Data processing: artificial intelligence – Knowledge processing system – Knowledge representation and reasoning technique
Reexamination Certificate
1999-07-21
2002-11-19
Black, Thomas (Department: 2121)
Data processing: artificial intelligence
Knowledge processing system
Knowledge representation and reasoning technique
C706S045000, C709S202000
Reexamination Certificate
active
06484155
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates to knowledge computing systems, and more particularly relates to large-scale distributed knowledge systems.
BACKGROUND OF THE INVENTION
Organizations managing complex operations such as military command and control, medical diagnosis and treatment, and enterprise management have turned to large-scale knowledge systems as a means to effectively capture and maintain their knowledge assets. Those assets are being deployed across distributed computing resources including intranets and the Internet.
The technology supporting such systems has progressed to the point of being able to capture and use knowledge on the order of thousands of knowledge units (algorithms/rules/frames/axioms) to support human-like problem solving. However, until now there has not existed a knowledge management system able to support dynamic distributed problem solving over multiple large-scale and distributed knowledge systems. Such a system must be capable of manipulating and coordinating hundreds of thousands to millions of knowledge units.
One reason for that shortcoming is that typical systems, such as knowledge-based systems, employ inference engines that operate only on knowledge resident in specific knowledge bases that they control directly. For example, database search engines have the capability to manage information finding and retrieval, but they are not capable of retrieving knowledge in general form or combining knowledge retrieved from multiple sources into integrated inference patterns. In contrast, humans perform inference processes that involve widely distributed knowledge of more than one representation type or domain.
In addition, existing systems for performing distributed problem-solving functions typically require human reasoning intervention to bridge gaps in inference sequences, particularly between domain-specific and type-specific automated knowledge processing subsequences. In cases where the human intervention occurs, it usually takes the form of case-specific procedures implemented in software or hardware interfaces.
One example of an existing knowledge processing system is described in U.S. Pat. No. 5,465,319 issued to Ahamed, assigned to AT&T Corp., and entitled “Architecture for a Computer System Used for Processing Knowledge.” Ahamed describes a knowledge machine for processing information and evolving knowledge. However, that system is limited because the knowledge stored must be organized in fixed links and in predetermined hierarchies, prior to any query of the knowledge. Moreover, the system described by Ahamed does not incorporate distributed inferencing that combines results of multiples of distinct, as well as overlapping and conflicting knowledge modules. In addition, the system of Ahamed requires centralized control of the allocation of knowledge processing hardware resources and of the execution of the knowledge representation of those hardware resources.
Another example of an existing knowledge processing system is described in U.S. Pat. No. 5,628,011 issued to Ahamed et al., assigned to AT&T Corp., and entitled “Network-Based Intelligent Information-Sourcing Arrangement.” Ahamed et al. describe a system in which knowledge bases are arranged hierarchically in a “knowledge ring.” Queries are collected and passed to the knowledge ring. The system restates the question in a way that the knowledge base is integrated, and, upon confirmation, responds to the query. However, that system is limited to the use of static knowledge trees. Any query must match an existing entry in the knowledge dictionary. The system of Ahamed et al. basically locates and retrieves pre-existing data by a search of the static knowledge tree, which is essentially a taxonomy of hierarchically-related subject names.
Yet another example of an existing knowledge processing system is described in U.S. Pat. No. 5,809,493 issued to Ahamed et al., assigned to Lucent Technologies, Inc. entitled “Knowledge Processing System Employing Confidence Levels.” Ahamed et al. describe a knowledge processing system which processes knowledge in the knowledge domain to generate incremental and integrated conclusions, and in the numerical domain to generate confidence levels. The knowledge processing system iteratively revises solutions to generate an optimal solution based on the confidence levels. However, that system fails to describe dynamic problem solving, but rather builds confidence in pre-existing problem solutions by building databases of previous attempts to solve the same problem. The described system uses pre-determined (non-dynamic) knowledge operations mappings.
Accordingly, a need exists for a knowledge management system for dynamic, distributed problem-solving systems.
SUMMARY OF THE INVENTION
The present invention overcomes the problems identified above by providing a knowledge management system that supports inquiries of distributed knowledge resources. Those inquiries may be in the form of questions or problem statements presented by a user. Interaction between a user and the knowledge resources is mediated by a collection of cooperative intelligent agents.
The cooperative intelligent agents incorporate generalized automated negotiation and distributed inference (i.e., problem-solving) processes. Using those processes in a hierarchical architecture, the invention analyzes input problem statements and organizes the problem statements as sets of tasks. In pursuit of each task, the invention solicits accessible knowledge repositories, represented by knowledge agents, for relevant knowledge, and then analyzes and integrates responses from those knowledge repositories. The invention may then provide the responses to a human user or a using process.
The invention adaptively and dynamically synthesizes problem-specific knowledge interfaces and reasoning procedures as the problem-solving process moves forward. The invention extends automated inference capability to make use of a large number of knowledge sources of different types, in different locations, and covering different domains of expertise.
In one aspect, the invention integrates established knowledge-based environments or other software-based knowledge by breaking them down into well-defined, independent “knowledge modules.” The knowledge modules contain knowledge and knowledge processing capabilities (e.g., inferencing, database management, algorithms, etc.) for a domain of knowledge. Knowledge modules are independent and may contain overlapping or exclusive knowledge. Each knowledge module is combined with a cooperative intelligent agent to form a unit that facilitates integration with other units. The units allow the system to support cooperative reasoning through distributed inference (and problem-solving) processing. The encapsulation of restricted knowledge content within the units and their well-defined interfaces aids verification. The resulting architecture supports the development of a continuously evolving, distributed knowledge system.
More specifically, the present invention provides a hierarchical knowledge management system having three general layers: a user interface layer, a meta agent layer, and a knowledge agent layer. Each layer in the system includes one or more intelligent agents responsible for one portion of the distributed problem-solving inferencing process. The user interface layer mediates the direct interactions with the user, which affords control of and a window into the system's activities. The meta agent layer analyzes user queries or problem formulations from the user interface layer, allocates tasks to the knowledge agent layer, resolves conflicts arising from the knowledge agent layer, and consolidates (including fusing and deconflicting) results provided by the knowledge agent layer. The knowledge agent layer provides an interaction mechanism for knowledge modules having associated knowledge agents within the knowledge agent layer. Each agent in the system includes inter-agent abstract communications facilities with the capability to negotiate with
Daniel, III Robert S.
Kiss Peter A.
Yalowitz Jeffrey S.
Booker Kelvin
Merchant & Gould
Sentar, Inc.
Whitaker John E.
LandOfFree
Knowledge management system for performing dynamic... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Knowledge management system for performing dynamic..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Knowledge management system for performing dynamic... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2937823