Architecting personalized delivery of multimedia information
Shoshana Loeb · Communications of the ACM · 1992
information filters are essential mediators between information sources and their users.In most cases, both the information sources and the information users possess no mutual knowledge that can guide them in finding the information most relevant for the users' momentary and long-term needs.Filters, which are positioned logically as "third parties" to the communication between users and sources, should possess both the knowledge and the functionality to examine the information in the sources and to forward the information "they judge" as relevant to individual users.Information filters can work on behalf of users as well as on behalf of sources.In the first case, which is the most common today, filters assist users in finding relevant information and overcoming what has been termed "the information flood."In the second case, filters can be used by sources to "target" information to potentially interested users.Currently it appears that many of the research issues involved in the proper design of high-performance filters are addressed only for a specific and relatively narrow class of sources and users.One reason for this is that it is exceedingly difficult at this time to provide general-purpose solutions to information-filtering requirements.In the general case, user needs do not seem to lend themselves to precise modeling [1], and information sources do not seem to provide precise enough descriptors of their contents (see [5] and references therein).Given this double imprecision, information filters are being optimized for specific context, specific types of users, and specific sources.In this article, as in [2], we view the information-filtering process as dependent on the application domain in which it operates and on the context in which it is used.We CONHUNI|ATIONII(iPTHBACH/