EXPECT in a Nutshell

EXPECT is an environment for developing knowledge-based systems that includes knowledge acquisition (KA) tools to extend and modify knowledge bases. Successful knowledge acquisition tools use a role-limiting approach where they help users provide the domain-dependent knowledge needed for a given kind of problem. Several such tools have been built to support KA for a specific problem-solving strategy: SALT for propose-and-revise [Marcus & McDermott 89], MOLE for cover-and-differentiate [Eshelman 1998], PROTEGE for skeletal plan refinement [Musen 1989], etc. Although having a role-limiting strategy provides very strong guidance for knowledge acquisition, these tools lack the flexibility that knowledge-based system construction needs [Musen 1992] because the problem-solving method is hard-coded and cannot be changed with the tool.

In contrast, EXPECT represents explicitly both factual knowledge and problem-solving knowledge and its knowledge acquisition tools can reason about each piece of knowledge and analyze its interactions with others in the context of problem solving. EXPECT is tightly integrated with Loom, a state-of-the-art knowledge representation system based on description logic.

EXPECT generates automatically an Interdependency Model (IM) that captures the relationships between different pieces of knowledge in a knowledge base. EXPECT's KA tool uses this IM to determine what KA tasks the user needs to complete. A basic type of interdependency occurs between ontologies and problem-solving knowledge, i.e., how the ontologies are used during problem solving and what problem-solving knowledge is required given the ontological knowledge about the domain. In effect, this basic IM determines which aspects of a very large background ontology are relevant to the problem at hand. For example, the class city may have attributes such as latitude, longitude, airports, seaports, mayor, subway-system, restaurant, etc. In a transportation application, when the user wants to add a city to the knowledge base, the IM would determine that only airport and seaport information is required and would not ask a user to specify any of the other attributes (such as the restaurants or the name of the mayor).

EXPECT's ontologies and factual knowledge includes concepts, instances, and the relations among them. Problem-solving knowledge is represented in a procedural-style language that is tightly integrated with the Loom representations. Subgoals that arise during problem solving are solved by methods. Each method description specifies: 1) its capability in terms of the kinds of goals that the method can achieve, 2) the type of result that the method returns, and 3) the method body that contains the procedure that must be followed in order to achieve the method's goal. A method body can contain nested expressions, including subgoal expressions that need to be resolved by other methods; control expressions such as conditional statements and some forms of iteration; and relational expressions to retrieve the fillers of a relation over a concept.

EXPECT's representation of problem-solving method capabilities and goals is tightly coupled with the domain ontologies in the knowledge base, can express task-related parameters explicitly, and is based on case grammars. This representation allows EXPECT to reason about goals and capabilities through subsumption and reformulation.

EXPECT's predecessor was the Explainable Expert Systems architecture [Swartout et al 1991], which exploited similar representations in order to generate adequate explanations of a knowledge base system's behavior in natural language.

In summary, EXPECT exploits explicit representations of knowledge to derive interdependency models that are useful to guide users in knowledge acquisition.

We have shown that this approach is more flexible than existing approaches by illustrating how EXPECT can acquire the same kinds of information as a well-known KA tool, and can support the acquisition of additional kinds of knowledge. EXPECT is unique in that it is able to guide the user to acquire and modify problem-solving knowledge, while other KA tools do not support these kinds of modifications.