Semantics and Pragmatics of Natural Language Processing
Lectures: Mondays and Thursdays 12:10--1pm in Appleton Tower Room 4.12.
The first lecture is at 12:10 on Monday 7th January, 2008.
This course examines
computational approaches to natural language semantics,
including the interpretation of multi-sentence discourse and
dialogue. Most of the course will focus on the use of formal methods,
yielding formal semantic representations which can be fully
integrated with automated reasoning systems. However, we will also
examine cases where
machine learning semantic information from corpora has proved
successful, and how such approaches might overcome the knowledge
The objective is for students to understand the central problems
in NL semantics, including phenomena on the syntax/semantics and
the semantics/pragmatics interfaces. They will learn about rich,
`hand-crafted' formal models for dealing with these phenomena, as
well as `shallow' methods, which use machine learning on corpora
and other linguistic resources to automatically acquire less
detailed models of meaning. By the end of the course, students
should have gained a working knowledge of the current techniques
used in computational semantics and pragmatics, their use in NLP
applications and their limitations.
- P. Blackburn and J. Bos  Representation and
Inference for Natural Language Volume I: Working With First Order
Logic, CSLI Publications.
- P. Blackburn and J. Bos [to appear] Representation and
Inference for Natural Language Volume II: Working With Discourse
Representation Structures, CSLI Publications.
Download it here.
- S. Bird, E. Klein and E. Loper 
Natural Language Processing In Python, available online from
This book is based on the Natural Language Toolkit (NLTK), which
is installed on DICE as a Python package. For more information on NLTK, see http://nltk.org.
- Week 1: Introduction
- Course Overview
- First-order logic; theory and implementation
- Blackburn and Bos Volume I: Introduction, pp.xi–xvi.
- Blackburn and Bos Volume I:
Chapter 1, pp.1–29.
- NLTK Book Chapter
12, up to and including Section 12.4.
- Week 2: Semantic Construction in the Grammar
- Lambda calculus plus FOL
- Demo/tutorial on using
Assignment 1 (level 10)
Assignment 1 (level 11) now
distributed. Due 1.00 pm, 11th February 2008.
Readings: Blackburn and Bos Volume I
- Week 3: Semantic Composition and Interpretation
- Using Feature-Based Grammars
Notes — revised
- From Syntax to Evaluation
Reading: NLTK Book, Chapters 11 and 12.
- Week 4: Inference and Underspecification
- Propositional Tableaux
Readings: Blackburn and Bos Volume I Chapters 3 and 4
- Week 5: Lexical Semantics
- Polysemy, typed feature structures and inheritance
- statistical models of lexical meaning
- A. Copestake  Representing Lexical Polysemy,
Proceedings of the AAAI Spring Symposium on Lexical
- M. Lapata and A. Lascarides  A Probabilistic Model of
Logical Metonymy, Computational
Linguistics, 29.2, pp263--317.
- Week 6: Introduction to Discourse
- Dynamic Semantics and its Motivation
- DRS Construction and Presuppositions
2 (Level 10) and Assignment
2 (Level 11)
is now distributed. Due 13th March 2008.
You will need the
following Python software to do this assignment:
discourse.zip. Instructions on how to
run this software are given in the assignment.
This software uses Prover9 and
Mace which are already installed on on DICE
machines. If you're interested, you can find information about these
Blackburn and Bos Volume II, chapters 1--3.
- Week 7: Discourse
- Problems in the dynamic semantic treatment of anaphora
- Algorithms for Resolving Pronouns
- Blackburn and Bos Volume II Chapter 4.
- D. Jurafsky and J.H. Martin  Speech and Language
Processing, Prentice Hall.
Chapter 18, pp.669--694
- Week 8: Discourse and Pragmatics
- Interpretation as Abduction
- Segmented Discourse Representation Theory
- J.R. Hobbs, M. Stickel, D. Appelt and P. Martin ,
Interpretation as Abduction, Artificial Intelligence,
Available from here
- A. Lascarides and N. Asher [in press] Segmented Discourse
Representation Theory: Dynamic Semantics with Discourse Structure, to
appear in H. Bunt and R. Muskens (eds.) Computing Meaning Volume
3, Kluwer Academic Publishers.
- Week 9: Discourse and Pragmatics
- Segmented Discourse Representation Theory Continued
- Statistical approaches to speech act recognition
Any one of the following:
A. Stolcke, K. Ries, N. Coccaro, E. Shriberg, R. Bates,
D. Jurafsky, P. Taylor, R. Martin, C. Van Ess-Dykema, & M. Meteer
(2000), Dialogue Act Modeling for Automatic Tagging and Recognition
of Conversational Speech,Computational Linguistics,
Can be retrieved from:
D. Marcu (1999) A decision-based approach to rhetorical parsing. In
Proceedings of the 37th Annual Meeting of the Association for
Computational Linguistics (ACL'99), pages 365-372, Maryland, MD.
Can be retireved from:
D. Jurafsky and J.H. Martin (200). Speech And Language Processing: An
Introduction to Natural Language Processing, Computational
Linguistics, and Speech Recognition. Prentice Hall.
Chapter 19, the section entitled Automatic Interpretation of
- Week 10
There will be two lectures each week (Monday and Thursday at
12:10pm), and students
are expected to attend and take part in them.
There will be two moderate-sized assessed assignments during the
module. Students are also encouraged to work on exercises in the
text, to gain confidence with the material.
- Alex Lascarides' office hours are on Wednesdays, 11am to 12 noon (Room
8, 2nd floor left, 2 Buccleuch Place).
Students with questions about or problems with the module
material are encouraged to come by for consultation during these
Assessment will be by two moderate-sized assignments during the
module, and a final paper. The total mark for the course will be
determined by the paper accounting for 40% of the mark and by the
two assessed assignments, each contributing 30%.
Students entering the level 11 course should have successfully
completed Linguistic and Computational Theories of Grammar (LCTG) or
equivalent. Those entering the level 10 coures should have Informatics
2A and 2B or something equivalent.
All students should have elementary knowledge of first order logic and
natural language syntax.
The level 11 version of SPNLP
is only open to MSc students and students in
and Linguistics program. Other students will be accepted into the
module only with permission of the instructor.
The level 10 version of SPNLP is open to all fourth years who have the