Date: Wed, 20 Nov 1996 22:29:01 GMT
Server: Apache/1.0.3
Content-type: text/html
Content-length: 12535
Last-modified: Thu, 11 Apr 1996 04:38:16 GMT
C661: Natural Language Processing
C661: Natural Language Processing
Instructor: Mike Gasser
[Make an
appointment with me.]
[Send me a message.]
Time: TuTh 1:00-2:15
Room: Woodburn 114
Announcements
Topics
Coursework
Other sources of information
Schedule
- Here is the final exam.
- A (fairly primitive) program for playing with Holographic
Reduced Representations (Plate) is at ~gasser/Apps/hrr on the
sharkestra, moose, and department SGIs.
Here is enough information to get you started
with it.
- Here
is a new (optional) paper by Paul Rodriguez on
context-free languages and simple recurrent nets.
This course provides an introduction to the field of natural language
processing (or computational linguistics), including both analysis and
generation.
Speech processing, machine translation, and computational
approaches to language acquisition and language evolution
are also given some attention.
A wide range of linguistic
phenomena, including phonology, morphology, syntax and semantics, and
pragmatics, will be treated,
and examples will come from various languages.
We will be concerned both with how well particular approaches solve
practical problems and with how well they model human data.
The course is divided into two relatively separate components.
The first deals with symbolic approaches to language
processing.
We will cover parsing and generation algorithms, emphasizing
modern unification-based approaches, but will spend
more of our time considering the sorts of grammars that support
parsing and generation.
With respect to theory and notation, we will stick mainly with
Head-Driven
Phrase Structure Grammar,
probably the most popular approach in computational
linguistics today.
The second component of the course
deals with statistical and connectionist approaches to language
processing, which, despite their very different
origins and motivation, share many underlying mechanisms as well as a
lack of built-in linguistic knowledge.
We will emphasize the acquisition of knowledge (phonological,
morphological, syntactic, semantic),
temporal processing, and
the relation between perception and the grammar/lexicon.
The course schedule, however, will be organized around topics rather
than approaches.
Thus we will look at morphology, approaches to parsing, and semantic
case, for example, in each case considering how both symbolic and
connectionist/statistical approaches deal with the problem.
For each topic we will also look at acquisition as well as processing.
Students should have some background in AI (such as
C563-564) and be
able to program in Scheme or Lisp. Some linguistics background would
also be very helpful but is not required.
Cognitive science students from outside of computer science are
encouraged to enroll.
Coursework includes
- Project (50%)
This may be done in collaboration with others in the class.
It should include a running program, though this can be based on
existing software in the case of connectionist models,
and a report which relates the work to other work in the area.
An attempt will be made to relate projects to each other by
constraining the type of language that is handled.
Grading: paper (25%); relevance (25%); originality, success, lessons learned
(50%)
Suggestions for projects
A simple story you might want to use
for your project
- Exams (40%)
There will be two exams, each covering half of the course.
You need only take the portion of each exam covering the approach
(symbolic or connectionist/statistical) which is not related to your
project.
- Discussion of papers (10%)
Students will be responsible for leading discussion of some of the
papers we will be reading.
Here's a schedule.
Readings for the class will be kept on reserve in Swain Library. A
copy will also be left in a box in the Computer Science Department
Copy Room.
Reading list
Discussants for readings
- Week 1: Introduction
- Week 2: Words -- Phonology, Morphology: A Symbolic Approach
- Week 3: Words -- Phonology, Morphology: Connectionist Approaches
- Week 4: Words -- Lexical Semantics: Symbolic and Connectionist
Approaches
- Week 5: Words -- Lexical Semantics: Statistical Approaches; Speech
- Week 6: Phrases and Sentences -- Context-Free Grammars, Augmented
Grammars
- Week 7: Phrases and Sentences -- Syntax and Semantics, Compositionality
- Week 8: Phrases and Sentences -- Unification Grammars
- Week 9: Phrases and Sentences -- Connectionist Syntax and Semantics
- Week 10: Phrases and Sentences: Connectionist Structure (cont.),
Parsing
- Week 11: Phrases and Sentences -- Parsing and Generation;
Spreading-Activation Approaches
- Week 12: Machine Translation, Statistical Approaches Again,
Discourses -- Symbolic Approaches
- Week 13: Discourses -- Symbolic Approaches
- Week 14: Discourses -- Symbolic and Connectionist Approaches
- Reading: St. John
- Tu
Progress report on project due
St. John (Wendy)
Using world knowledge
- Th
Discourse structure
Mélange of last-minute topics: metaphor, metonymy, humor,
deception (none of which I have a good account of)
- Week 15: Language Acquisition Revisited
- Week 16
- Mo (23:59:59)
Exams due (electronically)
- Tu (5:00-7:00pm)
Reports on projects
Project reports due
Last updated: 17 December 1995
URL: http://www.cs.indiana.edu/classes/c661/home.html
Comments:
gasser@salsa.indiana.edu
Copyright 1995,
The Trustees of
Indiana University