.ll 6.5i
.po 1.0i
.sz 12
.nr pp 12
.nh
.sp
UNIVERSITY OF VIRGINIA
.br
  SCHOOL OF ENGINEERING AND APPLIED SCIENCE
.br
    DEPARTMENT OF COMPUTER SCIENCE
.sp 2
.ce
\*(td
.sp 4
.ta 1.0i
MEMO TO: Faculty and Students
.sp 
FROM: James French
.sp
RE: Master's Project Presentation by Travis Emmitt
.sp 2
All faculty and students are cordially invited to attend
the Master's Project presentation by Travis Emmitt
to be held on Monday, 9 August 1999 at 2:00 PM in room 236-D. 
The committee members are James French and John Pfaltz.
.sp 2
.b
.ce
Cue-Validity Variance Database Selection Algorithm Enhancement
.sp 
.ce
ABSTRACT
.r
.sp
The Information Retrieval group has been executing and evaluating
different database selection algorithms within a controlled test
environment.  One of the more recent algorithms under investigation
is CVV, developed by Budi Yuwono (Ohio State University) and Dik Lee
(Hong Kong University of Science and Technology).  This algorithm
consists of two components: cue validity variance (CVV) and document
frequency (DF).  Despite the relatively poor performance of the
overall algorithm within our test environment, the CVV component was
unique enough to merit further study.
 
We modified the CVV algorithm by including two additional components:
query term weight (QTW) and inverse collection frequency (ICF), both
of which had been shown to be effective in other selection algorithms.
By varying the degree of contribution of each of the four components,
we were able to discover CVV variants that performed well overall,
while at the same time learning about the overall usefulness of each
of the components.