Representation and learning of concepts based on scope

The authors assume that meanings of concepts consist of relations among concepts (definitions of concepts) and a habit which an agent using concepts has. They formulate this idea as a new framework of a knowledge representation system of concepts by introducing scope, i.e., a set of concepts which is in the present consideration. In the framework, they treat relations among concepts by means of partial extensions, i.e., functions from a possible model of subtheory, which is determined by the present scope, to the extension of each concept at the model. A habit is treated as a restriction of motion of scope. Then, inference based on the framework and a connection between efficiency of reasoning and a semantic property are studied. This connection leads to the possibility of extending explanation-based learning.<
>