204596 : Special Topics in Information Retrieval

Second Semester 1998

Instructor: Dr. Arnon Rungaswang, fenganr@ku.ac.th, CPE room 1504

Class times: Wednesday evening, 18:00-21:00, CPE room 1102

Class size : minimun 12, maximum 16

Office Hours: Tuesday, 14:30-16:30, Wednesday, 9:30-11:30, or by appointment (send e-mail to fenganr@ku.ac.th).

Web's site of this course : http://www.cpe.ku.ac.th/~arnon/cpe204596.html

UPDATE: 25/1/99

- Current students

- Presentation schedule

- Presentation Slides

- IR Systems

- Local copy of TREC6's papers

- Local copy of UMASS presentation slides

Course Description

This is a seminar-style course which follows the 204553 "Information Storage and Retrieval" last semester, but is also run as a course for all other students who are interested in registering. During this course, we cover current and recent research in Information Retrieval (IR). We discuss how this research relates (or does not relate) to our lab's IR research. We try to include such specific topics as, visualization, passage handling, routing (feedback), query expansion, summarization, integration with databases, natural language processing, and so on.

If you attend, you are expected to participate in all seminars. Most importantly, that means reading the "assigned" papers before the seminar. You are also expected to raise questions and join in the discussion. In addition, each group of 3-4 students will be assigned a mini-project concerning real IR applications.

Seminar format

Sessions will take several forms:

Often, participants will be expected to read 1-3 papers (rarely more than 2) and "volunteers" will be responsible for presenting a 15-20 minute summary of the key points of each paper. Discussion of the papers will follow.
The sessions will also occasionally be devoted to more lecture-style seminars, where a paper or topic will be presented without having been read by the group in advance. Discussion of the material will follow.
Some sessions will include several small presentations of interesting IR-related developments, usually something found on the World Wide Web.
Guest lectures from other department, and outside KU will be invited to come to our seminar.

Dr. Asadaporn Nuchaprayoon, Visualization in IR (U. of Pittburgh)
Dr. Chuleerat J., Thai Text Retrieval (U. of George Washington)
...

When possible, guest lectures will be asked to relate their work to the papers.

Getting up to speed

If your background in IR is rusty, you should consider some background reading. Gerard Salton's 1989 book Automatic Text Processing (Addison-Wesley) is a reasonable starting point.

There is an excellent new book out, edited by Karen Sparck Jones and Peter Willett, called Readings in Information Retrieval (published by Morgan Kaufmann). It includes many important papers in IR. It is not a textbook, but is useful for anyone with strong interests in the field. If you interest to take a look at it, please let me know. But if you would like to process your own one, take a look at www.amazon.com.

Another source of online IR papers can be found at TREC site. Normally, papers distributed to the students will come from these TREC proceedings, and the above Readings book of Karen Sparck Jones. You are encouraged to take a look at TREC site now. Notice that we have a local copy of TREC6's papers here.

Grading Policy

Registered students are required to work for their grade as described below. I will attempt to monitor your progress, but ultimately it is your responsibility to ensure that you have satisfied all the requirements. Failing to do so will necessarily result in a lower grade. You are expected to:

Each student must present a minimum of one full paper during the semester. Sharing a presentation with someone does not count as full paper (unless it's a really long or complex paper). Papers that are short count as less than a full paper. You should expect to do the equivalent of presenting a full paper, even if you do it in bits and pieces.
Everyone should provide a brief summary of each paper for every meeting. The summary should be e-mailed to me before the seminar has started, preferably well before. A sample list of questions that you should address is available here. Summaries for papers that are unannounced are obviously not expected.
Each group must present a brief 15 minutes concerning the mini-project progress every two weeks. A brief summary what you present should be e-mailed me before the seminar has started, as well.
Group presentation of the mini-project is expected before 3 weeks counting backward from the last day of this course.
No mid-term, and no final examination.

Being a scribe

We will try to keep a record of what happened in all of the IR seminars on-line. This record will include a very brief outline of the papers as well as a summary of the discussion that followed.

The designated scribe for a seminar will write up an HTML page that includes the following:

List of papers presented during the seminar, who presented them, and a very brief summary of the papers. This brief summary should be a sentence or three at most. Note that sometimes "papers" will be a lecture, and sometimes they'll be very brief presentations of cool things on the web. Same idea for those.
The group's opinion of the paper, including its major strengths and weaknesses. Of course, if the group doesn't offer an opinion, there won't be much to write here. Opinions from the past have included things like "not worth the ink it's printed with". (As the scribe, you obviously have some latitude for editorial adjustments, but please be sure the discussion is represented accurately.)
A summary of any discussion that follows, proceeds, or occurs during the presentation(s).
Any other comments that you think are pertinent.
Each group will be assigned to update this HTML page, one group per week.

Include as many URL's as you can easily acquire. For example, pointers to someone's web page, pointers to project web pages, and so on.

These notes should be written up as quickly as possible so that they can be put on the web within 2 days of the seminar (sooner if possible). E-mail the notes or a pointer to them to me (fenganr@ku.ac.th) once they're done.

NOTE: What you write will be placed on the web so it is publicly readable by everyone in the whole wide world who is on the web (that's why it's called the World Wide Web). If for some reason that embarrasses or offends you (it shouldn't), please talk to me. You are welcome to leave your name off of the notes if it would make you feel more comfortable.