One drawback for current search engines is that they
usually sort documents based on the popularity of the documents,
therefore they are not suitable to the applications that require looking
for new, interesting and unique information which are not popular, for
example, the anomaly detection application. Information mining, which typically uncover trends in activity,
links, and hidden patterns/relationships is an integral part of knowledge
management. One of drawbacks in current information mining is that the mining results
are not used in the search, which is a most-frequently used form of
knowledge application .
Knowledge Pattern Search (KPS) architecture combines search and mining:
1) Indexes embedded in collaborative agents are generated from learning
and mining knowledge patterns from historical data; 2) Using semantic
machine understanding makes it possible to search for new, interesting
and unique information rather than popular information.