Abstract:
|
In this paper, we present a fully distributed and collaborative search
engine for web pages: Porqpine. This system uses a novel query-based model
and collaborative filtering techniques in order to obtain user-customized
results. All knowledge about users and profiles is stored in each user
node?s application. Overall the system is a multi-agent system that runs on
the computers of the user community. The nodes interact in a peer-to-peer
fashion in order to create a real distributed search engine where
information is completely distributed among all the nodes in the network.
Moreover, the system preserves the privacy of user queries and results by
maintaining the anonymity of the queries? consumers and results? producers.
The knowledge required by the system to work is implicitly caught through
the monitoring of users actions, not only within the system?s interface but
also within one of the most popular web browsers. Thus, users are not
required to explicitly feed knowledge about their interests into the system
since this process is done automatically. In this manner, users obtain the
benefits of a personalized search engine just by installing the application
on their computer. Porqpine does not intend to shun completely conventional
centralized search engines but to complement them by issuing more accurate
and personalized results. |