This course explores lexical semantics from computational, linguistic and philosophical perspectives. It covers theories of word meaning, differences in semantic structure between languages, and the relationship between word meaning and sentence meaning. Students will learn about primitives of meaning, semantic annotation, ontologies, and automated methods for acquiring semantic knowledge at scale. The course consists of 5 modules taught by different professors on topics like computational lexical semantics, deep semantics, ontologies, linguistic issues, and semantic annotation. There will be 4 assignments worth 25% each.

1. CS599: Lexical Semantics
Course Overview and Introduction
Patrick Pantel
USC Information Sciences Institute
http://www.isi.edu/~pantel
pantel@isi.edu
Theme
Simply put, this course explores the meaning of words
Lexical semantics:
is the study of how and what the words of a language denote
is deeply rooted in Linguistics, but is becoming an increasingly important part of
Natural Language Processing (NLP)
it covers theories of:
the classification and decomposition of word meaning
the differences and similarities in lexical semantic structure between different
languages
the relationship of word meaning to sentence meaning and syntax
We will explore different perspectives of lexical semantics, from the point of
view of:
NLP/AI, Semantics/Philosophy, Linguistics/Lexicography, and Ontologies/KR
(Knowledge Representation)
Some core issues we will cover:
primitives of meaning, the creation of semantically annotated corpora,
ontologies, automated methods for acquiring semantic knowledge on a large
scale, and a survey of related perspectives
© 2005 Patrick Pantel CS599 - Course Overview and Introduction 2
1

2. Demos
http://www.isi.edu/~pantel/demos.htm
© 2005 Patrick Pantel CS599 - Course Overview and Introduction 3
Some Definitions…
Important terms:
Syntax: the study of how words combine to form grammatical sentences
Semantics: the study of the meaning of words and how these combine to form
the meanings of sentences
Lexicography: the study of the design, compilation, use and evaluation of
general dictionaries (lexicons)
Interlingua: auxiliary language describing the language-independent meaning of
a sentence
Natural Language Processing vs. Computational Linguistics
Applications
Information Retrieval (IR): system used for searching a natural language
database (e.g. Google, Yahoo!, Microsoft Search)
Information Extraction (IE): system used to extract targeted information from
structured or unstructured information medium
Machine Translation (MT): system that translates spoken/written language
from one language to another (e.g. Systran, LanguageWeaver)
Question Answering (QA): specific IR system that answers natural language
questions with targeted text (instead of returning documents)
© 2005 Patrick Pantel CS599 - Course Overview and Introduction 4
2

3. Lexical Relations
Lexical relations are used as tools in this field:
synonymy
e.g., tome/book, love/adore
antonymy (opposition)
switching thematic roles associated with the verb (buy – sell)
stative verbs (live – die)
sibling verbs which share a parent (walk – run)
restitutive opposition: antonymy + happens-before
(damage - repair)
hyponymy/hypernymy (subclass/superclass)
e.g. stool is-a chair, apple is-a fruit, face is-a external-body-part
meronymy/holonymy (part-of/has-part)
e.g. wheel part-of car, chair part-of committee, face part-of head
cause, strength, temporal precedence, entailment…
© 2005 Patrick Pantel CS599 - Course Overview and Introduction 5
Logistics
Course: CS599 – Lexical Semantics (3 units)
Course website: http://www.isi.edu/~pantel/Content/Teaching/2005/cs599.htm
Class schedule: Tuesday/Thursday: 2:00 pm - 3:20 pm (GFS107)
Instructors: Prof. Eduard Hovy (CS)
Prof. Jerry Hobbs (CS)
Prof. Robert Belvin (Ling)
Prof. Patrick Pantel (CS)
Office hours: TBA
Grading: 4 assignments (25% each)
Readings: Instructors will recommend foundation and cutting edge
papers as weekly readings.
© 2005 Patrick Pantel CS599 - Course Overview and Introduction 6
3

4. What I need from you…
Please email the following information to pantel@isi.edu
before the next class
Full name + student number
Email address
Home department
Are you taking the class for credit or are you auditing it?
Are you familiar with the programming language Perl?
What other programming languages do you use?
Do you have any comments or concerns?
© 2005 Patrick Pantel CS599 - Course Overview and Introduction 7
Course Outline
Module 1 - Computational Lexical Semantics (Prof. Pantel)
Computational formulation of lexical semantics
Statistics and information theory
Automated methods for harvesting semantics (corpus- and web-based)
Applications (building a thesaurus, extracting paraphrases, discovering word
classes, inducing word senses, automatically linking learned knowledge into
formal ontologies)
August 25, 30; September 1, 6, 8, 27 (with Hovy)
Module 2 - Deep Lexical Semantics (Prof. Hobbs)
Introduction to deep lexical semantics
Interpretation as abduction
Cognition and the cognitive lexicon
Time and the word quot;Now“
Causality and modality
Similarity and the preposition quot;Like“
September 13, 15, 20; October 18, 20, 25, 27
© 2005 Patrick Pantel CS599 - Course Overview and Introduction 8
4

5. Course Outline
Module 3 - Ontologies (Prof. Hovy)
Semantic primitives
Introduction to ontologies
Ontologies for shallow semantics
Upper and middle models
Verb sense frames
The Omega ontology
September 22, 27 (with Pantel), 29; November 8, 10, 15, 17, 22
Module 4 - Linguistic Issues (Prof. Belvin)
Semantic cases, semantic fields
Thematic relations hypothesis
Lexical semantic decomposition
Verb classes and alternations
Formalisms and notation
Mapping
October 4, 6, 11, 13; November 1, 3, 29
© 2005 Patrick Pantel CS599 - Course Overview and Introduction 9
Course Outline
Module 5 – Annotation of Shallow Semantics (Prof. Hovy)
Sense annotation as in PropBank
The giant leap from senses to concepts
Annotation and verification of sense and concept creation
November 17, 22
© 2005 Patrick Pantel CS599 - Course Overview and Introduction 10
5

6. Grading
There will be four assignments, each worth 25%
Assignment 1: Corpus- and web-based knowledge
harvesting (programming)
Assignment 2: Deep lexical semantics
Assignment 3: Syntax to semantics mapping
Assignment 4: Annotation of shallow semantics
© 2005 Patrick Pantel CS599 - Course Overview and Introduction 11
A Few Prominent Figures
CS: Linguistics:
Christiane Fellbaum Bernard Comrie
William Croft
Graeme Hirst Charles Fillmore
Jerry Hobbs Ken Hale
Mitch Marcus Ray Jackendoff
Dan Moldovan George Lakoff
Sergei Nirenburg Ronald Langacker
Beth Levin
Martha Palmer James McCawley
James Pustejovsky Leonard Talmy
Roger Shank Anna Wierzbicka
… …
© 2005 Patrick Pantel CS599 - Course Overview and Introduction 12
6

7. Module 1 - Computational
Lexical Semantics
Prof. Patrick Pantel (CS)
August 25, 30; September 1, 6, 8, 27
Topics:
Computational formulation of lexical semantics
Distributional Hypothesis
Links the semantics of words to the syntactical uses
Words that occur in the same contexts tend to have similar
meanings
A bottle of tezgüno is on the table.
Everyone likes tezgüno.
Tezgüno makes you drunk.
We make tezgüno out of corn.
What is tezgüno?
© 2005 Patrick Pantel CS599 - Course Overview and Introduction 13
Module 1 - Computational
Lexical Semantics
Topics (continued…):
Statistics and information theory
Mutual information, log-likelihood, similarity metrics
Text- and web-mining algorithms for harvesting word
semantics, and semantic relations between words
Applications
building a distributional thesaurus
extracting paraphrases
discovering word classes
inducing word senses
automatically linking harvested knowledge into formal
ontologies
© 2005 Patrick Pantel CS599 - Course Overview and Introduction 14
7

8. Module 2 - Deep Lexical
Semantics
Prof. Jerry Hobbs (CS)
September 13, 15, 20; October 18, 20, 25, 27
Topics:
Introduction to deep lexical semantics
Interpretation as abduction
Cognition and the cognitive lexicon
Time and the word quot;Now“
Causality and modality
Similarity and the preposition quot;Like“
© 2005 Patrick Pantel CS599 - Course Overview and Introduction 15
Module 3 - Ontologies
Prof. Eduard Hovy (CS)
September 22, 27, 29; November 8, 10, 15, 17, 22
Topics:
Shallow semantics
When and why do we need shallow semantics?
MT toward interlinguas, better IR/IE/summarization
Ontology: conceptualization of knowledge domain
Controlled vocabulary of all concepts + a set of relations that link concepts +
language to make queries / assertions / inferences
Upper model, middle model, domain models
Dealing computationally with shallow semantics
Ontology vs. lexicon
Looking under the hood of ontologies
The WordNet termbank (ontology?)
The Omega ontology
© 2005 Patrick Pantel CS599 - Course Overview and Introduction 16
8

9. Module 4 - Linguistic Issues
Prof. Robert S. Belvin (Ling)
October 4, 6, 11, 13; November 1, 3, 29
Topics:
Linguistic study of lexical semantics with focus on benefits for
automated systems
Lexical semantics vs. lexicography vs. formal semantics
Different views of the language faculty and how that can impact
how we do semantic representation and mapping
Gruber's thematic relations hypothesis
Lexical semantic decomposition
some semantic primitives, cross-categorical semantic features,
languages which have transparent decomposition/morphologically
complex forms, lexicalization patterns (in English and cross-
linguistically)
© 2005 Patrick Pantel CS599 - Course Overview and Introduction 17
Module 4 - Linguistic Issues
Topics (continued…)
Verb Classes and Alternations
valence-changing operations (in English and cross-linguistically)
the quot;unaccusativequot; hypothesis
Some Formalisms and Notation
Lexical Conceptual Structure
Parsons Event Semantics
Image-schemas
Mapping
how do you get semantic representation out of syntactic structure?
subcategorization and selection
thematic role hierarchy and universal theta alignment hypothesis
semantic contribution of syntactic structure
© 2005 Patrick Pantel CS599 - Course Overview and Introduction 18
9

10. Module 5 - Annotation of
Shallow Semantics
Prof. Eduard Hovy (CS)
November 17, 22
Topics:
Manually acquiring shallow semantic knowledge is a
difficult task
Determine what you want
Derive a representational formalism that captures it
Get people to do it!! (or get machines!!!)
Lexical term creation, definition and extraction
Manual effort based on inter-annotator agreement
How do we ensure consistency and quality
Senses vs. concepts
© 2005 Patrick Pantel CS599 - Course Overview and Introduction 19
10