Bayesian Ontologies for Semantically Aware Systems Kathryn Blackmond Laskey C4I Center George Mason Univesity This presentation is based on the PhD research of Paulo Costa The Need Semantically aware systems are essential to the netcentric vision. Ontologies are a means to semantic awareness Representing and reasoning with uncertainty is essential for interoperability, knowledge sharing, and knowledge reuse. But... 2 1
The Need Semantically aware systems are essential to the netcentric vision. Ontologies are a means to semantic awareness Representing and reasoning with uncertainty is essential for interoperability, knowledge sharing, and knowledge reuse. But... Standard ontology languages provide no support for representing uncertainty in a principled way 3 Addressing the Need Long Term: - Establish a Bayesian framework for probabilistic ontologies to represent knowledge with associated uncertainty This Presentation: - Present MEBN as a logical basis for Bayesian ontologies - Describe PR-OWL, a MEBN-based extension to the Ontology language OWL 4 2
What is an Ontology? In Philosophy: the study of nature of being and knowing In Information Systems: many definitions 5 What is an Ontology? In Philosophy: the study of nature of being and knowing In Information Systems: many definitions An Explicit formal specification on how to represent the objects, concepts, and other entities that are assumed to exist in some area of interest an the relationships among them. (dictionary.com) In information science, an ontology is the product of an attempt to formulate an exhaustive and rigorous conceptual schema about a domain. An ontology is typically a hierarchical data structure containing all the relevant entities and their relationships and rules within that domain (Wikipedia.org). An ontology is a set of concepts - such as things, events, and relations - that are specified in some way (such as specific natural language) in order to create an agreed-upon vocabulary for exchanging information. (whatis.com) Is a formal specification of a conceptualization (Gruber) An Ontology formally defines a common set of terms that are used to describe and represent a domain. Ontologies can be used by automated tools to power advanced services such as more accurate Web search, intelligent software agents and knowledge management. (Owl Use Cases) A partial specification of a conceptual vocabulary to be used for formulating knowledge-level theories about a domain of discourse. The fundamental role of an ontology is to support knowledge sharing and reuse. (The Internet An ontology models the vocabulary and Reasoning Services project - IRS) meaning of domains of interest: the objects (things) in domains; the relationships among those things; the properties, functions, and processes involving those things; and constraints on and rules about those things (DaConta et 6 al., 2003) 3
What is an Ontology? In Philosophy: the study of nature of being and knowing In Information Systems: many definitions An Explicit formal specification on how to represent the objects, concepts, and other entities that are assumed to exist in some area of interest an the An ontology is a set of concepts - such as relationships among them. things, events, and relations - that are (dictionary.com) specified in some way (such as specific natural language) in order to create an agreed-upon vocabulary for exchanging In information science, an ontology is the product of an attempt to formulate an exhaustive and rigorous conceptual schema about a domain. An ontology is typically a hierarchical data structure containing all the relevant entities and their relationships and rules within that domain (Wikipedia.org). What is really information. (whatis.com) Is a formal specification of a conceptualization (Gruber) important? An Ontology formally defines a common set of terms that are used to describe and represent a domain. Ontologies can be used by automated tools to power advanced services such as more accurate Web search, intelligent software agents and knowledge management. (Owl Use Cases) A partial specification of a conceptual vocabulary to be used for formulating knowledge-level theories about a domain of discourse. The fundamental role of an ontology is to support knowledge sharing and reuse. (The Internet An ontology models the vocabulary and Reasoning Services project - IRS) meaning of domains of interest: the objects (things) in domains; the relationships among those things; the properties, functions, and processes involving those things; and constraints on and rules about those things (DaConta et 7 al., 2003) Ontologies in our Research Definition: An ontology is an explicit, formal representation of knowledge about a domain of application. This includes: - a) Types of entities that exist in the domain; - b) Properties of those entities; - c) Relationships among entities; - d) Processes and events that happen with those entities; where the term entity refers to any concept (real or fictitious, concrete or abstract) that can be described and reasoned about within the domain of application. 8 4
OWL "Pedigree" Web Languages HTML - XML - RDF Source: Adapted from McGuiness, D. (2003) Ontologies: What you should know and why you might care. Presentation available at www.ksl.stanford.edu/people/ dlm/talks/cognaoct2003final.ppt DAML Frame Systems OWL OIL Description Logics FACT - CLASSIC - DLP Logical vs. Plausible Three binary variables 2 = 8 possible combinations 10 5
Logical vs. Plausible Does mom have transportation to the doctor tomorrow? 1) Yes, if Lucy or Pete gives her a ride. Otherwise, no. Logical Plausible Yes? 75% No? 25% 11 Logical vs. Plausible Does mom have transportation to the doctor tomorrow? 1) Yes, if Lucy or Pete gives her a ride. Otherwise, no. 2) Pete can't make it tomorrow. Logical Plausible Yes? 50% No? 50% 12 6
Ontologies Definition: An ontology is an explicit, formal representation of knowledge about a domain of application. This includes: - a) Types of entities that exist in the domain; - b) Properties of those entities; - c) Relationships among entities; d) Processes and events that happen with those entities; e) Statistical regularities that characterize the domain; f) Inconclusive, ambiguous, incomplete, unreliable and dissonant evidence related to entities of the domain; - Uncertainty about all the above forms of knowledge; where the term entity refers to any concept (real or fictitious, concrete or abstract) that can be described and reasoned about within the domain of application. 13 Probabilistic Ontologies Definition: A probabilistic ontology is an explicit, formal representation of knowledge about a domain of application. This includes: - a) Types of entities that exist in the domain; - b) Properties of those entities; - c) Relationships among entities; d) Processes and events that happen with those entities; e) Statistical regularities that characterize the domain; f) Inconclusive, ambiguous, incomplete, unreliable and dissonant evidence related to entities of the domain; g) Uncertainty about all the above forms of knowledge; where the term entity refers to any concept (real or fictitious, concrete or abstract) that can be described and reasoned about within the domain of application. 14 7
What is Needed? Knowledge Base Enhanced Knowledge Racer Reasoners Typical Web Service/Agent's Knowledge Flow New Data Knowledge Base Logic Reasoner Uncertainty-free Information Logical Reasoning 15 What is Needed? Knowledge Base Racer New Data Evidence Knowledge Base Logic Reasoner Quiddity Uncertainty-free Information Logical Reasoning Evidence Enhanced Knowledge Reasoners Typical Web Service/Agent's Knowledge Flow Probabilistic KB Bayesian Reasoner Evidence Bayesian Reasoning 16 8
Bayesian Networks The Star Trek Problem: Discriminating Starships and making decisions with incomplete and uncertain knowledge 17 \ Needed: More than BNs What if multiple starships show up at the same time? One BN for each situation? 18 9
Multi-Entity Bayesian Networks 19 Situation-Specific BN 20 10
Multi-Entity Bayesian Networks Synthesis of Bayesian networks and first-order logic MEBN fragments (MFrags) represent probabilistic relationships among small set of related random variables Compose MFrags into MEBN Theories (MTheories) - collection of MFrags that satisfies consistency constraints - represents probability distribution over model structures of associated first-order logic theory Use situation-specific BN (SSBN) to reason over instances 21 PR-OWL Objective: - Extend OWL ontology language to represent MEBN Theories Possible approaches: - Upper Ontology (e.g. OWL-S) or Semantic Extension (e.g. SWRL) Initial Approach: An upper ontology for probabilistic systems 22 11
PR-OWL Overview 23 MEBN/PR-OWL 24 12
Kathryn Blackmond Laskey and Paulo Costa 25 Conclusions Annotating an ontology with probabilities is not enough Our approach: Define new datatypes and usage conventions to represent qualitative structural information and numerical probabilities in OWL This requires an underlying logic that combines first-order expressiveness with probability Tools are needed to build and visualize MEBN theories 26 13
Current Work In USA: - Research on MEBN (GMU) - Pr-OWL submission to the W3C - 2 nd URSW at the 5 th ISWC (Athens, GA - November 2006) In Brazil: - Development of a MEBN reasoner (University of Brasília) In Cyberspace: - Pr-OWL Website (www.pr-owl.org) -- work in progress by C4I Center affiliate faculty Paulo Costa - Development of a Protégé PR-OWL plugin (near future) 27 PR-OWL: The Final Frontier... These are the tools for the 21st Century Our ongoing mission -- To embrace uncertainty and acknowledge diversity To create agile tools for a changing world To build a Web that no one has seen before 14