Computational Semantics and Pragmatics
Lecturer: Raquel
Fernández (ILLC, University of Amsterdam)
Timetable: Tuesdays 13-15h in G3.13 and Wednesdays
15-17h in A1.06 (both rooms at Science Park 904).
Prerequisites:
There are no formal prerequisites, but I'll assume some basic
knowledge of semantics/pragmatics of natural language. Programming
skills are not required, although they are useful.
Overview:
Semantics and pragmatics are concerned with the study of natural
language meaning and its context of use in written texts and in
conversation. The computational counterparts of these disciplines
address these issues from an explicitely empirical and computational point of view,
combining insights from linguistic theory, computational linguistics,
and artificial intelligence. The course will introduce some of the
fundamental concepts in contemporary computational semantics and
pragmatics, exposing students to current research in topics such as
distributional lexical semantics, generation and resolution of
referring expressions, speech acts, and dialogue modelling.
Evaluation:
The course will consist of
lectures and discussions of research papers; everybody is expected
to play an active role in class. Students will be asked to write
a short paper by the end of the course (the last couple of weeks
will be dedicated to this). The final paper will account
for 70% of the overall grade. The remaining 30% will correspond to
exercises, and presentations and discussions of readings.
This website (especially the stuff below) will be updated throughout
the course. Please check it regularly. Its contents are subject to change.
Week 1
Main topic: Lexical Semantics
- 3 Sept: Course overview [slides]. Introduction to
lexical semantics [slides].
To do: Read the paper by Lapata below and find out what WordNet is
and how it is organised.
I recommend you to install Python and NLTK
(see nltk.org for instructions) and go
over the section on
WordNet (section 2.5 in chapter two) of the NLTK book. If you don't
know any Python, start from the beginning of the book; the book does
not assume you know any programming language.
- 4 Sept: Discussion of the following
paper:
More on lexical semantics: WordNet; theories of concepts and
typicality effects [slides].
To do: Read the overview papers on DSMs. Go over section 2.5 of
the NLTK book.
Week 2
Main topic: Distributional Semantic Models
- 10 Sept: Introduction to DSMs [slides].
Background readings and resources:
To do: Have a look at the paper by Abramova et al. and at
homework #1 (both are below).
- 11 Sept: More on DSMs [slides].
Brief discussion of the following paper exemplifying use of DSMs and
WordNet:
- Katja Abramova, Raquel Fernández, and Federico Sangati (2013)
Automatic Labeling of Phonesthemic Senses. In Proceedings of the 35th Annual Conference of the Cognitive Science Society, pp. 1696-1701. Berlin, Germany.
To do: Read the seminal paper on the generation of referring
expressions by Dale & Reiter (1995).
Homework #1: due on Friday 20 Sept. Submit via email a PDF
and code file named "yourname_hw1".
[suplementary material: python
code impleming a basic DSM and code documentation]
Week 3
Main topic: Referring Expressions (generation)
- 17 Sept: Introduction to the generation of referring
expressions [slides]
Readings and resources:
- 18 Sept: More on GRE [slides]
Week 4
Main topic: Referring in Dialogue
- 24 Sept: Introduction to referring in interactive
settings [slides]
Background readings:
We'll start the lecture by discussing the approach taken in this short paper
(you are welcome to have a look at it if you wish):
To do: Read the paper by Jordan & Walker (2005)
- 25 Sept: More on referring in interactive settings [slides]
Discussion of the following paper:
Homework #2:
due on Tuesday 1 Oct by 13:00 (before class).
[suplementary material: TUNA Corpus data sample]
Week 5
Main topic: Dialogue Modelling
- Background readings:
- 1 Oct: Intro to dialogue: speech act theory vs. joint action
model and interactive alignment model [slides]
- 2 Oct: More on dialogue: dialogue acts and dialogue
coherence, plus turn taking [slides]
Homework #3:
due on Friday 11 Oct by 14:00.
[suplementary material: Switchboard Corpus]
To do: Choose a topic for your final project/paper related to the
subjects covered in the course.
Send me an email about
your proposed topic as soon as possible but no later than Monday 7
Oct 13:00. If you want to meet up with me, I'll be available
that Monday between 16:00 and 18:00 (send me an email before Monday to make an appointment).
Week 6
Presentations of plans for project/final paper.
Attendance to
all presentations is mandatory.
5-minute presentations with up to 10 minutes for discussion (
with
flexibility in this case).
- 8 Oct: Julian, Lydia/Agnes, Sarah, Henning/Dominik
- 9 Oct: Remi, Suzanne, Robert
Week 7
Individual supervision meetings to discuss ongoing project work (in my
office at the ILLC, room F1.07 in SP 107; entrance through Nikhef at
SP 105).
Bring the current draft of your final paper to the meeting.
- Tuesday 15 Oct
- 15:30-16:00: Remi
- 16:00-16:30: Robert
- 16:30-17:00: Dominik & Henning
|
- Wednesday 16 Oct
- 14:00-14:30: Suzanne
- 14:30-15:00: Sarah
- 15:30-16:00: Julian
- 16:00-16:30: Agnes & Lydia
|
Week 8
- Thursday 24 Oct, room A1.06: final presentations of projects. Attendance to
all presentations is mandatory.
- 10:30-10:50: Robert
- 10:50-11:10: Sarah
- 11:10-11:20: break
- 11:20-11:40: Julian
- 11:40-12:00: Suzanne
- 12:00-12:40: lunch break
- 12:40-13:00: Remi
- 13:00-13:25: Agnes & Lydia
- 13:25-13:35: break
- 13:35-14:00: Dominik & Henning
- Sunday 27 Oct: final paper due
Guidelines for Final Projects
Final presentation
- Timeframe: up to 15 minutes for individual projects and up to
20 minutes for non-individual projects. In either case, there will
be 5 extra minutes for questions. Only very brief clarification
questions are possible during the talk.
- Evaluation: I will grade your presentation according to
these criteria. The presentation is
worth 10% of your final grade.
Final paper [ due on Sunday 27 Oct midnight ]
- Format: your final paper should be formatted using the Latex style files
of EACL 2014, which you can download from here (scroll down to
"Submission Format").
- Length: between 5 and 8 pages, including references.
- Structure: your paper should include a short abstract, an
Introduction section, a Related Work section, some sections where
you present your analyses and results, and a Conclusions
section. Survey papers will have a slightly different structure, but
they should definitely include an abstract as well as Introduction and
Conclusions sections.
- Evaluation: I will evaluate your paper in a similar
fashion as reviewers evaluate papers that are submitted to a
conference. Have a look at the review forms for ACL
2010 (the relevant ones are those tagged as "long"). The most
relvant criteria for the purpose of the course are: CLARITY
(including layout, style, orthographic and grammatical correctness),
CORRECTNESS (of the analysis and the methodology), MEANINGFUL COMPARISON,
SUBSTANCE, and REPLICABILITY. See the review forms for further details.