LBSC 708A/CMSC 838L
Information Retrieval Systems
Fall 2001
Term Project
Students registered for CMSC 838L will normally complete the
algorithms track and students registered for LBSC 708A will complete
the comparative evaluation track. A research publication track
is also available by special arrangemnt. Students wishing to choose
the publication track or to switch to a track not normally associated
with the course for which they registered should discuss their request
with the professor before the third class meeting.
Algorithms Track (CMSC 838L)
Students in the algorithms track will normally work individually,
although teams of students that wish to work together may propose more
ambitious projects. The key goal is to construct an information
retrieval system from scrach using algorithms of the student's choice.
Reusable components in the NIST IRF package may be used, but at least
one new algorithm must be developed by each student from first
principles. The following :
A written analysis of the proposed algorithm, including a
review of the relevant research literature and an analysis of time and
space complexity.
An implementation of the algorithm that is compatible with IRF.
This may be achieved either by coding in Java or by coding in C++ and
integrating with Java using JNI.
An oral presentation of evaluation results comparing the new
algorithm and the existing IRF algorithm using a standard test
collection.
Comparative Evaluation (LBSC 708A)
A term project will be completed by the end of the semester, typically
by teams of about three students. Each team will be asked to index a
standard text retrieval collection using a different text retrieval
system. Teams will perform a recall-precision effectiveness
evaluation, to measure both indexing and retrieval efficiency, and to
design and conduct a small user study of the retrieval system and the
associated user interface. Teams will present their results to the
class at the end of the semester and submit a written report.
Research Publication Track (By special arrangement)
In the research publication track, students will complete a research
project that that significant extends previous work and submit their
results for publication in a workshop, conference, or journal. The
research may address any topic that is important in an information
retrieval context. Examples might include:
Translation of proper names for cross-language retrieval
Support for redaction in electronic mail collections
Topic shift detection in recorded oral histories
Students wishing to choose the research publication track should
prepare a one-page proposal describing the research question that they
wish to explore, the method that they propose to use, and at least one
suitable venue at which the work could be presented. A literature
review will be due by the fifth class meeting and one-page status
reports will be due every three weeks thereafter.
Doug Oard
Last modified: Tue Aug 21 18:55:04 2001
