| | |
Summary: Using Star Clusters for Filtering
Javed Aslam Katya Pelekhov Daniela Rus
Department of Computer Science
Dartmouth College
Hanover, NH 03755
Abstract
We examine applications of clustering to the filtering
task. We use the on-line version of the star algorithm
[JPR98, JPR99] as the clustering tool because this algo-
rithm computes, with high precision, naturally occuring
topics in a collection and it admits an efficient on-line so-
lution for dynamic corpora. We describe several filtering
algorithms and show extensive experimental results using
the TREC collection.
1 Introduction
Our goal is to automate information access and orga-
nization by topic. In our previous work [JPR99, JPR98],
we presented an efficient algorithm for organizing static
and dynamic information by topic using the star cluster
algorithm. We do not impose the number of topics for the
|
