Ontology-Based Interconnection Of Military Knowledge

Many applications need sufficient background knowledge of military domain. Typical applications include military decision-making systems, intelligent tutoring systems, and military knowledge query systems. It is necessary for us to build a large knowledge base as the back ground knowledge, which is as correct and robust as possible. Moreover, facts have proved that it is necessary for computer to have intelligence, i.e., have enough knowledge and interconnections among knowledge. Knowledge interconnection is to build certain connections between two pieces of knowledge in terms of certain methods. The knowledge interconnection is decomposed into three classes: conceptual, statemental, and ontological. In this paper, we mainly consider conceptual, which focus on some connections between two concepts via the relation-links or attributes or values of attributes. A relation-link is a sequence of the form "C1,R1,C2,R2,…, Rn-1, Cn", where Ci is a concept,Rj is a relation, 1(i(n, 1(j(n-1. Conceptual can help us to find inconsistent knowledge in a large knowledge base.The concept interconnection has three kinds of forms: interconnection in fixed classification H, attribute-based interconnection, and relation-link-based interconnection. The relation-link-based interconnection is similar to semantic connection in MindNet.We define the least upper-concept lub(C,C') of C and C' as the least concept C" such that C,C'?C'' and C''?D for any C,C'?D. We define the great lower-concept glb(C,C') of C and C' as the great concept such that C''? C,C' and D?C'' for any D?C,C'. Our measure of interconnection-degree in H satisfies the following intuitive properties: * The more two concepts have the least upper-concepts(the great lower-concepts), the bigger the interconnection-degree is.* Interconnection-degree of two concepts having both the least-concepts and the great lower-concepts should be stronger than that of concepts having only one.* If the paths of two pairs of concepts are symmetric in H, then they have the same interconnection-degree.* The smaller the difference of levels in H is, the bigger the interconnection-degree of two concepts is.In order to discuss attribute-based interconnection, we differentiate structural-attributes, taxonomical attributes and unstructured attributes according to the domains of attributes. An attribute is structural if its domain D is structural, that is, there is a certain partial order over D. An attribute is taxonomical if its domain is taxonomical. An attribute is unstructured if its domain is unstructured. Our measure of the interconnection-degree of two concepts also satisfies the properties as follows:The more common attributes two concepts have, the bigger the* interconnection-degree is.* Different attributes should have different measures.* The values of attributes affect the interconnection-degree. Whole-part interconnection is an important attribute-based one. Similar to Winston, Chaffin, and Hermann, we give a taxonomy about the part-whole relation in military domain. Moreover, there are two important questions in whole-part interconnection: one is whether parts inherit the values of attributes of the whole, and the other is the relations between the values of certain attributes of parts in cuts and one of the same attribute of the whole. P represents a tree consisting of all parts of certain whole. A cut is a subset of P,, and has exactly one common element with any the most offshoot of the tree. We can use whole-part interconnection to check inconsistency. For example, if A' was cut-additional and C(A')(( C'(TC'(A'), then the values of A' are inconsistent. A' is called cut-additional if A' satisfies C(A')=( C'(TC'(A') for any T of P.We described two interconnection algorithms in section 5. The purpose of the first interconnection algorithm about two concepts C1 and C2 is output as much knowledge as possible. The algorithm mainly has tree steps. The first step is to consider the relations interconnection between C1 and C2. The second is t...