The Effects of a "Time-out" Period for a
                Distributed Information Filtering System 

                      Michael Boyles & Nila Patel
         Faculty Mentors: Rajeev Raje & Snehasis Mukhopadhyay



Department of Computer & Information Science, 
Indiana University Purdue University Indianapolis
723 West Michigan Street SL280, Indianapolis IN 46202



Information filtering is a technique used to classify, sort, and present 
documents according to a particular user's interest. Due to the vast 
amount of information accessible to a single user, an isolated filtering 
system may not be able to classify all incoming documents. 

Therefore, we have developed a distributed environment for information 
filtering, called D-SIFTER. By allowing many filters to collaborate with 
each other, we have managed to increase the number of successful 
classifications for a single user. D-SIFTER is currently implemented in 
the Java language and its Remote Method Invocation package. 

Our most recent research involved finding an optimal "time-out" period. 
That is, how long should one filter wait for the results of others? We 
conducted many experiments with D-SIFTER to examine this question. We 
concluded that once the optimal time-out was achieved, a further increase 
in it had no effect on the number of successful classifications.