Final Project

Due Wed, March 16th, 10:30am

The main objectives for the final project are (a) to give you the chance to explore some issues in KR in depth, and (b) give you a flavor for real research projects. As I mentioned in class, one goal to possibly aim for is to produce something like an AMIA symposium submission (which is conveniently also due on March 16th).

These projects may be carried out in teams of up to two students. I plan to allot both the 80 minutes of class on Thursday morning, March 10th, and the official two hour slot on March 16th for oral presentations. Ideally, we might have 10 distinct projects over the 15 students enrolled, which would give each team 20 minutes to present; more realistically, anything between 8 and 12 teams would be fine. For those that give oral presentations early (on the 10th), the written report is still not due until the 16th.

Types of projects :

For this project, it is insufficient to simply carry out a literature review in some topic. Instead, I will demand that you do something concrete or hands-on. What is is that you do depends on the sort of project you tackle. Here are three broad classes of projects:

• Evalation study: Carry out an experiment that evaluates some aspect of a known knowledge base or KB system(s). One interesting way to do this is to compare and contrast two systems. Which is better? Another way to think about this type of project is to pick two points along the expressivity / tractability tradeoff line, and experimentally demonstrate that tractability decreases as expressivity increases. Obviously, the scope of your experiment must be small. However, this sort of project requires minimal programming, so I will expect more in the way of quantitative results.
• Rational reconstruction: Based on descriptions in the literature, re-implement soime system in biomedical informatics. Your re-implementation need not be completely faithful to the original design (it could be improved!). Of course, this must have a Knowledge Base emphasis, but most systems necessarily include some sort of knowledge store. For reconstruction projects, your deliverable will simply be proof that your system works, hopefully in a similar way as the original system. In some cases, you might improve on an existing system that doesn't emphasis good KB design for sharability or for inferential capability.
• Application development project: Build your own system that uses a knowledge base to carry out some interesting inferential or query-answering capability in a biomedical or public health domain. In the space of 6-8 weeks, I'm not expecting ground-breaking research; but there are many interesting combinations of rule-based systems, description logics, or frame-based systems that have not been applied to biomedical and public health domains.

If you would like some specific examples of ideas for the sort of thing I would like to see for your final project please feel free to contact me. I can easily come up with a number of interesting projects.

Schedule of checkpoints:

(See also the assignments page)

1. Initial team selection, rough topic idea (Tues, week 4).
2. A one-page plan for the remainder of the quarter, including 1-2 paragraphs describing the type of project, as well as your overall goals and expectations for results. If a team project, you should include an explicit division of labor for the tasks listed in your plan. (Thur, week 6)
3. A progress report, up to 2 pages in length. May include a draft of the abstract of your written report, preliminary results, or direct questions to me about how to proceed in light of your initial endeavors. If a team project, the progress report should be individualized, indicated which team member has accomplished what so far. (Thur, week 8)

These checkpoints are not graded; they are primarily for your benefit, not mine. However, grading is always a subjective process. If you are late, or do a weak job on your checkpoints, this will give me a bad initial impression as to the content and thoughtfulness of your project.

Final Deliverables:

1. An oral presentation of your work. You will have ~15-20 minutes to present your results either during finals exam block (March 16th), or on the last day of class (March 10th). You should let me know if you have a strong preference one way or another. Once the number of groups are set, I can say more about when and how long the presentations will be. For team projects, it is fine to have only one person give the oral presentation.
2. For team projects, you must provide a document that specifies the division of labor between team members. This can be a jointly signed paper document, or a pair of individual emails (saying the same thing, hopefully!).
3. A written report that describes your work. This must be between 7-10 pages (1-inch margins, single-spaced, 12 point font text and figures). In most cases, the report should include at least some citations and discussion of relevant, related work. The written report should also include all "results", whether these be screenshots from your system or comparison tables, or demonstrations of working behavior in particular domains. Extensive system trace printouts or similarly lengthy details should be relegated to an appendix (which does not count against the 10 page limit.)

Grading criteria:

I see three components to grading these projects; (1) An assesment of the oral presentation, perhaps covering 30% of your grade, (2) An assessment of the written report, covering perhaps 40% of your grade, and (3) An overall assessment of the quality and content of your work, which is reflected in both the oral presentation and the written report (the remaining 30% or so). Note that I will not look closely at nor grade any code products. Your "results" should all be included in your written report (or appendixes), so those certainly are graded.

Detailed assessment of written report (but note that most of the content topics also apply to the oral presentation):