Expert Systems with Multiple Experts

Prof. Vagan Terziyan

University of Jyvaskyla, FINLAND

Kharkov State Technical University of Radioelectronics, UKRAINE

The multiple experts' knowledge base refinement is one of central problems in expert systems design. The group of voting-type methods is intended to derive the most supported opinion among experts. The different modifications of voting-type methods can be explained within a context of an appropriate problem. The context analysis appears to be quite important in managing multiple methods in knowledge acquisition.

The main goals of the lectures’ series are to describe problems taking place in knowledge acquisition with multiple knowledge sources, based on voting-type methods, knowledge representation with multilevel knowledge structures, context-sensitive reasoning and their use in some expert systems applications.

Contents

Expert Systems with Multiple Experts …..………………….….. 3

Topic 1. A Cognition Using Semantic Balance and

Refinement of Internal and External Knowledge ……… 4

Topic 2. Reasoning with Multilevel Contexts

in Semantic Metanetwork ……………………………... 21

Topic 3. Handling the Multiple Expert Knowledge

Based on the Most Supported Opinions ………………. 61

Topic 4. The Voting-Type Technique to Handle

the Multiple Expert Knowledge ………………………. 89

Topic 5. Temporal Knowledge Acquisition

from Multiple Experts ……………………………………. 148

Topic 6. Handling Interval Estimations

from Multiple Experts ………………………………….201

Author’s Contact Address …………………………..……….…..226

 

Expert Systems with Multiple Experts

(Main questions to be answered)

 

  1. What is knowledge within the context of multiple experts?
  2. How to acquire knowledge from multiple experts?
  3. How to represent knowledge taking into account context of multiple experts ?
  4. How to select the most supported knowledge from the multiple experts ?
  5. How to evaluate different experts and use this evaluation in knowledge refinement ?
  6. Which strategy to select to improve the environment cognition process ?
  7. How to use concept of semantic balance in the cognition process ?
  8. How to consider expert’s individuality as a context for reasoning with multiple experts knowledge ?
  9. How to handle inexact and incomplete knowledge obtained from different sources ?
  10. What are virtual experts and how they can be used to handle the multiple experts problems ?

 

A Cognition Using Semantic Balance and Refinement of Internal and External Knowledge

The goal of this topic is to study knowledge refinement method based on semantic balance of knowledge.

Reference: Grebenyuk V., Kaikova H., Terziyan V., Puuronen S., The Law of Semantic Balance and its Use in Modeling Possible Worlds, In: STeP-96 - Genes, Nets and Symbols, Publ. of the Finnish AI Society, Vaasa, Finland, 1996, pp. 97-103.

This topic presents a knowledge refinement strategy to handle incomplete knowledge during a cognition process.

The goal of this research is to develop formal tools that benefit the law of semantic balance. The assumption is used that a situation inside the object’s boundary in some world should be in balance with a situation outside it. It means that continuous cognition of an object aspires to a complete knowledge about it and knowledge about internal structure of the object will be in balance with knowledge about relationships of the object with other objects in its environment.

 

It is supposed that one way to discover incompleteness of knowledge about some object is to measure and compare knowledge about its internal and external structures in an environment.

 

Reasoning with Multilevel Contexts in Semantic Metanetwork

The goal of this topic is to study formal way to represent knowledge within several levels of contexts which can also been interpreted as levels of experts’ competence.

 

Reference: Terziyan V., Puuronen S., Multilevel Context Representation Using Semantic Metanetwork, In: Context-97 - International and Interdisciplinary Conference on Modeling and Using Context, Rio de Janeiro, Brazil, Febr. 4-6, 1997, pp. 21-32.

 

A multilevel semantic network is proposed to be used to represent knowledge within several levels of contexts. The zero level of representation is semantic network that includes knowledge about basic domain objects and their relations. The first level of presentation uses semantic network to represent contexts and their relationships. The second level presents relationships of metacontexts i.e. contexts of contexts, and so on at the higher levels. The topmost level includes knowledge which is considered to be "truth" in all the contexts. Such representation allows to reason with contexts towards solution of the following problems:

Handling the Multiple Expert Knowledge Based on the Most Supported Opinions

The goal of this topic is to study methods to handle knowledge elicited from multiple sources based on the most supported parts of this knowledge.

Reference: Puuronen S., Terziyan V., Modeling Consensus Knowledge from Multiple Sources Based on Semantics of Concepts, In: Albrecht M.& Thalheim B. (Eds.), Proceedings of the Workshop Challenges of Design, 15-th International Conference on Conceptual Modeling ER’96, Cottbus, Germany, October 1996, pp. 133-146.

An approach is considered to inference with knowledge obtained from multiple sources.

The sources are grouped into multilevel hierarchical structure, according to the type of knowledge obtained. The first level consists of experts who have knowledge about the basic objects and their relationships. The second level of experts includes those who have knowledge about the relationships of the experts at the first level and each higher level accordingly.

It would be shown how to derive the most supported opinion among the experts at each level. This is used to order the experts into categories of their competence defined as the support they get from their colleagues.

 

 

The Voting-Type Technique to Handle the Multiple Expert Knowledge

The goal of this topic is to study methods to handle knowledge elicited from multiple sources based on object-concept-source knowledge representation and voting technique of knowledge acquisition.

Reference: Puuronen, S., Terziyan, V., The Voting-type Technique in the Refinement of Multiple Expert Knowledge. In: Sprague, R. H., (Ed.), Proceedings of the Thirtieth Hawaii International Conference on System Sciences, Vol V, IEEE Computer Society Press, 1997, pp. 287-296.

Knowledge is represented using predicates that define relationships within three sets:

Sources express their comprehension of the use of each concept to describe each object by giving their votes: yes, no, and no-op.

We derive and interpret internal relations between any pair of subsets of the same type taken of the three sets: objects, concepts, and sources.

Intersections between the sets are interpreted as multilevel structures of knowledge and they are used for knowledge refinement.

The refinement technique presented is based on the derivation of the most supported opinion of the group of experts and refining it further using a multilevel structure of knowledge sources.

 

Temporal Knowledge Acquisition from Multiple Experts

The goal of this topic is to study a group of knowledge acquisition methods based on rank-navigation technique.

Reference: Kaikova H., Terziyan V., Temporal Knowledge Acquisition From Multiple Experts, In: Shoval P. & Silberschatz A. (Eds.), Proceedings of NGITS’97 - The Third International Workshop on Next Generation Information Technologies and Systems, Neve Ilan, Israel, June-30 - July 3, 1997, pp. 44 - 55.

The area of knowledge acquisition from multiple experts is considered. The reasonable consensus among experts is being achieved after several advances due to flexible rank-navigation technique. The example of the domain of Allen’s temporal relations is used to answer to the following questions:

 

 

Several possible strategies to provide the dynamic rank-navigation are presented and compared based on the example.

Handling Interval Estimations from Multiple Experts

The goal of this topic is to study the method to handle inexact knowledge represented by intervals acquired from multiple knowledge sources.

 

Reference: Terziyan V., Puuronen S., Kaikova H., A Decontextualization Method for Estimated Intervals, Submitted to: 7-th IPMU Conference - Information Processing and Management of Uncertainty in Knowledge-Based Systems’98, Paris, France, July 6-10, 1998.

This topic covers the context sensitive approach to handle interval knowledge acquired from multiple knowledge sources.

Each source gives its estimation of the value of some parameter x. The goal is to process all the intervals in a context of trends caused by some noise and derive resulting estimation that is more precise than the original ones and also takes into account the context noise.

The main assumption used is that if a knowledge source guarantees smaller measurement error (estimated interval is shorter), then this source in the same time is more resistant against the effect of noise. This assumption allows us to derive and process trends among intervals and end up to shorter resulting estimated interval than any of the original ones.