150 likes | 266 Views
Knowledge Acquisitioning. Definition. The transfer and transformation of potential problem solving expertise from some knowledge source to a program. How Done??. Could be achieved by a computer program that creates associations using a large body of case data
E N D
Definition • The transfer and transformation of potential problem solving expertise from some knowledge source to a program
How Done?? • Could be achieved by a computer program that creates associations using a large body of case data • Knowledge Elicitation: Series of interviews between the domain expert and the knowledge engineer • Interaction between a domain expert and a computer program
Poor Productivity • Knowledge Elicitation results in poor productivity ( 2-5 rules per day) because of: • The knowledge engineer has to learn something about the domain before communicating with the domain expert • Experts think less in terms of general principles, and more in terms of typical objects • Finding a good notation for expressing domain knowledge is a very difficult
Stages of Knowledge Acquisition • Identification : • Identify the class of problems that the system is expected to solve, and the criteria that solutions must meet • Conceptualization: • Uncover the key concepts and the relationships between them, which include: Data characteristics , flow of information and the domain structure.
Cont… • Formalization: • Understand the nature of the search space, and the character of the search that will be conducted. • Important issues: completeness and certainty of information • Implementation: • Turning the formalization of knowledge to an executable program • Important issues: Data Structures, modules independency • Testing: • Evaluation by running the program on a large and representative sample of test cases • Possible errors are missing, incomplete or incorrect rules
KADS • KADS is framework for a modeling approach of knowledge engineering. • Knowledge-Based systems are not just containers of knowledge • They are operational model that exhibits some desired behavior and impacts real-world phenomena • Knowledge elicitation is not just eliciting domain knowledge but also interpreting this data with respect of some conceptual framework, and formalize it in such a way that the program can actually use the knowledge
KADS Principles • Introduction of multiple models to cope with knowledge engineering complexity • KADS four-layer framework for modeling the required expertise • The reusability of generic model components • The process of differentiating simple models into more complex ones • The importance of structure-perceiving transformation of models of expertise into design and implementation
Cont… • The motivation of KADS is to manage complexity. • The first principle provides multiple models to help knowledge engineering in facing some issues: • Defining the problem that the expert system is meant to solve • Defining the function of the expert system • Defining the tasks that must be performed to fulfill the expert system’s function
KADS Models • Organizational model of the environment in which the system will function • Application model of the problem to be solved • Task model which shows how the function will be fulfilled by breaking the desired behavior down into component tasks • Model of cooperation or communication which is responsible for decomposing the problem solving behavior into primitive tasks and then distributing the tasks across agents • Model of expertise which is the analysis of different kind of knowledge • Design Model that suggests computational techniques and representation mechanisms
Previous KADS Version models • Knowledge conceptualization • Formal description of knowledge using concepts and conceptual relations • Epistemological analysis • Uncover properties of the conceptual knowledge • Logical analysis • Knowledge about how to perform reasoning of tasks • Implementation analysis • Deals with mechanisms upon which other levels of analysis are based
KADS four-layer framework • It is used in differentiating between different kinds of knowledge according the roles they play in problem solving • Knowledge category: • Strategic: controls execution of tasks • Task: what tasks to perform • Inference: how to reason in a domain • Domain: concepts, property and relations
Ontological analysis • This approach describes systems in terms of entities, relations between them and transformations between entities • Ontological analysis assumes that the problem can be reduced to a search problem, but doesn’t focus upon the method of search ( OPAL System ) • These analyses are abstract, but they are valuable in structuring an ill-structured task such as the hard task of finding a suitable framework for organizing the knowledge in the process of knowledge elicitation
Cont… • Main categories for structuring domain knowledge: • Static ontology: consist of domain entities with their properties and relations • Dynamic ontology: defines the states in problem solving, and how one state can be transformed to another • Epistemic ontology: knowledge that guides and constraints state transformations