Multi-Perspective Retrieval, Ranking and Visualization of Web Data

We propose a new system to present information in
a database from different perspectives and draw
correlations between the perspectives. Although the
tool may be used to understand the contents of any
database, our studies focus on its application to
retrieve, rank and cluster Web documents. The major
bottleneck in the development of our system is the
speed of retrieval and ranking of Web pages since
several different sets of queries (each of which
represent different perspectives) must be processed
in parallel. We developed a small package to quickly
compute the rankings of Web pages for a set of
input queries. The fast speed of the computation
demonstrates that implementation of parallel
rankings using up-to-date information on changes in
Web sites is possible. Our data are Web pages from a
Japanese newspaper, which consists of Japanese and
English text. This is a preliminary report on the
development of our system.