From Semantic Multimedia Wiki

Alignment of W3C Ontology for Media Resource and M3O

The Ontology for Media Resource^[1] developed by the W3C defines a core vocabulary for multimedia annotation. The ontology targets at an unifying mapping of common media formats like EXIF or Dublin Core. The core challenge for this alignment of the Ontology for Media Resource concerns the mapping of properties to either information object or information realization as provided by the Information Realization Pattern of the M3O.

Description

The Ontology for Media Resource defines a core vocabulary for multimedia annotation. The core challenge for this alignment of the Ontology for Media Resource concerns the mapping of properties to either information object or information realization.

In the following we describe the steps taken for aligning the Ontology for Media Resource and the M3O.

Understanding

The Ontology for Media Resource presents a property-centric approach to ontology modeling and consists of 28 predicates including properties like title and language. Some properties are specified in further detail, e.g., through role or type properties. Only entities such as multimedia items and persons are represented as resources. Any other information such as roles or types are represented using primitive values, e.g., strings. The Ontology for Media Resource defines neither structural patterns nor formal logical constraints beyond the domain and range specification for each property. Unlike the M3O, there is no distinction between information object and information realization.

Grouping

The Ontology for Media Resource's documentation ^[1] provides a number of conceptual groupings for certain aspects of multimedia description, e.g., identification or fragmentation. However, this information is not accessible in a machine readable format. With the alignment of the Ontology for Media Resource and the M3O, we have provided grouping information by defining the appropriate axioms. For example an IdentificationAnnotation concept hasPart some TitleAnnotationConcept, LanguageAnnotationConcept, and LocatorAnnotationConcept.

Mapping

For mapping the ontology for Media Resource to the M3O, we define a subconcept of the AnnotationConcept for each predicate of the ontology. For example, we define a LocatorAnnotationConcept to match the locator property. Concrete values are expressed using the Data Value Pattern of the M3O. To this end, we define appropriate Region concepts. In the case of the LocatorAnnotationConcept, we define a LocatorRegion with the property hasRegionDataValue and an URI specifying a concrete location on the web.

Of primary concern for this alignment is the mapping with the Information Realization Pattern. By taking into account the difference between information objects and information realizations, we can improve semantic precision of the aligned ontology. To this end, we examine each attribute of the Ontology for Media Resource for its inherent meaning and constrain it to the appropriate concept of the Information Realization Pattern of the M3O. As an example, the locator property of the Ontology for Media Resource annotates media files that are locatable on the web. This is a quality only applicable for information realizations and is expressed in the definition of the LocatorAnnotationConcept.

We express the type property of the Ontology for Media Resource through specialization, e.g., by specifying an ImageRealization, a subclass of the InformationRealization, as the type for the considered media item.

Finally the fragments facet of the Ontology for Media Resource has been modeled using the Decomposition Pattern of the M3O. The functionality indicated by the namedFragmentsproperty can be obtained by decomposing multimedia items using the M3O Decomposition Pattern and by using the M3O Annotation Pattern to annotate the resulting fragment with a FragmentLabelAnnotationConcept.

Consistency Checking and Documentation

The ontology has been validated using Fact++ and Pellet.

Application of the Aligned Ontology

We explicitly distinguish between an ImageObject and an ImageRealization that realizes the ImageObject. The specific type for each media is expressed through specialization of the corresponding InformationObject and InformationRealization concepts. The ImageObject is annotated with some TitleAnnotationConcept, where the title "Mona Lisa" is expressed using the Data Value Pattern. The ImageRealization is annotated with some LocatorAnnotationConcept that parametrizes a Region for an URI locatable on the web.

Application using the Ontology for Media Resource aligned with the M3O.

Ontology

ofmr.owl [1]