1. Lecture
Lexical Semantics
CS 4705

2. What is lexical semantics?
Meaning of Words
Lexical Relations
WordNet
Thematic Roles
Selectional Restrictions
Conceptual Dependency

3. What is a word? Lexeme: an entry in the lexicon that includes
an orthographic representation
a phonological form
a symbolic meaning representation or sense
Dictionary entries:
Red (‘red) n: the color of blood or a ruby
Blood (‘bluhd) n: the red liquid that circulates in the heart, arteries and veins of animals
Word Sense Disambiguation
For any given lexeme, how can its sense be reliably distinguished?
Lex. Rel. III: Metaphor, Metonymy
What is metaphor?
That doesn't scare Digital.
What is metonymy?
GM killed the Fiero.
Extension of existing sense to a new meaning.
Lexical Relations IV: Synonomy
What is synonomy? Substitutability.
How big is that plane?
How large is that plane?
Compare:
A big fat apple
?A large fat apple
A big sister
?A large sister
Influences on substitutability:
subtle shades of meaning differences
polysemy
register
collocational constraints
Lexical Relations V: Hyponomy
What is hyponomy?
General: hyponym
Specific: hypernym
Example: ``car'' is a hyponym of ``vehicle''
and ``vehicle'' is a hypernym of ``car.''
Test: ``That is a car'' implies ``That is a vehicle''
What is ontology?
What is taxonomy?
What is object hierarchy?
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Semantic Networks
Used to represent relationships between words
Example: WordNet - created by George Miller's team at Princeton
WordNet (1.6)
WordNet is the most widely used hierarchically organized lexical
database for English -- Fellbaum (1998).
\vspace{.1in
\epsfxsize=1\textwidth
\fig{\file{figures{fig16.01.ps
Format of WordNet Entries
WordNet sense entries consist of a set of synonyms, a dictionary-style
definition (or gloss), and some examples of uses.
\fig{\file{figures{fig16.02.ps
Sense Distribution for WordNet Verbs
\fig{\file{figures{fig16.03.ps
Lexical (N) Relations in WordNet
\fig{\file{figures{fig16.04.ps
Verb Relations in WordNet
\fig{\file{figures{fig16.05.ps
Adj. and Adv. Relations in WordNet
\fig{\file{figures{fig16.06.ps
Synsets in WordNet
WordNet is organized around the notion of synset.
{ chump, fish, fool, gull, mark, patsy, fall guy, sucker, schlemiel,
shlemiel, soft touch, mug }
Important: It is this exact synset that makes up one of the sense for
each of the entries listed in the synset.
Theoretically, each synset can be viewed as a concept in a taxonomy --
like the concepts described in Chapter 14.
Hyponomy in WordNet
fig16.07.ps
Internal Structure of Words
What are the meaning components underlying word sense?
Thematic Roles (theta-roles)
What is a thematic role?
E w,x,y,z Giving (x) ^ Giver(w,x) ^ Givee(z, x)
^ Given(y,x) $
E w,x,z Breaking (x) ^ Breaker(w,x) ^ Broken(z,x)
Generic Thematic Roles
fig16.08.ps
Examples of Thematic Roles
fig16.09.ps
Early Theories of Thematic Roles
: "The beginning of Lexical Semantics" (Fillmore; Gruber;
Jackendoff (based on Gruber))
Two fundamentally different approaches to linguistics
Gruber/Jackendoff: account for semantics and use grammar derived to say
something about syntax
Fillmore: account for syntax and use that to describe semantics
Thematic Level
Why posit a thematic level distinct from that of syntactic
subcategorization?
capture similarity between different (but related) uses of same
lexical item)
obviate need for subcategorization frames: mapping from syntax to
lexical semantics
Selectional Restrictions
What are selectional restrictions?
Recall the "Godzilla" example.
Selectional Restriction Implementation
A WordNet approach: hamburgers are edible
fig16.10.ps
Primitive Decomposition
Jim killed his philodendron
Jim did something to cause his philodendron to become not alive
Schank's Primitives
Conceptual Dependency
fig16.11.ps
Pred. Independence vs. Dependence
Predicate-Independent
single set of roles is chosen independent of the type of predicates involved (no reference to
type of predicates)
Schank
Predicate-Dependent
roles identified by particular positions arguments
occupy with respect to primitive predicates
Decomposition vs. Non-Decomposition
Decomposition / Compositional Approach (Schank, Jackendoff) vs.
Non-decomposition / Noncomposition Approach (Fillmore)
Within compositional approaches: exhaustive (Schank) vs. nonexhaustive
(Jackendoff)
Schank: Motivation
Underlying Motivation: "Strong AI"
Focus: understanding. Argues that the representation is reversible.
Rejects syntax during analysis. Allows it during generation.
Attempts to come up with well-defined system of rules and
conceptualizations.
Inferences, expectation, syntax, conversational norms, real world.
Conceptual Structure (CD): Language-independent conceptual level.
Schank: Kill vs. Die
Schank: Problem 1
"John caused Mary to die" vs. "John killed Mary"
Identically substitutable?
Flaw of all compositional approaches of this nature.
Schank: Problem 2
The decompositions are very complex.
Too specific
Why are these conceptualizations so radically distinct from the
syntactic realization?
Talks CD from NL understanding point of view -what about generation?
The NLP Bottleneck
Acquisition of Computational Lexicons
For Next Time
Chapter 18

4. Do dictionaries give us definitions?
Right (‘rt) adj: located nearer the right hand esp. being on the right when facing the same direction as the observer
Left (‘left) adj: located nearer to this side of the body than the right
Do dictionaries give us definitions?
Some are circular
All are defined in terms of other lexemes
You have to know something to learn something
What can we learn from dictionaries?
Relations between words:
Oppositions, similarities, hierarchies

5. Homonomy
Homonyms: Words with same form but different, unrelated meanings, or senses (multiple lexemes)
A bank holds investments in a custodial account in the client’s name.
As agriculture is burgeoning on the east bank, the river will shrink even more
Word sense disambiguation: what clues?
Similar phenomena
homophones - read and red (different orth. form)
homographs - bass and bass (different phon. form)

6. Ambiguity: Which applications will these cause problems for?
General semantic interpretation
Machine translation
Spelling correction
Speech recognition
Text to speech
Information retrieval

7. What is polysemy?
Word with multiple but related meanings (same lexeme)
They rarely serve red meat.
He served as U.S. ambassador.
He might have served his time in prison.
What’s the difference between polysemy and homonymy?
Homonymy:
Distinct, unrelated meanings
Different etymology? Coincidental similarity?

8. For either, practical task:
Polysemy:
Distinct but related meanings
idea bank, sperm bank, blood bank, bank bank
How different?
Different subcategorization frames?
Domain specificity?
Zeugma: Can the two candidate senses be conjoined?
?He served his time and as ambassador to Norway.
For either, practical task:
What are its senses? (related or not)
How are they related? (polysemy ‘easier’ here)
How can we distinguish them?

9. Metaphor, Metonymy What is metaphor? What is metonymy?
Father of the atom bomb.
What is metonymy?
GM killed the Fiero.
The ham sandwich wants his check.
Both extend existing sense to new meaning
Metaphor: use completely different concept (but cf conventional metaphors like GM)
Metonymy: use related concepts

10. Synonomy Substitutability: different lexemes with the same meaning
How big is that plane?
How large is that plane?
How big are you? Big brother is watching.
What influences substitutability?
Polysemy (large vs. old sense)
register: He’s really cheap/?parsimonious.
collocational constraint: roast beef, ?baked beef
convention: economy fare/?price

11. Hyponomy General: hypernym (super…ordinate)
dog is a hypernym of poodle
Specific: hyponym (under..neath)
poodle is a hyponym of dog
Test: That is a poodle implies that is a dog
What is ontology? Object in some domain
What is taxonomy? Structuring of those objects
What is object hierarchy? Structured hierarchy that supports feature inheritance

12. Semantic Networks Used to represent lexical relationships
e.g. WordNet (George Miller et al)
Most widely used hierarchically organized lexical database for English
Synset: set of synonyms, a dictionary-style definition (or gloss), and some examples of uses --> a concept
Databases for nouns, verbs, and modifiers
Applications can traverse network to find synonyms, antonyms, hierarchies,...

13. Is a rock edible?
What are the parts of a human body?
What is a cheeseburger?
What are its parts?
What is the opposite of ambitious?

14. Thematic Roles E w,x,y,z Giving (x) ^ Giver(w,x) ^ Givee(z, x)
^ Given(y,x)
E w,x,z Breaking (x) ^ Breaker(w,x) ^ Broken(z,x)
A set of roles:
agent, experiencer, force, theme, result, content, instrument, beneficiary, source, goal,...
The dog ate the cheeseburger.
What is cheeseburger?
The sniper shot his victim with a rifle.
What is rifle?

15. Early Theories of Thematic Roles
: "The beginning of Lexical Semantics" (Fillmore; Gruber; Jackendoff (based on Gruber))
Two fundamentally different approaches to linguistics
Gruber/Jackendoff: account for semantics and use grammar derived to say something about syntax
Fillmore: account for syntax and use that to describe semantics

16. Why do we need a thematic level?
We already have syntactic subcategorization
Capture similarity between different (but related) uses of same lexical item)
Avoid need for subcategorization frames: mapping from syntax to lexical semantics

17. What are Selectional Restrictions?
George ate a cheeseburger/his lunch/dirt.
Jim killed his philodendron
?His philodenron killed Jim.
The flu killed Jim.

18. Schank's Conceptual Dependency
Eleven predicate primitives represent all predicates
Objects decomposed into primitive categories and modifiers
But few predicates result in very complex representations of simple things
Ex,y Atrans(x) ^ Actor(x,John) ^ Object(x,Book) ^ To(x,Mary) ^ Ptrans(y) ^ Actor(y,John) ^ Object(y,Book) ^ To(y,Mary)
John caused Mary to die vs. John killed Mary

19. Next time Word sense disambiguation and information retrieval
Chapter 17