Re: ACTION-178: What is ISSUE-116 (Axiomatic Triples for OWL R) about?

Ian Horrocks wrote:
> Michael (and Ivan),
> 
> I don't have any problem with axiomatic triples in principle. However, 
> while they may be (relatively) harmless in principle, I worry that they 
> could be extremely damaging from an implementation perspective.
> 
> Presumably, making axiomatic triples be part of OWL RL (Full) would mean 
> extending the rule set so that it would generate such triples. There 
> could be a very large (perhaps even infinite) number of such triples. 
> This might be a serious burden on implementations and lead to a 
> significant degradation in performance.
> 
> I CCed Zhe on this in the hope that we can get a view on this from an 
> OWL R implementer.
> 

Although not Zhe:-) but I did implement RDFS a while ago using Herman 
ter Horst's approach. What it does is for a specific graph was to look 
at the rdf:_i properties, looks at the maximum 'i', and use the axioms 
(eg, rdf:_1 rdf:type rdfs:ContainerMembershipProperty) for that interval 
only. That takes care of the infinite issue for each specific graph.

Yes, performance might be an issue, Zhe is better positioned to answer that.

Ivan


> ian
> 
> 
> 
> 
> 
> 
> On 11 Aug 2008, at 15:15, Michael Schneider wrote:
> 
>> [cc'ing Ivan Hermann as the raiser of ISSUE-116.]
>>
>> Dear WG!
>>
>> At F2F3, ACTION-178 was put on me to explain my understanding of
>>
>>   ISSUE-116: "Should Axiomatic Triples be added to OWL-R Full?"
>>
>> Here is my understanding, after a short explanation what axiomatic 
>> triples are.
>>
>> (A) What are axiomatic triples?
>> -------------------------------
>>
>> These are typical examples of axiomatic triples (taken from RDFS):
>>
>>   (1) rdf:Property rdf:type rdfs:Class
>>   (2) rdf:type rdf:type rdf:Property
>>
>> These axiomatic triples specify the part of the universe to which 
>> certain URIs of the "system vocabulary" belong: 'rdf:Property' denotes 
>> a class, while 'rdf:type' refers to a property.
>>
>> Note that in RDF, there is no real distinction between the vocabulary 
>> specifying the language (such as 'rdf:type'), and the custom 
>> vocabulary used within an ontology. Axiomatic triples interpret the 
>> system vocabulary, by telling what's true for them independently of 
>> the concrete ontology, in which they are used.
>>
>> Axiomatic triples are defined in RDF and RDFS, and also in pD*. OWL 1 
>> Full also contains lots of axiomatic triples for the OWL vocabulary, 
>> though they are written in a different notation there.
>>
>> (B) Which axiomatic triples do typically exist?
>> -----------------------------------------------
>>
>> Aside of type information (as in the examples (1) and (2)), RDFS also 
>> specifies the domain and range for all properties in the RDFS 
>> vocabulary, e.g.:
>>
>>   (3) rdf:type rdfs:domain rdfs:Resource
>>   (4) rdf:type rdfs:range rdfs:Class
>>
>> (C) What is the scope of the issue?
>> -----------------------------------
>>
>> The main question at F2F3 was whether this issue is also about 
>> *additional* genuine OWL R Full axiomatic triples, or just about RDFS 
>> axiomatic triples. From Ivan's mail it becomes pretty clear that this 
>> issue includes axiomatic triples, which go beyond RDFS (Ivan confirmed 
>> this to me in a private mail). One of his stated examples is:
>>
>>   "owl:FunctionalProperty,rdfs:subClassOf,rdf:Property"
>>
>> This assertion clearly does *not* exist in RDFS, since 
>> 'owl:FunctionalProperty' doesn't belong to the RDFS vocabulary. 
>> (Whether this example would be a useful axiomatic triple or not for 
>> OWL R Full is not a topic here.)
>>
>> (D) What would be the additional axiomatic triples?
>> ---------------------------------------------------
>>
>> Defining axiomatic triples is in most times a pretty straight forward 
>> task. For all OWL URIs in the OWL R vocabulary the type is stated 
>> (Resource, Class, Property), and for properties also the range and 
>> domain. All axiomatic triples which I have in mind would follow the 
>> examples (1) to (4) above.
>>
>> For example, for 'owl:complementOf', I would propose the axiomatic 
>> triples:
>>
>>     owl:complementOf
>>         rdf:type rdf:Property ;
>>         rdfs:domain rdfs:Class ;
>>         rdfs:range rdfs:Class .
>>
>> Since OWL 2 Full will also have axiomatic triples (the old ones from 
>> OWL 1 Full, and new ones for the new vocabulary), OWL R Full can 
>> simply take what it needs from OWL 2 Full.
>>
>> (E) Could additional axiomatic triples lead to problems?
>> --------------------------------------------------------
>>
>> I suppose that if RDFS axiomatic triples do not already lead to 
>> problems, then additional axiomatic triples for the OWL R vocabulary 
>> will not lead to problems, either.
>>
>> In general, I expect axiomatic triples, as long as they have the form 
>> given in (D), to be semantically harmless (well, mostly harmless :)).
>>
>>> From a spec'ing effort perspective, adding them is an easy task (I am 
>>> offering to do this, should the WG decide to add them).
>>
>> Regards,
>> Michael
>>
>> -- 
>> Dipl.-Inform. Michael Schneider
>> FZI Forschungszentrum Informatik Karlsruhe
>> Abtl. Information Process Engineering (IPE)
>> Tel  : +49-721-9654-726
>> Fax  : +49-721-9654-727
>> Email: Michael.Schneider@fzi.de
>> Web  : http://www.fzi.de/ipe/eng/mitarbeiter.php?id=555
>>
>> FZI Forschungszentrum Informatik an der Universität Karlsruhe
>> Haid-und-Neu-Str. 10-14, D-76131 Karlsruhe
>> Tel.: +49-721-9654-0, Fax: +49-721-9654-959
>> Stiftung des bürgerlichen Rechts
>> Az: 14-0563.1 Regierungspräsidium Karlsruhe
>> Vorstand: Rüdiger Dillmann, Michael Flor, Jivka Ovtcharova, Rudi Studer
>> Vorsitzender des Kuratoriums: Ministerialdirigent Günther Leßnerkraus
>>
> 

-- 

Ivan Herman, W3C Semantic Web Activity Lead
Home: http://www.w3.org/People/Ivan/
PGP Key: http://www.ivan-herman.net/pgpkey.html
FOAF: http://www.ivan-herman.net/foaf.rdf