Information Technology Reference
In-Depth Information
Figure 2. Collaborative search system
and make your judgments about them. The system
accumulates these into a description of your quest
and your judgments, and adaptively develops a
model by which it computes similarity between
your quest and stored profiles of quests that have
been conducted earlier by other people. When it
identifies similar quests, it then looks at materials
that had been found useful by the owners of those
quests, and computes some kind of consensus,
judgment or ranking. The system then provides an
“Ant-link” which carries you directly from your
present search point to those materials that had
been judged relevant by the other users. Different
from other recommendation systems, which use
implicit indicators of relevance (topic similarity
of old search sessions), AntWorld needs users' ac-
tive contribution about the relevance of retrieved
results, which may put extra requirement on the
user side.
However, we don't believe that should be a
concern based on two observations. First, it will
not be a problem in any system targeting an in-
tranet environment. In such situation, searching
and judging is part of the information seekers job.
Second, we believe that with the development of
social networks, the new generation of web users
may be more willing to contribute.
Add to the old model of AntWorld is the learn-
ing algorithms. These learning algorithms work
with your quest and judgments, and your responses
to suggestions made. These are integrated with
the present models of content aspects and quality
aspects, to improve the rule by which stored quests
are matched to your present quest. This will involve
looking at the retrieved web pages from your quest
in much more sophisticated ways than was done
in the AntWorld System (which simply used term
frequencies). Details are discussed below.
The key is that three things will change: we
will move to a larger class of representation meth-
ods; we will consider a large range of matching
methods and, most significantly, we will work
Search WWH ::
Custom Search