- From: Smith, Michael K <michael.smith@eds.com>
- Date: Wed, 18 Sep 2002 11:10:21 -0500
- To: Jim Hendler <hendler@cs.umd.edu>, Christopher Welty <welty@us.ibm.com>, www-webont-wg@w3.org
In the context of the wine onotlogy and looking as Guus' region example I was thinking a little about distributed geographical ontologies. The following scenario seems simple and desirable: Assume GeoOnt is an ontology for regional geographcial descriptions, including countries, regions, states, provinces, cities, towns, and their relations. Assume the following import GeoOnt. TX-Ont - Definition of entitities in the state of Texas OK-Ont - Definition of entitities in the state of Oklahoma And so forth for other geographical areas. Now both TX-Ont and OK-Ont are going to want to refer to adjacent states. The authors of TX-Ont can refer to Oklahoma by 1. Defining a new identifier, txont:Oklahoma, that is unrelated to okont:Oklahoma. If someone chooses to use both TX-Ont and OK-Ont together, it will be up to them to identify all of the correspondences and create the needed equivalences. 2. Use okont:Oklahoma, with a okont: namespace declaration, but no imports. This does permit inconsistencies to arise, but that's life in a distributed, web-based world. 3. Import the entire OK-Ont ontology. In a well-connected but distributed use of GeoOnt, a consequence of this would be that in order to reason about traveling from Austin to San Antonio (both in Texas) I would import the transitive closure of the entire world's geographical ontology. Possibility 2 seems simplest. It enables another party to import both TX-Ont and OK-Ont when they need to and reason about them without having to jump through the hoops in 1. AND, they don't get the complete geography of Siberia clogging up their system as an inevitable consequence. - Mike -----Original Message----- From: Jim Hendler [mailto:hendler@cs.umd.edu] Sent: Wednesday, September 18, 2002 10:12 AM To: Christopher Welty; www-webont-wg@w3.org Subject: Re: LANG: owl:ontology At 9:49 AM -0400 9/18/02, Christopher Welty wrote: >Jim, > >In your example below, that the builder of URI2 may not have "intended" >for all of URI1 to be loaded when importing URI1:foo is too bad. One of >the most common and (I think) important reasons to build a million-node >graph around a set of symbols to to ensure those symbols are used >correctly, and to prevent them from being used incorrectly. The ontology >helps to enforce that the intended meaning is maintained. > >If someone wants to import just one symbol from my ontology, then they >have to get all of my ontology so that they get all the constraints and >inter-relationships that help clarify the meaning of that symbol wrt >others. I *don't care* if that's what the user intended - that's what I, >the designer of the ontology, intended: dont misuse my symbols! > >If this isn't enforced, then borrowing symbols from other ontologies is >meaningless and serves no purpose. There is nothing to gain from using >URI1:foo if you aren't getting all of the ontology around URI1:foo - you >may as well just make up your own symbol in URI2, because that's all it >is. If you are borrowing URI1:foo because "you mean the same thing", then >importing the whole ontology that enforces that meaning should have no bad >consequence. > >Of course ontologies can always be built in a modular way so that if there >is a small subset of the symbols that can be exported idependently of >others (a top level ontology would be one kind of example) then they are >all in a smaller ontology that is imported, but that is more of an >engineering issue. > >-Chris Chris, what can I say except I can't disagree more (but am willing to look for compromise). Here's real world experience - one of my students wanted to put on her web page that she had a stuffed penguin on top of her monitor. She eventually created the following in her ontology >:isOnTopOf a rdfs:Property; > rdfs:Domain cyc:ExistingObject; > rdfs:Range cyc:ExistingObject. > >:Penguin a daml:Class; > rdfs:subClassOf cyc:Bird. > in her instances information she asserted >:penguin1 a our:Penguin; > our:isOnTopOf :mymonitor. > So she has linked her stuff into CYC in some sense, but it clear to me that I must be able to read her ontology and use it without having to bring in the entirety of the HUGE cyc graph. Now, we can get into a deep philosophical issue - do we want to say she intended to include all the semantics of open cyc (answer: she said "no" :->) or more importantly, should the CYC folks be saying by packaging it all together it is all of nothing. But what I'm after is that she found something useful, it made it easy for her to express what she wanted on the Semantic Web, and I believe strongly (my drop dead issue) that we want to encourage this kind of use (see my article at http://www.cs.umd.edu/~hendler/AgentWeb.html which has some suggested use cases for ontology extension, modification, linking, etc). In short, I see this as a use case that drives what Peter said: At 10:40 AM -0400 9/18/02, Peter F. Patel-Schneider wrote: > >As far as I can see the only viable route is to be able to use resources >without committing to anything related to that resource. To commit to >something in some other ontology/document, use imports. -- Professor James Hendler hendler@cs.umd.edu Director, Semantic Web and Agent Technologies 301-405-2696 Maryland Information and Network Dynamics Lab. 301-405-6707 (Fax) Univ of Maryland, College Park, MD 20742 240-731-3822 (Cell) http://www.cs.umd.edu/users/hendler
Received on Wednesday, 18 September 2002 12:10:35 UTC