CAISOR Knowledge Representation Framework

Components of the KR Framework

The Knowledge Representation Framework (KRF) is an approach to the representation and organization of structured information both in conventional software and in knowledgebases and knowledge-based systems. This framework has the following major parts.

Syntax and Languages

The KRF is based on the use of Knowledge Representation Expressions (KR Expressions) as a more readable and expressive alternative to S-expressions, JSON notation and similar notations for representing structured information. KRF notation is closer to the notation of logic and of extended set-theory, through its use of a variety of brackets, and of infix notation.

A number of languages in the KRF use KR expressions as their syntactic framework, just like KIF, KQML, PDDL and several others use S-expressions for their syntax. This includes the Common Expression Language which plays a role similar to KIF, the Document Scripting Language for markup of documents and webpages, and several others.

The Session Command Language is a KRE-based scripting language which is used for giving commands to sessions with KRF-style software, including software agents.

Representation and Aggregates

The basic representational structure in KRF is the use of entities which may be described by assigning values to attributes for those entities. However, KRF differs in the following ways from RDF-based representation systems, such as OWL:

• The values of attributes may be arbitrary KR expressions, and not merely entities or sets of entities. KR expressions have a recursive syntax allowing for sets, sequences, records and other structures.
• Entities are not restricted to being atomic symbols; composite entities can be formed using functions with their arguments (like in Prolog, for example).

An entity description is a textual expression for an entity and the assignments of values to its attributes. Sets of entity descriptions can be aggregated into entityfiles, and sets of these can form knowledgeblocks on one or more levels.

In addition to the knowledgeblock hierarchy, the KR Framework defines autonomous agents as software artifacts, software individuals as aggregates consisting of several agents that are functionally diverse i.e. they play different roles that contribute to the behavior of the individual, and societies of individuals that may cooperate through message-passing.

Purposes

The KR Framework is used for the following purposes.

• As the notational and conceptual basis for lecture notes and other educational resources in the area of Artificial Intelligence, with an emphasis on Knowledge Representation.
• As the defining framework for the Leonardo software system which is the first implementation of the Framework.
• As the defining framework for the Common Knowledge Library, a freely available knowledgebase which at present contains more than 60,000 entities each with a number of attribute-values.
• As the basis for our work in the area of information analysis.

The work on the Knowledge Representation Framework is a major part of the CAISOR Research Agenda . Other parts of the agenda concern information analysis and open-access publication . These activities make use of both the KR Framework and its implementation in the Leonardo software system.

Additional Information

Please consult the other left-side menu items for additional information on the various aspects of the Knowledge Representation Framework, and refer to the following reports:

PM-krf-019	Knowledge Representation Framework: Introduction and Overview
PM-krf-009	Knowledge Representation Framework: Overview of Languages and Mechanisms
PM-caisor-001	The CAISOR Agenda: A Comprehensive Research Agenda for Software Architecture and Open-Access Knowledgebases

PS.

Both the present webpage and the project's reports, such as those mentioned above have of course been produced using KRF-based software, markup languages and information structures.