My work in the area of knowledge representation essentially involves conceptual analysis (of biology domain in my case). While performing the task of elicitation of knowledge, I use concept maps (refined concept maps) to get the experts (and novices) to articulate their knowledge. While performing the task of formalization of knowledge, I play a role of knowledge engineer wherein i re-encode the knowledge of the subject matter (which is expressed in natural language) into AI language i.e. applying constraints, rules and facts.
My earlier post on my work on building the knowledge organizers deals with formalization part of conceptual analysis. I do know that this research work is just a beginning of a massive project, and I really intend to continue working on it.
knowledge acquisition involves elicitating, analyzing, and formalizing the subject matter. during knowledge elicitation, the knowledge engineer asks expert to articulate the tacit knowledge in natural language. during formalization the knowledge engineer encodes the knowledge that is elicited from experts in rules and facts of some AI language.
conceptual analysis is central to knowledge acquisition. conceptual analysis is the task of analysing concepts expressed in natural language and making the implicit relations into explicit relations.
understanding of logic and language and philosophy (ontology) are pre-requisite for working on conceputal analysis. in computer science field, it is also called as conceptual modelling. other than its application in computer sciences, conceptual analysis can also be applied in different contexts such as interviewing an experimenter, child; an automobile expert in manufacturing or designning.
conceptual analysis determines the general principles which constitutes of semantic and episodic memory for defining an expert system. while episodic memory deals with information related to facts, the semantic memory includes five kinds of information—taxonomy, definitions, constraints, schemata, behavior.
although it is possible to perform conceptual modelling by using automated tools, the basic tools of usage are paper and pencil. One can cite two different tools—concept maps (Novak), and conceptual graphs (Sowa)—as tools for conceptual modelling.
concept maps serve as a common notation for experts and knowledge engineer, since they are informally used, are flexible, simple, and yet because of these features they lack in formal structure for knowledge representation and cannot be used to express logic and semantics. however, concept maps can be formalized by applying constraints.
conceptual graphs follows the version of semantic networks designed involving logic. they can be mapped directly to and from natural language; they can be translated to and from other AI formalisms; they can support automated knowledge acquisition tools.
Reference: John Sowa (1992), Conceptual Analysis as a basis for knowledge acquisition.