Richard J. Haberlin Jr.
Ontologies are a fundamental enabling technology for system interoperability. They provide machine-interpretable representation of domain semantics, thus allowing interchange of information with unambiguous, shared meaning. However, a fundamental aspect of many real-world problems is uncertainty, which traditional ontologies do not represent. Representation of uncertainty in real-world problems requires probabilistic ontologies, which integrate the inferential reasoning power of probabilistic representations with the first-order expressivity of ontologies. The Reference Architecture for Probabilistic Ontology Development (RAPOD) catalogues and defines the processes and artifacts necessary for the development, implementation and evaluation of explicit, logical and defensible probabilistic ontologies developed for knowledge-sharing and reuse in a given domain. Creating an architecture for a given domain problem results in a reusable blueprint for similar designs that facilitates successful development from conceptualization to operation. With the ever-increasing volume of information delivered to the decision maker, the semantic technology community seeks to provide advanced decision support through data compilation, screening, transformation and probabilistic inference. Input may include raw data, documents, interviews, and mathematical models stored in databases, ontologies, and probabilistic ontologies. Because each decision support system (DSS) is domain-specific, it has a narrow focus of applicability and will only address a narrow set of decisions. This paper provides an example implementation of the RAPOD in the form of an architecture for a Maritime Domain Decision Support System.
Naval commanders often receive uncertain or incomplete information about contacts of interest, and it is not uncommon that military ships share similar capabilities among varied classes. There is a need for military commanders to distinguish the specific class of a potential adversary ship to best prepare for the threat. A Military Ship Probabilistic Ontology for an area of operations can assist the commander in determining the most likely ship class. Given a varied amount of data about a contact of interest, this probabilistic ontology is used to support inferential reasoning for assessing key aspects such as the most probable ship class from a set of ship classes, or how likely it is that the ship is a member of the domain set. Using the RAPOD, an architecture is instantiated to develop a Military Ship Probabilistic Ontology used for decision support.