Position Paper: Using Ontology-based Rules for Situation Awareness and Information Fusion Christopher J. Matheus Versatile Information Systems, Inc. 5 Mountainview Drive Framingham MA 01701 USA cmatheus@vistology.com Background Versatile Information Systems, Inc. (VIS) has been applying Semantic Web technologies to the problems of situation awareness and information fusion for more than five years. Situation awareness involves the real-time processing of event-based information coming from an evolving situation in an attempt to understand what is happening. In our view, situation awareness primarily comes down to identifying higher-order relations that come into being within a situation and that have particular relevance to the problem at hand as defined by the user’s goals or objectives 1 , 2 . By higher-order relations we mean relations involving multiple objects; OWL ObjectProperties represent the simplest of such relations involving two objects but situation awareness is often interested in more complex relations involving several objects. We contrast these higher-order relations with those that merely define characteristics of an individual object; DataProperties fall into this category. The analytical processes that go into establishing situation awareness necessarily involve information fusion, the processes by which data/information from multiple sources are combined to produce new enhanced information that incorporates aspects of the raw, original sources. Our approach to both situation awareness and data fusion involves the use of formal ontologies to describe the fundamental events, objects and relations of a situation’s domain and logical rules to define ways of fusing information and identifying higher-order relations relevant to the situation at hand. We have been working with DAML/OWL for representing our formal ontologies since the language’s inception. For rules we began by using RuleML, which we systematically converted to Jess 3 rules using XSLT scripts. The rule we have written have always been grounded in the classes and properties of one or more OWL ontologies; we refer to this as constructing rule-based domain knowledge on top of OWL ontologies 4 . With the advent of SWRL in the fall of 2003 we adopted it as our primary rule language because of its close coupling with OWL, our primary language of discourse. We have since developed a graphical SWRL rule editor called RuleVISor, which we are planning to release to the community in the near future. A distinguishing feature of RuleVISor is that it greatly facilitates the development of rules based on ontologies by permitting the user to drag and drop OWL ontology elements into the head and body of a rule. Challenges in Using SWRL on top of OWL We strongly advocate the use of formal methods for representing ontologies and reasoning with rules. We have attempted to whole-heartedly adopt OWL and SWRL as formal languages for the use in our research and development systems. Doing so however has come at a cost due to several challenges encountered in trying to use these languages for practical applications. Struggles with the Binary-Property Limitation The restriction to binary properties and the lack of a join mechanism in OWL has forced us at times to construct rather unnatural ontologies and/or represent complex properties in the form of rules rather than more simple ontological constructs. In SWRL, the restriction to the use of binary predicates that it inherited from OWL has significantly hampered our ability to readily codify domain knowledge. We find ourselves first writing rules using higher-order predicates and then manually translating them into SWRL rules with the addition of “relation classes” to capture the