Automating the creation of information filters

User Interfaces-user interface management systems (UIMS) SHOSHANA LOEB has been a member of technical staff in the information networking lab at Bellcore since 1989. Her research centers on uncovering the interplay between network and application architectures as it applies to the design of open networks and appealing multimedia application prototypes. Before joining Bellcore she was a faculty member in Permission to copy without fee all or part of this material is granted provided that the copies are not made or distributed for direct commercial advantage, the ACM copyright notice and the title of the publication and its date appear, and notice is give that copying is by permission of the Association for Computing Machinery. To copy otherwise, or to republish, requires a fee and/or specific permission. ne of the big Problems in using filters to deal with large information spaces is the effort Involved in creating, maintaining, and evolving filters over time. The effort involved can cause enough overload to make filter management as much trouble as the overload Problems filters were intended to solve. In order to address this situation the INFOSCOPE [1] system employs rule-based agents that recognize a user's usage patterns and make suggestions based on them. These suggestions help users create and maintain their own sets of filters. Agents keep a constantly evolving user model of individual interests. AS users read Usenet news messages, data about their interactions is stored in a knowledge base. Agents use rules to measure and record interesting terms found in read and deleted mail, as well as rules for tracking the timeliness of those measurements. Terms are collected from header fields and meaningless terms (like if, and, was...) are eliminated. Frequency and recency might be determined by ensuring terms are read in 30% or more of the last six sessions, or 50% of the last 10 sessions. TheSe tests are fully adjustable by individual users and therefore the rules might be different for each user. A pattern that is frequently SUpported by additional readings of matching Items will continue to trigger a suggestion, while patterns that are less frequently supported, or have been allowed to sit without action will be removed from the suggestion queue. Using this approach agents make suggestions that are completed filters based on observed reading patterns. They are displayed in a graphi-cal dialog box with editable text fields for altering the terms to be inCluded In or …