Comparing knowledge elicitation techniques: a case study

Three knowledge elicitation techniques were used to extract knowledge bases from experts on lighting for industrial inspection tasks. The techniques were: I a structured interview; II twenty questions —imputing rules from information requests; and III a card sort. The first two techniques generate protocols, and in these cases two knowledge engineers independently extracted production rules from the protocols. In the third technique rules were derived from the classification of lighting solutions. The first two techniques led to about the same number of rules, the card sort to less. There was a slightly higher percentage of agreement between the rules extracted separately by the knowledge engineers for the normal interview than for the twenty questions technique. Disagreements between the knowledge engineers were resolved by discussion, and an expert system to select special lights for inspection tasks was implemented. Of the rules finally agreed on, the percentage which could be implemented in the expert system was less for the twenty questions technique than for the others. The agreed rules were mailed to the expert, who indicated whether she agreed, disagreed, or wanted to make a modification. This secondary stage of elicitation revealed no evident difference between the three techniques in terms of the proportion of rules validated by the expert.     

Causal domain model driven knowledge acquisition for expert diagnosis system development

Despite the successful operation of expert diagnosis systems in various areas of human activity these systems still show several drawbacks. Expert diagnosis systems infer system faults from observable symptoms. These systems usually are based on production rules which reflect so called shallow knowledge of the problem domain. Though the explanation subsystem allows the program to explain its reasoning, deeper theoretical justifications of program's actions are usually needed. This may be one of the reasons why in recent years in knowledge engineering there has been a shift from rule-based systems to model-based systems. Model-based systems allow us to reason and to explain a system's physical structure, functions and behaviour, and thus, to achieve much better understanding of the system's operations, both in normal mode and under fault conditions. The domain knowledge captured in the knowledge base of the expert diagnosis system must include deep causal knowledge to ensure t he desired level of explanation. The objective of this paper is to develop a causal domain model driven approach to knowledge acquisition using an expert–acquisition system–knowledge base paradigm. The framework of structural modelling is used to execute systematic, partly formal model-based knowledge acquisition, the result of which is three structural models–one model of morphological structure and two kinds of models of functional structures. Hierarchy of frames are used for knowledge representation in topological knowledge base (TKB). A formal method to derive cause–consequence rules from the TKB is proposed. The set of cause–consequence rules reflects causal relationships between causes (faults) and sequences of consequences (changes of parameter values). The deep knowledge rule base consists of cause–consequence rules and provides better understanding of system's operation. This, in turn, gives the possibility to construct better explanation fa cilities for expert diagnosis system. The proposed method has been implemented in the automated structural modelling system ASMOS. The application areas of ASMOS are complex technical systems with physically heterogeneous elements.     

Model for knowledge and legal expert systems

This paper presents a four layer model for working with legal knowledge in expert systems. It distinguishes five sources of knowledge. Four contain basic legal knowledge found in published and unpublished sources. The fifth consists of legal metaknowledge. In the model the four basic legal knowledge sources are placed at the lowest level. The metaknowledge is placed at levels above the other four knowledge sources. The assumption is that the knowledge is represented only once. The use of metaknowledge at various levels should make it possible to use the appropriate knowledge for the problem presented to the system. The knowledge has to be represented as closely to the original format as possible for this purpose. Suitable representation formalisms for the various types of knowledge in the five knowledge sources are discussed. It is not possible to indicate a best representation formalism for each knowledge source.     

DETECTOR: A knowledge-based system for injection molding diagnostics

A knowledge-based system (KBS) for diagnosis of multiple defects in injection molding is presented. The general scheme for knowledge representation based on fuzzy set theory has been shown useful in representing inexact and incomplete information for developing the KBS. An optimality criterion is created for selecting a simple and best cover to explain the given problem. An efficient search algorithm for finding such cover is also discussed.     

Knowledge acquisition for fuzzy expert systems

Expert systems have been successfully applied to a wide variety of application domains. to achieve better performance, researchers have tried to employ fuzzy logic to the development of expert systems. However, as fuzzy rules and membership functions are difficult to define, most of the existing tools and environments for expert systems do not support fuzzy representation and reasoning. Thus, it is time-consuming to develop fuzzy expert systems. In this article we propose a new approach to elicit expertise and to generate knowledge bases for fuzzy expert systems. A knowledge acquisition system based upon the approach is also presented, which can help knowledge engineers to create, adjust, debug, and execute fuzzy expert systems. Some control techniques are employed in the knowledge acquisition system so that the concepts of fuzzy logic could be directly applied to conventional expert system shells; moreover, a graphic user interface is provided to facilitate the adjustment of membership functions and the display of outputs. the knowledge acquisition system has been integrated with a popular expert system shell, CLIPS, to offer a complete development environment for knowledge engineers. With the help of this environment, the development of fuzzy expert systems becomes much more convenient and efficient. © 1995 John Wiley & Sons, Inc.     

汽轮发电机组故障诊断专家系统的研究

建立一个基于知识的汽轮发电机组故障诊断专家系统ＫＢＦＤＥＳ，该系统采用“框架＋规则”知识表示技术，并将框架驻留在内存，而将规则驻留在虚拟盘上；在诊断过程中采用广义的不精确推理策略，并对重要信息用一个组合神经网络进行智能识别；系统将神经网络技术和ＩＤ３算法结合，可以实现从诊断实例自动获取知识。     

Unification Algorithms for Eliminating and Introducing Quantifiers in Natural Deduction Automated Theorem Proving

A natural deduction system was adapted from Gentzen system. It enables valid wffs to be deduced in a very natural way. One need not transform a formula into other normal forms. Robinsons unification algorithm is used to handle clausal formulas. Algorithms for eliminating and introducing quantifiers without Skolemization are presented, and unification theorems for them are proved. A natural deduction automated theorem prover based on the algorithms was implemented. The rules for quantifiers are controlled by the algorithms. The Andrews challenge and the halting problem were proved by the system.     

Knowledge acquisition and knowledge representation with class: the object-oriented paradigm

Knowledge acquisition and knowledge representation are the fundamental building blocks of knowledge-based systems (KBSs). How to efficiently elicit knowledge from experts and transform this elicited knowledge into a machine usable format is a significant and time consuming problem for KBS developers. Object-orientation provides several solutions to persistent knowledge acquisition and knowledge representation problems including transportability, knowledge reuse, and knowledge growth. An automated graphical knowledge acquisition tool is presented, based upon object-oriented principles. The object-oriented graphical interface provides a modeling platform that is easily understood by experts and knowledge engineers. The object-oriented base for the automated KA tool provides a representation independent methodology that can easily be mapped into any other object-oriented expert system or other object-oriented intelligent tools.     

Can computers carry content ‘inexplicitly’?

I examine whether it is possible for content relevant to a computer's behavior to be carried without an explicit internal representation. I consider three approaches. First, an example of a chess playing computer carrying emergent content is offered from Dennett. Next I examine Cummins response to this example. Cummins says Dennett's computer executes a rule which is inexplicitly represented. Cummins describes a process wherein a computer interprets explicit rules in its program, implements them to form a chess-playing device, then this device executes the rules in a way that exhibits them inexplicitly. Though this approach is intriguing, I argue that the chess-playing device cannot exist as imagined. The processes of interpretation and implementation produce explicit representations of the content claimed to be inexplicit. Finally, the Chinese Room argument is examined and shown not to save the notion of inexplicit information. This means the strategy of attributing inexplicit content to a computer which is executing a rule, fails.I wish to thank Fred Dretske, JOhn Perry, and an anonymous reviewer for helpful comments and suggestions. Earlier versions of this paper were read at the American Philosophical Association Pacific Division Meeting in San Francisco in March, 1993, and at the 7th International Conference on Computing and Philosophy in Orlando in August, 1992.     

A weight adaptation method for fuzzy cognitive map learning

Fuzzy cognitive maps (FCMs) constitute an attractive modeling approach that encompasses advantageous features. The most pronounces are the flexibility in system design, model and control, the comprehensive operation and the abstractive representation of complex systems. The main deficiencies of FCMs are the critical dependence on the initial experts beliefs, the recalculation of the weights corresponding to each concept every time a new strategy is adopted and the potential convergence to undesired equilibrium states. In order to update the initial knowledge of human experts and to combine the human experts structural knowledge with the training from data, a learning methodology for FCMs is proposed. This learning method, based on nonlinear Hebbian-type learning algorithm, is used to adapt the cause–effect relationships of the FCM model improving the efficiency and robustness of FCMs. A process control problem is presented and its process is investigated using the proposed weight adaptation technique.     

A framework for intelligent design of manufacturing cells

One of the major thrusts of agile/lean/responsive manufacturing strategies of the twentyfirst century is to introduce advanced information technology into manufacturing. This paper presents a framework for robust manufacturing system design with the integration of simulation, neural networks and knowledge-based expert system tools. An operation/ cost-driven cell design methodology was applied to concurrently consider cell physical design and the complexity of cell control functions. Simulation was exercised to estimate performance measures based on input parameters and given cell configurations. A rulebased expert system was employed to store the acquired expert knowledge regarding the relation between cell control complexities, cost of cell controls, performance measures and cell configuration. Neural networks were applied to predict the cell design configuration and corresponding complexities of cell control functions. Training of neural networks was performed with both forward and backward methods by using the same pair of data sets. Hence, trained neural networks will be able to predict either input or output parameters. This innovative new design methodology was illustrated via a successful implementation exercise resulting in actually acquiring an automated cell at industrial settings. The experience learned from this exercise indicates that the proposed design methodology works well as an effective decision support system for cell designers and the management in determining appropriate cell configuration and cell control functions at the design stage.     

A decision support tool for the diagnosis of breast cancer based upon Fuzzy ARTMAP

This paper presents research into the application of the fuzzy ARTMAP neural network model to the diagnosis of cancer from fine-needle aspirates of the breast. Trained fuzzy ARTMAP networks are differently pruned so as to maximise accuracy, sensitivity and specificity. The differently pruned networks are then employed in a cascade of networks intended to separate cases into certain and suspicious classes. This mimics the predictive behaviour of a human pathologist. The fuzzy ARTMAP model also provides symbolic rule extraction facilities and the validity of the derived rules for this domain is discussed. Additionally, results are provided showing the effects upon network performance of different input features and different observers. The implications of the findings are discussed.     

Optimal component-insertion sequenceplanning methodology for the semiautomatic assembly of printed circuit boards

One of the primary concerns in the manufacturing of printed circuit boards (PCBs) is the definition of optimum component-insertion sequences for manual and automated operations. The problem is dynamic and multiple feasible solutions can be found. The research work described in this paper has developed a component-insertion sequencing methodology and a proof-of-concept expert system for printed circuit boards. This system solves the component-insertion sequencing problem for semi-automated work cells which utilize a light guiding system to identify for the operator where to assemble the next component. The innovation of the methodology is the application of Artificial Intelligence and Expert Systems techniques to represent the human reasoning involved in semi-automated PCB assembly planning. Based on established assembly criteria, sequencing decision rules and data available from a CAD system, the methodology leads to optimum component-insertion sequences.     

A dialectical model of assessing conflicting arguments in legal reasoning

Inspired by legal reasoning, this paper presents a formal framework for assessing conflicting arguments. Its use is illustrated with applications to realistic legal examples, and the potential for implementation is discussed. The framework has the form of a logical system for defeasible argumentation. Its language, which is of a logic-programming-like nature, has both weak and explicit negation, and conflicts between arguments are decided with the help of priorities on the rules. An important feature of the system is that these priorities are not fixed, but are themselves defeasibly derived as conclusions within the system. Thus debates on the choice between conflicting arguments can also be modelled.The proof theory of the system is stated in dialectical style, where a proof takes the form of a dialogue between a proponent and an opponent of an argument. An argument is shown to be justified if the proponent can make the opponent run out of moves in whatever way the opponent attacks. Despite this dialectical form, the system reflects a declarative, or relational approach to modelling legal argument. A basic assumption of this paper is that this approach complements two other lines of research in AI and Law, investigations of precedent-based reasoning and the development of procedural, or dialectical models of legal argument.Supported by a research fellowship of the Royal Netherlands Academy of Arts and Sciences, and by Esprit WG 8319 Modelage.     

Symbolical Reasoning about Numerical Data: A Hybrid Approach

By combining methods from artificial intelligence and signal analysis, we have developed a hybrid system for medical diagnosis. The core of the system is a fuzzy expert system with a dual source knowledge base. Two sets of rules are acquired, automatically from given examples and indirectly formulated by the physician. A fuzzy neural network serves to learn from sample data and allows to extract fuzzy rules for the knowledge base. A complex signal transformation preprocesses the digital data a priori to the symbolic representation. Results demonstrate the high accuracy of the system in the field of diagnosing electroencephalograms where it outperforms the visual diagnosis by a human expert for some phenomena.     

Representing and using legal knowledge in integrated decision support systems: DataLex WorkStations

There is more to legal knowledge representation than knowledge-bases. It is valuable to look at legal knowledge representation and its implementation across the entire domain of computerisation of law, rather than focussing on sub-domains such as legal expert systems. The DataLex WorkStation software and applications developed using it are used to provide examples. Effective integration of inferencing, hypertext and text retrieval can overcome some of the limitations of these current paradigms of legal computerisation which are apparent when they are used on a stand-alone basis. Effective integration of inferencing systems is facilitated by use of a (quasi) natural language knowledge representation, and the benefits of isomorphism are enhanced. These advantages of integration apply to all forms of inferencing, including document generation and casebased inferencing. Some principles for development of integrated legal decision support systems are proposed.     

Symbolic and numeric knowledge integration in multiple fault troubleshooting

Many expert system researchers have reported in recent years that situation-action symbolic production rules frequently fail to provide adequate knowledge representation schemes without resorting to numeric computation. However, despite the need to integrate symbolic and quantitative computation into one coherent framework of knowledge, surprisingly few architectures have been proposed for achieving this goal. This paper explores the integration of qualitative and numeric processing in expert systems. We address the topic with respect to the construction of expert systems that perform the tasks of design and multiple fault troubleshooting. This paper shows that these tasks can be handled effectively when an appropriate interface is established between the heuristic and the numeric knowledge-based components. Specifically, we demonstrate how to interface heuristic knowledge with non-linear optimization models in order to allow an expert system greater expressiveness. An actual example is presented from the machining domain.     

SAPT—Knowledge-based CAPP system

The system for automatic programming technology (SAPT system) is a part of the designer-expert system. The new developments in artificial intelligence are promising for design and use knowledge-based expert systems. For a manufacturing process planning knowledge base, the method of knowledge representation and reasoning strategy are given. GT and TT concepts are discussed in detail from a KB approach. Production rules are given for knowledge sources, tasks and meta-knowledge. For two real life, examples, a LISP-program has been written and executed. The results obtained are encouraging and further research is in progress.     

Synthesis of fuzzy logic and Dempster–Shafer Theory for the simulation of the decision-making process in stock trading systems

Modern computerized stock trading systems (mechanical trading systems) are based on the simulation of the decision-making process and generate advice for traders to buy or sell stocks or other financial tools by taking into account the price history, technical analysis indicators, accepted rules of trading and so on. Two stock trading simulating systems based on trading rules defined using fuzzy logic are developed and compared. The first is based on the so-called “Logic-Motivated Fuzzy Logic Operators” (LMFL) approach and aims to avoid certain disadvantages of the classical Mamdani’s method, which has been developed for use in fuzzy logic controllers and not for solving the decision-making problems of stock trading. The LMFL   approach is based on the modified mathematical representation of t$t$-norm and Yager’s implication rule. The second trading system combines the tools of fuzzy logic and Dempster–Shafer Theory (DST  ) to represent the features of the decision-making process more transparently. The fuzzy representation of trading rules based on the theory of technical analysis is used in these expert systems. Since the theory of technical analysis is based on the indicators used by experts to predict stock price movements, the method maps these indicators into new inputs that can be used in a fuzzy logic system. The only required inputs to calculate these indicators are past sequences (history) of stock prices. The method relies on fuzzy logic to choose an appropriate decision when certain price movements or certain price formations occur. The optimization procedure based on historical (teaching) data is used as it significantly improves the performance of such expert systems. The efficiency of the developed expert systems is measured by comparing their outputs versus stock price movements. The results obtained using real NYSE$NYSE$ data allow us to say that the developed expert system based on the synthesis of fuzzy logic and DST provides better results and is more reliable. Moreover, such a conjunction of fuzzy logic, DST and technical analysis, makes it possible to make a profit even when trading against a dominating trend.