of the bank. –Newscast: “The White House released new figures today.” –Waitperson: “The ham sandwich spilled his drink.”

32. 2002.11.07 - SLIDE 32IS 202 – FALL 2002 Hyponymy/Hyperonymy ISA relation Related to Superordinate and Subordinate level categories –hyponym(robin,bird) –hyponym(bird,animal) –hyponym(emu,bird) A is a hypernym of B if B is a type of A A is a hyponym of B if A is a type of B

33. 2002.11.07 - SLIDE 33IS 202 – FALL 2002 Basic-Level Categories (review) Brown 1958, 1965, Berlin et al., 1972, 1973 Folk biology: –Unique beginner: plant, animal –Life form: tree, bush, flower –Generic name: pine, oak, maple, elm –Specific name: Ponderosa pine, white pine –Varietal name: Western Ponderosa pine No overlap between levels Level 3 is basic –Corresponds to genus –Folk biological categories correspond accurately to scientific biological categories only at the basic level

34. 2002.11.07 - SLIDE 34IS 202 – FALL 2002 Psychologically Primary Levels SUPERORDINATE animal furniture BASIC LEVEL dog chair SUBORDINATE terrier rocker Children take longer to learn superordinate Superordinate not associated with mental images or motor actions

35. 2002.11.07 - SLIDE 35IS 202 – FALL 2002 Meronymy Parts-of relation –part of(beak, bird) –part of(bark, tree) Transitive conceptually but not lexically: –The knob is a part of the door. –The door is a part of the house. –? The knob is a part of the house ?

36. 2002.11.07 - SLIDE 36IS 202 – FALL 2002 Antonymy Lexical opposites –antonym(large, small) –antonym(big, small) –antonym(big, little) –but not large, little Many antonymous relations can be reliably detected by looking for statistical correlations in large text collections. (Justeson &Katz 91)

37. 2002.11.07 - SLIDE 37IS 202 – FALL 2002 Thesauri and Lexical Relations Polysemy: Same word, different senses of meaning –Slightly different concepts expressed similarly Synonyms: Different words, related senses of meanings –Different ways to express similar concepts Thesauri help draw all these together Thesauri also commonly define a set of relations between terms that is similar to lexical relations –BT, NT, RT

38. 2002.11.07 - SLIDE 38IS 202 – FALL 2002 What is an Ontology? From Merriam-Webster’s Collegiate: –A branch of metaphysics concerned with the nature and relations of being –A particular theory about the nature of being or the kinds of existence More prosaically: –A carving up of the world’s meanings –Determine what things exist, but not how they inter- relate Related terms: –Taxonomy, dictionary, category structure Commonly used now in CS literature to describe structures that function as Thesauri

39. 2002.11.07 - SLIDE 39IS 202 – FALL 2002 Lecture Overview Review –Probabilistic Models of IR –Relevance Feedback Lexical Relations WordNet Can Lexical and Semantic Relations be Exploited to Improve IR? Credit for some of the slides in this lecture goes to Marti Hearst and Warren Sack

40. 2002.11.07 - SLIDE 40IS 202 – FALL 2002 WordNet Started in 1985 by George Miller, students, and colleagues at the Cognitive Science Laboratory, Princeton University Can be downloaded for free: –www.cogsci.princeton.edu/~wn/ “In terms of coverage, WordNet’s goals differ little from those of a good standard college-level dictionary, and the semantics of WordNet is based on the notion of word sense that lexicographers have traditionally used in writing dictionaries. It is in the organization of that information that WordNet aspires to innovation.” –(Miller, 1998, Chapter 1)

41. 2002.11.07 - SLIDE 41IS 202 – FALL 2002 Presuppositions of WordNet project Separability hypothesis: T –The lexical component of language can be separated and studied in its own right Patterning hypothesis: –People have knowledge of the systematic patterns and relations between word meanings Comprehensiveness hypothesis: –Computational linguistics programs need a store of lexical knowledge that is as extensive as that which people have

42. 2002.11.07 - SLIDE 42IS 202 – FALL 2002 WordNet: Size POSUniqueSynsets Strings Noun 107930 74488 Verb 10806 12754 Adjective 21365 18523 Adverb 4583 3612 Totals144684 109377 WordNet Uses “Synsets” – sets of synonymous terms

43. 2002.11.07 - SLIDE 43IS 202 – FALL 2002 Structure of WordNet

44. 2002.11.07 - SLIDE 44IS 202 – FALL 2002 Structure of WordNet

45. 2002.11.07 - SLIDE 45IS 202 – FALL 2002 Structure of WordNet

46. 2002.11.07 - SLIDE 46IS 202 – FALL 2002 Unique Beginners Entity, something –(anything having existence (living or nonliving)) Psychological_feature –(a feature of the mental life of a living organism) Abstraction –(a general concept formed by extracting common features from specific examples) State –(the way something is with respect to its main attributes; "the current state of knowledge"; "his state of health"; "in a weak financial state") Event –(something that happens at a given place and time)

47. 2002.11.07 - SLIDE 47IS 202 – FALL 2002 Unique Beginners Act, human_action, human_activity –(something that people do or cause to happen) Group, grouping –(any number of entities (members) considered as a unit) Possession –(anything owned or possessed) Phenomenon –(any state or process known through the senses rather than by intuition or reasoning)

48. 2002.11.07 - SLIDE 48IS 202 – FALL 2002 WordNet Usage Available online (from Unix) if you wish to try it… –Login to irony and type “wn word” for any word you are interested in –Demo…

49. 2002.11.07 - SLIDE 49IS 202 – FALL 2002 Lecture Overview Review –Probabilistic Models of IR –Relevance Feedback Lexical Relations WordNet Can Lexical and Semantic Relations be Exploited to Improve IR? Credit for some of the slides in this lecture goes to Marti Hearst and Warren Sack

50. 2002.11.07 - SLIDE 50IS 202 – FALL 2002 Lexical Relations and IR Recall that most IR research has primarily looked at statistical approaches to inferring the topicality or meaning of documents I.e., Statistics imply Semantics –Is this really true or correct? How has (or might) WordNet be used to provide more functionality in searching? What about other thesauri, classification schemes and ontologies?

51. 2002.11.07 - SLIDE 51IS 202 – FALL 2002 Natural Language Processing and IR The main approach in applying NLP to IR has been to attempt to address –Phrase usage vs individual terms –Search expansion using related terms/concepts –Attempts to automatically exploit or assign controlled vocabularies

52. 2002.11.07 - SLIDE 52IS 202 – FALL 2002 NLP and IR Early research showed that (at least in the restricted test databases tested) –Indexing documents by individual terms corresponding to words and word stems produces retrieval results at least as good as when indexes use controlled vocabularies (whether applied manually or automatically) –Constructing phrases or “pre-coordinated” terms provides only marginal and inconsistent improvements

53. 2002.11.07 - SLIDE 53IS 202 – FALL 2002 NLP and IR Not clear why intuitively plausible improvements to document representation have had little effect on retrieval results when compared to statistical methods –E.g. Use of syntactic role relations between terms has shown no improvement in performance over “bag of words” approaches –Semantics is even harder to accomplish WordNet alone can’t disambiguate word senses in texts

54. 2002.11.07 - SLIDE 54IS 202 – FALL 2002 Using NLP Strzalkowski TextNLPrepres Dbase search TAGGER NLP: PARSERTERMS

55. 2002.11.07 - SLIDE 55IS 202 – FALL 2002 Using NLP INPUT SENTENCE The former Soviet President has been a local hero ever since a Russian tank invaded Wisconsin. TAGGED SENTENCE The/dt former/jj Soviet/jj President/nn has/vbz been/vbn a/dt local/jj hero/nn ever/rb since/in a/dt Russian/jj tank/nn invaded/vbd Wisconsin/np./per

56. 2002.11.07 - SLIDE 56IS 202 – FALL 2002 Using NLP TAGGED & STEMMED SENTENCE the/dt former/jj soviet/jj president/nn have/vbz be/vbn a/dt local/jj hero/nn ever/rb since/in a/dt russian/jj tank/nn invade/vbd wisconsin/np./per

57. 2002.11.07 - SLIDE 57IS 202 – FALL 2002 Using NLP PARSED SENTENCE [assert [[perf [have]][[verb[BE]] [subject [np[n PRESIDENT][t_pos THE] [adj[FORMER]][adj[SOVIET]]]] [adv EVER] [sub_ord[SINCE [[verb[INVADE]] [subject [np [n TANK][t_pos A] [adj [RUSSIAN]]]] [object [np [name [WISCONSIN]]]]]]]]]

58. 2002.11.07 - SLIDE 58IS 202 – FALL 2002 Using NLP EXTRACTED TERMS & WEIGHTS President 2.623519 soviet 5.416102 President+soviet 11.556747 president+former 14.594883 Hero 7.896426 hero+local 14.314775 Invade 8.435012 tank 6.848128 Tank+invade 17.402237 tank+russian 16.030809 Russian 7.383342 wisconsin 7.785689

59. 2002.11.07 - SLIDE 59IS 202 – FALL 2002 NLP & IR Research Issues Is natural language indexing using more NLP knowledge needed? Or, should controlled vocabularies be used Can NLP in its current state provide the improvements needed How to test

60. 2002.11.07 - SLIDE 60IS 202 – FALL 2002 NLP & IR Research Areas Lewis and Sparck Jones (CACM 1996) suggest research in three areas –Examination of the words, phrases and sentences that make up a document description and express the combinatory, syntagmatic relations between single terms –The classificatory structure over document collection as a whole, indicating the paradigmatic relations between terms and permitting controlled vocabulary indexing and searching –Using NLP-based methods for searching and matching

61. 2002.11.07 - SLIDE 61IS 202 – FALL 2002 NLP & IR: Possible Approaches Indexing –Use of NLP methods to identify phrases Test weighting schemes for phrases –Use of more sophisticated morphological analysis Searching –Use of two-stage retrieval Statistical retrieval Followed by more sophisticated NLP filtering

62. 2002.11.07 - SLIDE 62IS 202 – FALL 2002 Can Statistics approach Semantics? One approach is the Entry Vocabulary Index (EVI) work being done here… (The following slides are from my presentation at JCDL 2002)

63. 2002.11.07 - SLIDE 63IS 202 – FALL 2002 What is an Entry Vocabulary Index? EVIs are a means of mapping from user’s vocabulary to the controlled vocabulary of a collection of documents…

64. 2002.11.07 - SLIDE 64IS 202 – FALL 2002 Start with a collection of documents.

65. 2002.11.07 - SLIDE 65IS 202 – FALL 2002 Classify and index with controlled vocabulary. Index Ideally, use a database already indexed

66. 2002.11.07 - SLIDE 66IS 202 – FALL 2002 Problem: Controlled Vocabularies can be difficult for people to use. “pass mtr veh spark ign eng” Index

67. 2002.11.07 - SLIDE 67IS 202 – FALL 2002 Solution: Entry Level Vocabulary Indexes. Index EVI pass mtr veh spark ign eng” = “Automobile”

68. 2002.11.07 - SLIDE 68IS 202 – FALL 2002 EVI example EVI 1 Index term: “pass mtr veh spark ign eng” User Query “Automobile” EVI 2 Index term: “automobiles” OR “internal combustible engines”

69. 2002.11.07 - SLIDE 69IS 202 – FALL 2002 But why stop there? Index EVI

70. 2002.11.07 - SLIDE 70IS 202 – FALL 2002 “Which EVI do I use?” Index EVI Index EVI Index EVI

71. 2002.11.07 - SLIDE 71IS 202 – FALL 2002 EVI to EVIs Index EVI Index EVI Index EVI EVI 2

72. 2002.11.07 - SLIDE 72IS 202 – FALL 2002 Find Plutonium In Arabic Chinese Greek Japanese Korean Russian Tamil Why not treat language the same way?

73. 2002.11.07 - SLIDE 73IS 202 – FALL 2002 Find Plutonium In Arabic Chinese Greek Japanese Korean Russian Tamil Statistical association Digital library resources