1. Class attendance and completion of the readings by the assigned date are essential. This is not a straight lecture class - you will be expected to be able to discuss the articles. To this end:
2. At least one analytical question for each of the readings must be mailed to the instructors by 5pm on each Wednesday afternoon. These questions will form the basis for discussion.
3. Each student will lead the discussion of an article once or twice during the semester. Leading the dicussion means
• preparing a handout summarizing the article. This handout is to be sent to the class mailing list by Monday night at midnight -- in order to serve as a reading guide for the other members of the class.The summary should indicate and explicate the important parts of the article, the problems with the theory espoused or system explained, and the relevance for previous research discussed in class; and
• delivering an in-class  presentation consisting of:

(a)  an existing conversational/dialogue interactive system that deals with the discourse phenomenon under discussion.  The presentation should include a short description of the system architecture and a discussion of how well or poorly the particular discourse phenomenon in question is treated, and how it might be extended.
AND / OR
(b)  your own design sketch of a computational system or application that depends on or exemplifies the phenomenon treated in that week's readings, and leading a discussion in class on that computational system.  A design sketch should present a system of the student's own invention that incorporates the phenomenon under discussion into its functioning. The system can be as fantastical as you desire (and may incorporate *some* elements that are impossible given the state of research today), but the part of the system that uses the class phenomenon should be down-to-earth, possible, and clear. The goal of this exercise is to help the class understand the utility of discourse phenomena for interactive systems, and to understand in a concrete way how to incorporate them into system design.
4. Students will also take turns leading class discussion on the homework assignments (probably in pairs).
5. The final assignment will consist of two parts: 
1. a working interactive system, or good prototype for a system, or addition to an existent system, that deals in some way with a discourse phenomenon. That is, for example, you might build an interactive system that uses understanding of discourse markers to predict the relationship between two clauses. You must give a demonstration of your system in operation on the final day of class; 
2. a write-up of the system in ACL paper format (http://www.cis.udel.edu/%7Eacl2k/style/MSWord_template.rtf). 

Design sketches and homework assignments may be profitably used to lead up to the final project.
Students will come up with the idea for their final project by the 8th week of the semester and will turn in a proposal at that time. Final project presentations will take place during the last class session.
6. Because we expect that you will be interested in different discourse types (conversational stories, conversations between humans and animals, dialogues during Quake playing . . .) we will be asking you to collect data of your favorite type. But, because collecting and analyzing raw data is very time consuming, we will also be basing assignments on pre-collected data. You may always substitute your own data for pre-collected data in assignments, with prior consent of the instructors.

Some of the course readings will be available electronically, and linked to the syllabus.   Other readings will be passed out a week in advance.

Various computational linguistics tools will be made available to students, as indicated in the syllabus. Use of these tools will be overseen by Tim (bickmore@media.mit.edu).