[ rdfs:comment """ I think there are three major ways we can advance the ViPER data format. We could rewrite ViPER to use MPEG-7 natively, switch to an OWL/RDF based description logic system, or more conservatively expand the system to include relations and take links to descriptors as attributes. The first would be to take on the often-suggested MPEG-7 format. While this appears to be a simple proposition at first, it would involve modifying the format to have the description definitions in a seperate file, or something. I haven't quite figured it out. This goes back to something I thought about when switching over to XML - using XML-Schema as the section of the files. I didn't do it at the time, since there wasn't a convenient method of accessing the datatype information through the DOM. ViPER should be rewritten as rdf using owl. This would give it a stronger advantage over mpeg-7 while also giving some tools and a neater format (n3). Should it use rdf for the viper and xsd for the data? I would choose to allow the user to select the type of attribute. The main problem I see is developing a robust and straightforward representation of "temporally qualified triples". I find this particularly interesting in light of other work to describe media using description logics like OWL. Finally, a more conservative approach would extend the current model with a 'extends' attribute on the 'config' tag. There would also be two additional data types, 'relation' and 'subtype-value'. The first one would be like having an ObjectProperty that is restricted to other descriptors, and the second is roughly equivalent to restricting the ObjectProperty to be a subclass of a given class, or disjunction of classes. This page represents an idea for representing ViPER data in OWl. First is the ontology: """ ] . ################################################################### # preliminary OWL schema # will require an xml schema document, still, to define # interval and range data types. ###### :MediaFile a owl:Class ; rdfs:comment "A Media file that is described by some viper:Descriptors. " ; rdfs:subClassOf [ a owl:Restriction ; owl:onProperty :hasFileDescriptor ; owl:maxCardinality "1" ] . # Without going to OWL-DL, there is no way to do this. # So I have to have 3 different :hasDescriptor # properties that do not inherit from one place. # If I could use owl:oneOf, then I could # do it. This goes to an OWL-DL v. OWL-Lite debate. # see http://www.w3.org/TR/owl-guide/#OwlVarieties :hasFileDescriptor a owl:ObjectProperty ; rdfs:range :MediaFile ; rdfs:domain :FileDescriptor . :hasContentDescriptor a owl:ObjectProperty ; rdfs:range :MediaFile ; rdfs:domain :ContentDescriptor . :hasObjectDescriptor a owl:ObjectProperty ; rdfs:range :MediaFile ; rdfs:domain :ObjectDescriptor . :Descriptor a owl:Class. :frameRange a owl:DatatypeProperty ; rdfs:range :Descriptor ; rdfs:domain vd:spanList . :timeRange a owl:DatatypeProperty ; rdfs:range :Descriptor ; rdfs:domain vd:spanList . # like "12:13 16:17" # Don't need an id, that is handled by rdf. # should have a name, perhaps? Or maybe have # the name be handled by the rdfs:label # of the class. Too bad you can't say that # a Descriptor can't be directly instantiated. :FileDescriptor a owl:Class; rdfs:subClassOf :descriptor ; rdfs:subClassOf [ a owl:Restriction ; owl:onProperty :hasAttribute ; owl:allValuesFrom :StaticAttribute ] . :ContentDescriptor a owl:Class; rdfs:subClassOf :descriptor ; rdfs:subClassOf [ a owl:Restriction ; owl:onProperty :hasAttribute ; owl:allValuesFrom :StaticAttribute ] . :ObjectDescriptor a owl:Class; rdfs:subClassOf :descriptor . :hasAttribute a owl:ObjectProperty ; rdfs:domain :Descriptor ; rdfs:range :Attribute . # If we want to use anything less than # OWL Full and have reference types mixed with # data types, we need to make everything Objects :Attribute a owl:Class . :DynamicAttribute a owl:Class ; rdfs:subClassOf :Attribute . :frameSpan a owl:DatatypeProperty ; rdfs:range :DynamicAttribute ; rdfs:domain vd:span . :timeSpan a owl:DatatypeProperty ; rdfs:range :DynamicAttribute ; rdfs:domain vd:span . :StaticAttribute a owl:Class ; rdfs:subClassOf :Attribute . ######################################### # Next comes an example document. # owl:Ontology rdf:about <> ; owl:imports ; owl:imports . :FileInformation rdfs:subClassOf viper:FileDescriptor . :named a owl:ObjectProperty ; rdfs:subPropertyOf viper:hasAttribute ; rdfs:label "has file name of" ; rdfs:range :FileInformation ; rdfs:domain viperdata:svalue . :Movement rdfs:subClassOf viper:ObjectDescriptor . :Arrive rdfs:subClassOf :Movement . # appear on scene :Exit rdfs:subClassOf :Arrive . # from inside a vehicle :Depart rdfs:subClassOf :Movement . #leave scene :Enter rdfs:subClassOf :Depart . # get into a vehicle, disappearing :Pass rdfs:subClassOf :Arrive ; # Walk through the scene rdfs:subClassOf :Depart . :Approach rdfs:subClassOf :Movement . :Leave rdfs:subClassOf :Movement . :Move rdfs:subClassOf :Movement . # w/o entering or exiting scene. # perhaps should be subclass of leave and approach :goTowards a rdfs:ObjectProperty . # applies to approach activities :goFrom a rdfs:ObjectProperty . # applies to leave activities :involvesPerson a rdfs:ObjectProperty . # applies to Movements :involvesVehicle a rdfs:ObjectProperty . # also to Movements a viper:Sourcefile ; viper:hasFileDescriptor [ a :FileInformation ; :someAttribute [ a :svalue ; :hasValue "A file" ] ] ; viper:hasObjectDescriptor [ a :Leave ; :- :leaveEvent01 ; :goFrom :person01 ] ; # what about dynamic # attributes? viper:hasObjectDescriptor [ a :Person ; :- :person01 ; : ] . ########################################### How to express in the old format? First, allow inheritance of descriptors. .... .... will inherit all Movement's attributes. Not sure how to handle overloading yet. Locale This should be easy to build into the API. Not sure how difficult it would be to add to viper-gt 3.6, and not clear about how to evaluate these.