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


Contents

Announcements
Topics
Coursework
Other sources of information
Schedule


Announcements


Topics

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.


Coursework and Prerequisites

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

  1. 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

  2. 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.

  3. 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


Class Newsgroup


Some Other Sources of Information


Schedule


To the IU Bloomington Home Page. To the IUB
Computer Science Department Home Page.

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