350 likes | 353 Views
Progress on the Risk Ontology Complying with the Risk Registry / Case Studies. Nick Bassiliades , Dimitris Vrakas Logic Programming & Intelligent Systems group Dept. of Informatics Aristotle University of Thessaloniki, Greece. Outline. Introduction Previous ontology version
E N D
Progress on the Risk OntologyComplying with the Risk Registry / Case Studies Nick Bassiliades, Dimitris Vrakas Logic Programming & Intelligent Systems group Dept. of Informatics Aristotle University of Thessaloniki, Greece
Outline • Introduction • Previous ontology version • Risk Registry Schema • New ontology version • Next Steps
Why do we need Ontologies? • All the variables associated with the Risk Assessment Process must be defined in the Risk Ontology(ies) • Inputs / Outputs to RAT • Past cases or Models • Others • Why? • To facilitate integration of risk assessment practices from different domains • To eliminate misunderstandings concerning the use of terms • To allow the use of various ways to describe the same term (synonyms, translations, etc) • To enable reasoning in a higher level of abstraction (general rules that apply to a group of specific cases)
IRIS Risk Glossary + Case Studies Schema The Ontology Server Case Studies
Outline • Introduction • Previous ontology version • Risk Registry Schema • New ontology version • Next Steps
Previous OntologyObject Properties Relating Riskto Other Concepts
Outline • Introduction • Previous ontology version • Risk Registry Schema • New ontology version • Next Steps
Risk Registry Schema • Different variables / chara-cteristics / attributes of Risk • Properties of class Risk • Different types of Risk • Subclasses of class Risk • Spatio-temporal information for the triggering of a risk occurrence • Explanation on the values that attributes of Risk can take • Determine ranges of properties
Risk Registry Data (1/2) • Case study administrative information • Each line of the table is a different instance/object of the corresponding risk class • Although currently content is just simple text, they have been modeled as objects to capture future deployment of fine-grained domain-specific ontologies • Risk Identification Methodology suggests two more properties: probability and value • Another risk property • Plural suggests 1-to-many relationship … • …but actually risk relationship to factor is 1-to-1 • Multiple Risk instances for each alternative factor • Currently 1-to-1 relationship • Needs to map to fine-grained domain-specificontologies ?? • Needs to be represented as a workflow??
Risk Registry Data (2/2) • There are no examples • Assumed simple text • There are no examples, therefore ideas from previous ontology have been used • A separate object, because it may hold internal properties (distribution, value) • Use of the same name implies a single object with two properties (description, value) • Although singular is used, sample data suggest multiple impacts • However, NOT 1-to-many relationship … • … but several 1-to-1 relationships with risk (multiple instances)
Outline • Introduction • Previous ontology version • Risk Registry Schema • New ontology version • Next Steps
New ontology versionClass Hierarchy Risk Properties • Abstract classes for inheritance purposes • value • hasProbability • Types of Impact • Hierarchy slightly changed from previous Administrative Information • Types of Risk • Previous types dropped
New ontology versionRiskVariables (Factor, Component, Mechanism and Impact)
New ontology versionCreatingOntology Instances from Registry • 4 impacts • Thus 4 instances 1 factor 1 mechanism 3 components
Outline • Introduction • Previous ontology version • Risk Registry Schema • New ontology version • Next Steps
Next StepsRiskOntology Refinement • Work with engineers to improve risk ontology • Follow advancements of case studies • E.g. Impact of one risk may lead to occurrence of another risk • Chained risks (need to be modeled?) • E.g. Maintain groups of risks with similar factors/impacts?
Next StepsHarmonize with Domain Ontologies • Integrate fine-grained domain ontologies • explore how risk variables can break down to more detailed structured descriptions (and not just text) • interrelate risk variables to each other through process workflow (mechanism) • Also, harmonize with integration (DSS) ontologies
Next StepsOntology Population & Querying • Currently data from risk registries have been encoded manually (proof-of-concept) • Suggestion: automate in the future? • Link ontology to a risk registry database (??) • Risk insertion/retrieval in/from database will be enhanced with ontology reasoning capabilities • Suggestion: Build a knowledge portal (??) • Better dissemination of case studies • Better exploitationofthe ontology • Risk Knowledge Management
Thank you! • New Ontology can be found at: http://lpis.csd.auth.gr/ontologies/2010/iris.owl • Previous Ontology can be found at: http://lpis.csd.auth.gr/ontologies/2009/iris.owl