Probabilistic Ontologies Incremental Modeling Using UnBBayes La´ ecio L. Santos Department of Computer Science University of Bras´ ılia Bras´ ılia, Distrito Federal, Brazil Email: laecio@gmail.com Rommel N. Carvalho Department of Research and Strategic Information Brazilian Office of the Comptroller General Brasilia, Distrito Federal, Brazil Email: rommel.carvalho@cgu.gov.br Marcelo Ladeira and Li Weigang Department of Computer Science University of Bras´ ılia Bras´ ılia, Distrito Federal, Brazil Email: {mladeira,weigang}@unb.br Abstract—Although various formalisms for building proba- bilistic ontologies (POs) have been developed, this area still lacks adequate methodologies to guide the experts in the development and application of these ontologies. The Uncertainty Reasoning Process for Semantic Technology (URP-ST) has been proposed to remedy this deficiency by providing a process for working with probabilistic ontologies from creation to inference. Based on the concept of building POs incrementally through iterations, the Uncertainty Modeling Process for Semantic Technologies (UMP- ST), part of URP-ST, focus in evolving previous versions of a PO to handle a new set of requirements. This paper presents the implementation of UMP-ST plug-in in UnBBayes and how UnBBayes can be used to follow all steps of the URP-ST. This paper also present an experimental application, the Maritime Domain Awareness, to show the advantages of developing prob- abilistic ontologies using UMP-ST’s iterative approach. I. INTRODUCTION The Semantic Web (SW) extends the Web by adding semantic information which allow machines to understand information before understandable only by humans. To fill the lack of a standard mechanism for handling uncertainty, some methods with different formalisms have been proposed, including Probabilistic OWL (PR-OWL) [1], OntoBayes [2], and BayesOWL [3]. However, to build a probabilistic ontol- ogy (PO) is a complex task, in which the developers have frequently asked questions about which steps they should be following. The difficulty increases because there are no tools dedicated to this task. As proposed by Carvalho [4], the Uncertainty Reasoning Process Semantic Technologies (URP-ST) is a methodology for working with probabilistic ontologies. The URP-ST pro- vides a framework that describes the steps of the modeling domain and populating the knowledge base and inference. For the modeling task, Carvalho presented the Uncertainty Modeling Process Semantic Technologies (UMP-ST), which is based on Unified Process (UP) with an iterative approach in modeling and building the probabilistic ontologies from a set of requirements. UnBBayes 1 is an open source framework that allows the user to work with some methods of different formalism based on Bayesian networks [5]. In 2011, a plug-in was implemented in UnBBayes for the users to work in the first two phases of the UMP: Requirements and Analysis & Design [6], [7]. 1 http://sourceforge.net/projects/unbbayes/ This implementation, combined with existing resources in UnBBayes, such as editing probabilistic ontologies using PR- OWL and inference using Multi-Entity Bayesian Networks (MEBN) [8], allows to work also in the other two phases of the UMP (Implementation and Test), and permite the use of the tool in all stages of the more general metodology URP-ST. A PO for Maritime Domain Awareness to identify ships of interest (Ships that bring some kind of threat) is presented as a case study to illustrate the advantages of the plug-in developed for UnBBayes. This use case was created during the project PROGNOS (Probabilistic OntoloGies for Net-centric Opera- tion Systems) [9], through a partnership between academia and experts from the U.S. Navy. A probabilistic ontology was modeled for the domain using PR-OWL/MEBN [10] technologies. This paper presents the implementation of UMP-ST plug-in in UnBBayes. It also aims to show how to model the Maritme Domain Awareness ontology iteratively using this plug-in, and how each step of the URP-ST can be followed using the framework. Using the proposed approach, several common errors in probabilistic modeling ontologies can be avoided. This paper is organized as follows: Section II presents the fundamental knowledge of URP-ST, UMP-ST, PR-OWL and MEBN. Section III introduces UnBBayes and the plug- in developed for UMP-ST. In Section IV the use case for Maritime Domain Awareness is described to show how this domain is modeled using UMP-ST’s incremental approach and how to follow the steps of the URP-ST using UnBBayes. Section V presents some concluding remarks. II. FUNDAMENTAL KNOWLEDGE This session presents the concepts of URP-ST, UMP-ST, MEBN and PR-OWL to provide a basic understanding about these formalisms and the development in these areas. A. URP-ST The Uncertainty Reasoning Process for Semantic Technol- ogy (URP-ST) was developed by Carvalho [4], [11] as an approach for modeling probabilistic ontologies and how to use it to make plausible reasoning in applications that use semantic technologies. It is divided into three steps, as shown in Figure 1. The first step is to model the domain including the survey of requirements, analysis, design and implementation,