Word Sense Disambiguation

1
Word Sense Disambiguation

• Evgenia Ivanova
• eivanova_at_sfs.uni-tuebingen.de
• January 8, 2007

2
Word Sense Disambiguation

• WordNet

3
Outline

• Basic Principles
• Statistics
• Nouns
• Adjectives
• Verbs

4
WordNet

• WordNet is an on-line lexical reference system
  whose design is inspired by current
  psycholinguistic theories of human lexical
  memory. English nouns, verbs and adjectives are
  organized into synonym sets, each representing
  one underlying lexical concept. Different
  relations link the synonym sets.

5
WordNet Basic Principles

• On-line free for download
• Based on psycholinguistic principles
• Division of the lexicon according to the
  syntactic categories
• Organization in terms of word meanings

6
WordNet - Statistics
7
WordNet - Nouns

• We might want to know
• What kinds of the concept there are?
• What are the coordinated terms concepts?
• Of what parts does the concept consists?

8
WordNet - Nouns

• Basic organization Lexical Inheritance System
• Hypernymy (_at_?) X is a kind of
• Hyponymy (?) is a kind of X

9
WordNet - Nouns

• The noun arm has 6 senses (first 5 from tagged
  texts)
• 1. (563) arm -- (a human limb technically the
  part of the superior limb between the shoulder
  and the elbow but commonly used to refer to the
  whole superior limb)
• 2. (168) weapon, arm, weapon system -- (any
  instrument or instrumentality used in fighting or
  hunting "he was licensed to carry a weapon")
• 3. (21) branch, subdivision, arm -- (an
  administrative division of some larger or more
  complex organization "a branch of Congress")
• 4. (16) arm, branch, limb -- (any projection that
  is thought to resemble a human arm "the arm of
  the record player" "an arm of the sea" "a
  branch of the sewer")
• 5. (5) arm -- (the part of an armchair or sofa
  that supports the elbow and forearm of a seated
  person)
• 6. sleeve, arm -- (the part of a garment that is
  attached at the armhole and that provides a cloth
  covering for the arm)

10
WordNet - Nouns

• Sense 2
• weapon, arm, weapon system -- (any instrument or
  instrumentality used in fighting or hunting "he
  was licensed to carry a weapon")
• gt instrument
• gt device
• gt instrumentality,
  instrumentation
• gt artifact, artefact
• gt whole, unit
• gt object, physical
  object
• gt physical entity
• gt entity

11
WordNet - Nouns

• Sense 2
• weapon, arm, weapon system -- (any instrument or
  instrumentality used in fighting or hunting "he
  was licensed to carry a weapon")
• gt bow
• gt gun
• gt knife
• gt light arm
• gt stun gun, stun baton
• gt sword, blade, brand, steel
• gt weapon of mass destruction, WMD, W.M.D.
  

12
WordNet - Nouns

• We might want to know
• What kinds of the concept there are?
• What are the coordinated terms concepts?
• Of what parts does the concept consists?

13
WordNet - Nouns

• Sense 1
• arm -- (a human limb technically the part of the
  superior limb between the shoulder and the elbow
  but commonly used to refer to the whole superior
  limb)
• -gt limb
• gt hind limb
• gt forelimb
• gt flipper
• gt leg
• gt thigh
• gt arm
• gt forearm

14
WordNet - Nouns

• We might want to know
• What kinds of the concept there are?
• What are the coordinated terms concepts?
• Of what parts does the concept consists?

15
WordNet - Nouns

• Meronymy
• If Y has X as a part ?
• x1, x2, is_a_myronym y1, y2,
• Transitive
• Asymmetrical

16
WordNet - Nouns

• Sense 1
• arm -- (a human limb technically the part of the
  superior limb between the shoulder and the elbow
  but commonly used to refer to the whole superior
  limb)
• HAS PART brachial artery, arteria brachialis
  
• HAS PART cephalic vein, vena cephalica
• HAS PART forearm
• HAS PART hand, manus, mitt, paw
• HAS PART ulnar nerve, cubital nerve, nervus
  ulnaris
• HAS PART biceps brachii, musculus biceps
  brachii
• HAS PART triceps brachii, musculus triceps
  brachii
• HAS PART elbow, elbow joint, human elbow,
  cubitus
• HAS PART wrist, carpus, wrist joint,
  radiocarpal joint
• HAS PART arm bone -- (a bone in the arm)

17
WordNet - Nouns

• 3 Types of myronymy
• Is a component of
• family, family unit
• HAS MEMBER child, kid
• Is a member of
• tree
• HAS PART trunk, tree trunk, bole
• Is the stuff made from
• tree
• HAS SUBSTANCE heartwood, duramen

18
WordNet - Adjectives

• Two basic types of adjectives
• Descriptive
• x is Adj
• A(x) Adj
• Relational
• pertaining to

19
WordNet - Adjectives

• Descriptive Adjectives
• Antonymy
• HEAVY LIGHT
• MASSIVE AIRY
• PONDEROUS FLOATY

20
WordNet - Adjectives

• Sense 1
• heavy (vs. light) -- (of comparatively great
  physical weight or density "a heavy load" "lead
  is a heavy metal" "heavy mahogany furniture")
• light (vs. heavy) -- (of comparatively little
  physical weight or density "a light load"
  "magnesium is a light metal--having a specific
  gravity of 1.74 at 20 degrees C")
• gt airy
• gt buoyant, floaty

21
WordNet - Adjectives

• Sense 3
• airy -- (having little or no perceptible weight
  so light as to resemble air "airy gauze
  curtains")
• INDIRECT (VIA light) -gt heavy -- (of
  comparatively great physical weight or density
  "a heavy load" "lead is a heavy metal" "heavy
  mahogany furniture")

22
WordNet - Adjectives

• Relational Adjectives
• Are mostly derived from nouns or from other
  relational adjectives
• Occur mostly in attributive position
• Do not often combine with descriptive adjectives
• Have no antonyms

23
WordNet - Adjectives

• Sense 2
• stellar, astral -- (being or relating to or
  resembling or emanating from stars "an astral
  body" "stellar light")
• Pertains to noun star (Sense 1)
• gtstar -- ((astronomy) a celestial body of
  hot gases that radiates energy derived from
  thermonuclear reactions in the interior)
• gt celestial body, heavenly body --
  (natural objects visible in the sky)

24
WordNet - Adjectives

• Reference-Modifying Adjectives
• can be used only attributively
• often have direct antonyms
• Color Adjectives
• are introduced via opposition chromatic vs
  
• achromatic
• Sense 1
• red, reddish, ruddy, blood-red, carmine,
  cerise
• INDIRECT (VIA chromatic) -gt achromatic --
  (having no hue)

25
WordNet - Verbs

• 15 Files
• Bodily Functions Care
• Change
• Communication
• Competition
• Consumption
• Contact
• Cognition
• Creation

• Motion
• Emotion
• Stative
• Perception
• Possession
• Social Interaction
• Weather

26
WordNet - Verbs

• Sense 1
• surfboard, surf -- (ride the waves of the sea
  with a surfboard "Californians love to surf")
• gt glide
• gt travel, go, move, locomote
• Sense 2
• browse, surf -- (look around casually and
  randomly, without seeking anything in particular
  "browse a computer directory" "surf the internet
  or the world wide web")
• gt search, seek, look for
• Sense 3
• surf, channel-surf -- (switch channels, on
  television)
• gt switch, shift, change

27
WordNet - Verbs

• Sense 1
• surfboard, surf
• gt Somebody ----s
• gt Somebody ----s PP
• Sense 2
• browse, surf
• gt Somebody ----s
• gt Somebody ----s something
• Sense 3
• surf, channel-surf gt Somebody ----s

28
WordNet - Verbs

• Verb Sentence Frames
• 1 Something ----s
• 2 Somebody ----s
• 3 It is ----ing
• 4 Something is ----ing PP
• 5 Something ----s something Adjective/Noun
• 6 Something ----s Adjective/Noun
• 7 Somebody ----s Adjective
• 8 Somebody ----s something

29
WordNet - Verbs

• 9 Somebody ----s somebody
• 10 Something ----s somebody
• 11 Something ----s something
• 12 Something ----s to somebody
• 13 Somebody ----s on something
• 14 Somebody ----s somebody something
• 15 Somebody ----s something to somebody
• 16 Somebody ----s something from somebody
• 17 Somebody ----s somebody with something

30
WordNet - Verbs

• 18 Somebody ----s somebody of something
• 19 Somebody ----s something on somebody
• 20 Somebody ----s somebody PP
• 21 Somebody ----s something PP
• 22 Somebody ----s PP
• 23 Somebodys (body part) ----s
• 24 Somebody ----s somebody to INFINITIVE
• 25 Somebody ----s somebody INFINITIVE

31
WordNet - Verbs

• 26 Somebody ----s that CLAUSE
• 27 Somebody ----s to somebody
• 28 Somebody ----s to INFINITIVE
• 29 Somebody ----s whether INFINITIVE
• 30 Somebody ----s somebody into V-ing something
• 31 Somebody ----s something with something
• 32 Somebody ----s INFINITIVE
• 33 Somebody ----s VERB-ing
• 34 It ----s that CLAUSE
• 35 Something ----s INFINITIVE

32
WordNet - Verbs

• Entailment
• P entails Q ?? (PtrueQtrue)\/(PfalseQfalse)
• Unilateral
• Negation reverses the direction of entailment

33
WordNet - Verbs

• Entailment
• Temporal -Temporal
• Inclusion
  Inclusion
• Backward
• 
  Presupposition
• 
  Cause
• Co-extensivness Proper Inclusion
• Troponymy -Troponymy

34
WordNet - Verbs

• Sense 2
• travel rapidly, speed, hurry, zip -- (move very
  fast "The runner zipped past us at breakneck
  speed")
• gt flit, flutter, fleet, dart -- (move
  along rapidly and lightly skim or dart "The
  hummingbird flitted among the branches")
• gt run -- (move fast by using one's feet,
  with one foot off the ground at any given time
  "Don't run--you'll be out of breath" "The
  children ran to the store")
• gt zoom, zoom along, whizz, whizz along --
  (move along very quickly)

35
WordNet - Verbs

• Opposition Relation
• Deadjectival verbs
• Sense 4 shorten
• Antonym of lengthen (Sense 2)
• Morphologically marked verbs
• Sense 1 tie, bind
• Antonym of untie (Sense 1)

• Converses
• Sense 3 give
• Antonym of take (Sense 8)
• Stative verbs
• Sense 1 differ
• Antonym of equal (Sense 1)

36
Word Sense Disambiguation

• Building Semantic
• Concordance

37
Outline

• What Is a Semantic Concordance?
• How it is built?

38
Semantic Concordance

• Semantic concordance is a textual corpus
  and a lexicon so combined that every substantive
  word in the text is linked to its appropriate
  sense in the lexicon via semantic tags Miller,
  1993

39
Semantic Concordance

• Brown Corpus (103 passages)
• The Red Badge of Courage (novella by Stephen
  Crane) (45,600 words)

40
Preproceessing

• Original text
• SGML
  format
• Preprocessing Tokenization
• POS
  Tagging
• Text in ConText
• file format

41
ConText File Format

• Elements
• context elements (ltcontextgt)
• paragraph elements (optional) (ltpgt)
• sentence elements (ltsgt)
• word form elements
  (ltwfgt)
• punctuation elements
  (ltpgt)

42
ConText File Format

• Attribute-value pairs
• cmd the state of semantic tagging
• (tag/done/ignore)
• pos POS tags from preprocessing
• lemma base form of word
• lexsn sense key
• ot other tags (metaphor/idiom)
• pn proper noun (person/loc/institution/other)

43
Developing Procedure

• Sequential annotation
• Verification
• Notes to lexicographers
• Changes to WordNet
• Annotation
• etc.

44
Quality Control

• Check, that all words are tagged
• Check every 11th word
• Check all problem cases
• Calculate the error margin
• Repeat partially the verification

45
References

• Miller, G., (1993)
• Five Papers on WordNet. International Journal
  of Lexicography 4(3).
• Landes et al. (1998)
• Building Semantic Concordances. In C.Fellbaum
  (Ed.), WordNet An Electronic Lexical Data-base,
  pp.199-216.
• Fellbaum et al. (1998)
• Performance Confidence in a Semantic
  Annotation Task. In C.Fellbaum (Ed.), WordNet An
  Electronic Lexical Data-base, pp.217-237.

46
References

• http//wordnet.princeton.edu/