A significant problem in many information filtering systems is the dependence on the user for the creation and maintenance of a user profile, which describes the user's interests. NewsWeeder is a netnews-filtering system that addresses this problem by letting the user rate his or her interest level for each article being read (1-5), and then learning a user profile based on these ratings. This paper describes how NewsWeeder accomplishes this task, and examines the alternative learning methods used. The results show that a learning algorithm based on the Minimum Description Length (MDL) principle was able to raise the percentage of interesting articles to be shown to users from 14% to 52% on average. Further, this performance significantly outperformed (by 21%) one of the most successful techniques in Information Retrieval (IR), term-frequency/inverse-document-frequency (tf-idf) weighting.

NewsWeeder: learning to filter netnews

An apparatus, method, and computer program product for information filtering in a computer system receiving a data stream from a computer network, the data stream having raw informons embedded therein, at least one of the raw informons being of interest to a user, the user being a member client of a community. The method includes the steps of providing a dynamic informon characterization having profiles encoded therein, including an adaptive content profile and an adaptive collaboration profile; adaptively filtering the raw informons responsive to the dynamic informon characterization, and producing a proposed informon; presenting the proposed informon to the user; receiving a feedback profile from the user, responsive to the proposed informon; adapting the adaptive content profile, the adaptive collaboration profile, or both responsive to the feedback profile; and updating the dynamic informon characterization responsive to the previous adapting step. The apparatus includes a plurality of processors for providing interactive, distributed filtering of information, extracted from a computer network data stream in response to multiple attribute profiles.

Information system and method for filtering a massive flow of information entities to meet user information classification needs

A computer implemented method on a computer readable media is provided for determining the source of a problem in a complex system of managed components based upon symptoms. The problem source identification process is split into different activities. Explicit configuration non-specific representations of types of managed components, their problems, symptoms and the relations along which the problems or symptoms propagate are created that can be manipulated by executable computer code. A data structure is produced for determining the source of a problem by combining one or more of the representations based on information of specific instances of managed components in the system. Computer code is then executed which uses the data structure to determine the source of the problem from one or more symptoms.

Apparatus and method for event correlation and problem reporting

A system for selecting options for modifying a network architecture in accordance with user preferences using a knowledge base and a database. The knowledge base includes rules to select modification categories and the database includes information on a plurality of equipment and a plurality of tariffs. To select the options, the system generates a base-line simulation from the network architecture, which base-line simulation includes performance data on the network architecture under a predetermined scenario. Modification categories based on the user preferences are selected using rules in the knowledge base, and equipment information and tariff information corresponding to the network modification categories are selected from the database to create modification options for the network modification categories. The modification options are evaluated using rules in the knowledge base. The system then generates a plurality of new simulations of the network architecture as modified by the modification options, and compares the performance data of the base-line simulation with performance data of the new simulations to specify differences in performance data between the base-line simulation and each new simulation. Finally, the modification options (each corresponding to one of the new simulations) are ranked using the specified differences in performance data between the base-line simulation and each new simulation, user preferences, and rules of the knowledge base.

Automated network simulation and optimization system

Our research examines the role of prior knowledge in learning about really new products that make existing product knowledge obsolete. Those with higher prior knowledge incorrectly generalize from their knowledge of existing products in the same product market and assume that they already know most of what is necessary to use the new product properly. Three studies demonstrate that, compared to consumers with lower prior knowledge, those with higher prior knowledge learn less about a new product. Further, higher knowledge consumers are able to learn better but learn less due to motivational deficits. Study 1 shows that, when given a cue that the new product is truly different, those with higher prior knowledge learned more than those with lower prior knowledge. In Study 2, we demonstrate that inferior learning of new product information by those with higher prior knowledge is caused by inattention at encoding rather than to reconstructive errors at retrieval. When incentives for learning were provided after encoding but prior to retrieval, those with higher prior knowledge learned less than those with lower prior knowledge. The reverse was true when incentives were provided prior to encoding. We show that these results hold both when prior knowledge is manipulated experimentally (Studies 1 and 2) and when it is an individual difference factor (Study 3).

https://papers.ssrn.com/sol3/Delivery.cfm/SSRN_ID339383_code021011410.pdf?abstractid=339383&mirid=1&type=2

Prior Knowledge and Complacency in New Product Learning

A model-based reasoning system for diagnosing faults in a computer network has an information base that stores information about the computer network. This information includes the topology of the network, a class hierarchy containing a component library and diagnostic expertise, and a diagnosis episode. A diagnosis engine receives a pointer to a situation. Each situation includes a focus role and a most general class for the focus role. Based upon this situation, the diagnosis engine creates a diagnosis episode, using the information stored in the information base.

Model based reasoning system for network fault diagnosis

Recent attempts to develop larger and more complex knowledge-based systems have revealed the shortcomings and problems of centralized, single-agent architectures and have acted as a springboard for research in distributed AI (DAI). Although initial research efforts in DAI concentrated on issues relating to homogeneous systems (that is, systems using agents of a similar type or with similar knowledge), there is now increasing interest in systems comprised of heterogeneous components. The workshop on cooperation among heterogeneous intelligent agents, held July 15 during the 1991 National Conference on Artificial Intelligence, was organized by Evangelos Simoudis, Mark Adler, Michael Huhns, and Edmund Durfee. It was designed to bring together researchers and practitioners who are studying how to enable a heterogeneous collection of independent intelligent systems to cooperate in solving problems that require their combined abilities.

/pdf/aaaii-workshop-on-cooperation-among-heterogeneous-1ottkpe59p.pdf

AAAII workshop on cooperation among heterogeneous intelligent agents

Expert systems generally contain the knowledge of human domain experts in the form of heuristic rules. The heuristics encapsulate the rules of thumb that the experts have gained through extensive experience in the field. These experience-proven heuristics are used by the experts as a major part of their problem-solving approach. When confronted with rare and novel problems to which the heuristics may not apply, human experts can apply, in addition, their fundamental understanding—or “deep knowledge”—of the domain. In GTE’s COMPASS and PROPHET expert systems, we have elicited and used expert knowledge that is at an intermediate level, lying between knowledge based on direct experience and knowledge based on fundamental domain principles. We call this intermediate level expertise knowledge. Use of this type of knowledge allowed us to extend the range of both the resulting systems beyond the heuristics gained from the experts’ field experience.

Eliciting and using experiential knowledge and general expertise

The Digital Services Research Group and its predecessor groups and offshoots in Digital Equipment Corporation have been mobilizing leading-edge AI research to bear on real-life problems that face the corporation and its customers. The general strategy of the group is to explore emerging techniques relevant to service and support needs through developing rapid prototypes, deploying these prototypes, and incorporating feedback from users. With over 32 major projects undertaken during the past decade, we have worked on broad spectrum of problems and explored a variety of advanced AI techniques. This article describes the current AI activities in five areas: (1) enterprise advisory systems, (2) natural language processing and textual information retrieval, (3) largescale knowledge base management and access, (4) software configuration management, and (5) intrusion detection. We also list some future research directions.

Mark Ronald Adler

Papers

Model based reasoning system for network fault diagnosis

AAAII workshop on cooperation among heterogeneous intelligent agents

Eliciting and using experiential knowledge and general expertise

AI research and applications in Digital's service organization