Knowledge modeling directed by situation-specific models

Declarative representation of strategic control knowledge

“…We then review the general process of constructing explicit SSMs, highlighting the roles of subgoaling and goalsequencing knowledge in this process. [This process is discussed extensively in the context of a KBS architecture called Ace-Ssm (Benaroch, 1998a) and a knowledge modeling approach called Ssm-Dkm (Benaroch, 1998b).] Finally, in light of this review, we sharpen our statement of the specific issues involved in representing and using declarative goal-sequencing knowledge.…”

“…This same knowledge can be captured naturally using our declarative formalism but not using conventional logic programming formalisms. Another situation where our representation is especially useful arises in the context of applications that use ontological knowledge to support, for example, meta-reasoning or inter-application communication via a middleware layer (Benaroch, 1998a). Because of the ontology-based nature of our representation, such applications could express even their goal-sequencing knowledge declaratively.…”

“…In SSM terms, we know that all the candidates added to the SSM up to a given moment are organized following the topology of the disease taxonomy, and so the most specific candidates must be the roots of the lowest-level SSM subgraphs. As such, roots of competing SSM subgraphs (candidates) bear father, child or sibling relationships, and principles in S G are designed to exploit these relationships so as to effect the search in different ways (Benaroch, 1998b;Clancey, 1988). For instance, a principle that switches the current focus node with one of its children focuses attention to a more specific disease within the last disease category considered (or within the currently pursued SSM subgraph), and a principle that switches the current focus node with an ancestor of candidates in the differential transfers attention to a broader disease category.…”

“…The problem-space paradigm readily maps to the SSM construction process (Benaroch, 1998a). The initial state, SSM i , is an SSM containing only initially provided problem inputs (e.g.…”

“…SSMs for typical diagnosis tasks are causal arguments having the structure of a proof). Benaroch (1998a) showed in the context of Ace-Ssm (architecture for constructing explicit SSMs) that a declarative specification of the structure underlying SSMs created for a target task defines a meta-level goal structure for the task, and it hence constitutes declarative ontologybased subgoaling knowledge. In Ace-Ssm, starting with an explicit SSM reflecting what is initially known about a target problem situation, problem solving uses this subgoaling knowledge (or meta-level goal structure) to gradually post new subgoals in the course of evolving the SSM into a well-defined goal state.…”