New trends in development of databases and expert systems seems to underline the role of graphical specification tools, visual information modeling and formal verification procedures. This paper incorporates these new ideas and, moreover, tries to present putting them in engineering practice. The main goal is to move rule representation, called eXtended Tabular Trees. The main idea behind XTT is to build a hierarchy of Object-Attribute-Value Tables (OAV table). The basic component for knowledge specification is an OAV table. It is analogous to a relational database table; however, it contains conditional part and decision columns. Moreover, the attribute values can be non-atomic ones. Each row provides specification of a single rule. The OAV tables can be connected with one another through appropriate links specifying the control flow in the system. The design specification is automatically translated into Prolog code, so the designer can focus on logical specification of safety and reliability. On the other hand, formal aspects such as completeness, determinism, etc. are automatically verified on-line during the design, so that it verifiable characteristics are preserved. From practical point of view, the design process is performed with a intelligent tool named Mirella. 1.

As shown in numerous studies, a significant part of published clinical guidelines is tainted with different types of semantical errors that interfere with their practical application. The adaptation of generic guidelines, necessitated by circumstances such as resource limitations within the applying organization or unexpected events arising in the course of patient care, further promotes the introduction of defects. Still, most current approaches for the automation of clinical guidelines are lacking mechanisms, which check the overall correctness of their output. In the domain of software engineering in general and in the domain of knowledge-based systems (KBS) in particular, a common strategy to examine a system for potential defects consists in its verification. The focus of this work is to present an approach, which helps to ensure the semantical correctness of clinical guidelines in a three-step process. We use a particular guideline specification language called Asbru to demonstrate our verification mechanism. A scenario-based evaluation of our method is provided based on a guideline for the artificial ventilation of newborn infants. The described approach is kept sufficiently general in order to allow its application to several other guideline representation

In order to address problems encountered in the design of rule-based and expert systems the paper discusses a proposal of a new approach to rule-based system design and verification using an integrated CASE tool supporting visual design of rule-based systems. The Mirella tool allows for dynamic on-line specification of components of the knowledge with simultaneous analysis of the rule-based system during the design phase. The main idea is to move the design procedure to a more abstract, logical level, where knowledge specification is based on use of abstract rule representation, called eXtended Tabular Trees, supported by Mirella CASE tool. It consists of visual design environment integrated with Prolog-based analysis and verification engine. Along with editor`s built in on-line checking capabilities it improves system performance and safety. Selected implementation aspects are also highlighted. The environment is implemented in ANSI C in the GNU/Linux environment using Gtk/GNOME/SWI-Prolog development platform. This makes it both efficient and portable. The tool has a multilayer, multimodule architecture which gives it flexibility and allows for future extensions. 1.

During the last years, a number of formal specification languages for knowledge-based systems has been developed. Characteristic for knowledge-based systems are a complex knowledge base and an inference engine which uses this knowledge to solve a given problem. Specification languages for knowledge-based systems have to cover both aspects. They have to provide means to specify a complex and large amount of knowledge and they have to provide means to specify the dynamic reasoning behaviour of a knowledge-based system. This paper focuses on the second aspect. For this purpose, we survey existing approaches for specifying dynamic behaviour in related areas of research. In fact, we have taken approaches for the specification of information systems (Language for Conceptual Modelling and TROLL), approaches for the specification of database updates and logic programming (Transaction Logic and Dynamic Database Logic), and the generic specification framework of Abstract State Machines. Keywords...

Traditional approaches to validation and verification of KBS aim at investigating properties of a KBS which are independent of the particular task of the KBS, and are phrased in terms of the implementation language of the final system. In contrast to this, we propose an approach to validation and verification of KBS which exploits task-specific properties of a KBS, and which is based on an implementationindependent conceptual model of the system. We illustrate our approach by investigating the taskspecific properties of a conceptual model for a wide class of diagnostic systems. 1 Introduction 1.1 Motivation Much of the current work on V&V of knowledge-based systems (see e.g. [Lydiard, 1992, Ayel & (eds.), 1991, Hamilton & Kelly, 1991] for some collections) has the following characteristics: Firstly, it is aimed at verifying properties of the implementation of a KBS (e.g. in rules or frames). Secondly it verifies properties which are independent of the particular task of the KBS (e....

This paper examines how formal specification techniques can support the verification and validation (V&V) of knowledge-based systems. Formal specification techniques provide levels of description which support both verification and validation, and V&V techniques feed back to assist the development of the specifications. Developing

This paper describes a new approach, the HeKatE methodology, for design and development of complex rule-based systems for control and decision support. The main paradigm for rule representation, namely the eXtended Tabular Trees (XTT), ensures high density and transparency of visual knowledge representation. Contrary to traditional, flat rule-based systems, the XTT approach is focused on groups of similar rules rather than single rules. Such groups form decision tables which are connected into a network for inference. Efficient inference is assured thanks to firing only rules necessary for achieving the goal, identified by context of inference and partial order among tables. In the paper the new version of the language – XTT 2 – is presented. It is based on the ALSV(FD) logic, also described in the paper. Another distinctive feature of the presented approach is a top-down design methodology based on successive refinement of the project. It starts with Attribute Relationship Diagram (ARD) development. Such a diagram represents relationships among system variables. Based on the ARD scheme, XTT tables and links between them are generated. The tables are filled with expert-provided constraints on values of the attributes. The code for rules representation is generated in a human-readable representation called HMR, and interpreted with provided inference engine called HeaRT. A set of software tools supporting the visual design and development stages is described in brief.