CS 772 ADVANCED NATURAL LANGUAGE PROCESSING CONCEPTS

Instructor:

Dr. Michael T. Cox
cs772fac@cs.wright.edu
http://www.mcox.org/

Office:

RC343 (phone 775-5126)

Office Hours:

5:55 PM-6:55 PM TTh

Prerequisites:

CS 771 Natural Language Processing Techniques

Meeting Times:

TTh 7:00-8:15, Room RC302 Russ Engineering.

Credit Hours:

4 (3hr lecture, 2hr lab)

Syllabus

Course Description:

"Advanced Natural Language Processing (NLP) Concepts" is a graduate level course in natural language understanding by computer, designed for students interested in or specializing in Artificial Intelligence (AI) or language. We will cover basic methods for parsing and generating meaning for English text from an AI perspective. The course is taught from two perspectives: from a cognitive perspective that attempts to develop computational theories of human language abilities, and from an applied systems perspective that uses language in a large range of tasks. The course will concentrate on semantics, rather than syntax. Topics include advanced parsing techniques, semantic correspondence theory, verb acquisition, information extraction, the impact of memory constraints on language, speech processing, and the use of world knowledge for language tasks.

The intent of the course is to provide a background in NLP, an exposure to the major issues and methods in the field, and experience in writing NLP programs. Classes will consist of active discussions in addition to lectures; thus, class participation will be important and encouraged. Because of this reason, and because lectures will contain information not covered in the text and for which the student will be responsible on tests, attendance is crucial. Class lectures and discussions will be complemented by programming assignments and supplementary reading assignments. All programming will be done in Common Lisp.

Objectives:

The main objectives of the course are

to provide an advanced survey of recent research in NLP
to develop a conceptual understanding of the basic issues and major topics in NLP
to develop design skills for building language understanding systems.

Required and Suggested Texts:

Ram, A, & Moorman, K. (1999). Understanding Language Understanding. Cambridge, MA: MIT Press.
J. Allen, Natural Language Understanding (second edition). Benjamin/Cummings, 1995.
Norvig, P. Paradigms of Artificial Intelligence Programming: Case Studies in Common Lisp, Morgan Kaufmann, 1992. ISBN 1-55860-191-0. (Advanced AI programming techniques.) http://www.norvig.com/paip.html

Supplementary Required Readings

Lisp and AI Programming Books and Resources:

Herner, A. (1999). Natural Language Processing Demonstrations on the Net (Organized in parallel with chapters from Allen's text). http://www.wright.edu/~herner.2/
Steele, G.L. Jr., (1990). Common Lisp: The Language, 2nd ed., Digital. ISBN 1-55558-041-6. (The definitive Common Lisp manual.) file:///usr/local/cltl/cltl2.html (AVAILABLE FROM CHEOPS ONLY)
Taylor, L., The Lisp Programmer's Guildhall. http://www.cc.gatech.edu/computing/classes/cs2360/ghall/
Touretzky, D.S. (1990). Common Lisp: A Gentle Introduction to Symbolic Computation, Benjamin/Cummins. ISBN 0-8053-0492-4. (For Lisp novices.)

Academic Dishonesty:

Academic dishonesty will not be tolerated in class, on homework, or during examinations. For a list of examples of cheating see the Code of Student Conduct section of the Wright State University Student Handbook. Search for "Academic Dishonesty Defined" to find the exact location where the examples are listed.

Student Evaluation:

There will be a term project (in place of a midterm exam) constituting 20% of the grade. The final exam will be worth 35% of the overall grade for the class. You will be tested on your understanding of the principles and your ability to apply them to new and different problems. Homework assignments will provide 15% and a major programming project will provide the additional 30%. Class participation and attendance will also be significant biasing factors in the evaluation for those students on the border between letter grades.

All assignments must be done independently. You are encouraged to talk to the instructor, if you encounter any difficulties. You may discuss the problem, and the issues that arise, with other students, as long as you clearly indicate who you discussed the problem with. You may not share or discuss your code. In other words, talk about the problem, not about the program itself. (This is supposed to be a common sense guideline; please talk to the instructor, if you are unsure about what is permissible.) If you use algorithms or methods described in books or papers, make sure you include a comment in your program with the appropriate references.

Many assignments will contain one or more extra credit problems. These are designed for students who need extra credit to remedy a poor score, and also for students who wish to explore particular topics in greater depth. The final grades will be assigned on a curve based on regular credit problems, after which students with extra credit will be scaled up appropriately. This ensures that extra credit can only help those students who have it, but not hurt those who don't.

Homework

NLP Frequently Asked Questions

College Computing Resources:

Allegro Common Lisp is available on one machine in the Department of Computer Science and Engineering. This is cheops.cs.wright.edu. You will be given class accounts to access this software. College computer labs, software availability, and the user help desk are listed on-line.

Communication:

Most communication regarding the course will be done electronically. Please read your electronic mail (e-mail) regularly; the elm command or others can be used to do this. Announcements will be posted on the newsgroup wright.cs.cs772 which you should also read regularly; if your browser is not configured to use this link, the rn command can be used to read the group. The gnus command within emacs can also.

You may get in touch with the instructor by sending e-mail to cs772fac@cs.wright.edu. You are encouraged to send any questions about the assignments to the wright.cs.cs772 newsgroup so that everyone can benefit from the questions and answers.