- From: <tim.glover@bt.com>
- Date: Fri, 7 Apr 2006 16:51:13 +0100
- To: <henry.story@bblfish.net>
Henry Story notes: I am redirecting this message to the SemWeb list, as I do think that discussing concrete problems is a lot more useful than most of the rest of the conversation we have been having recently here. I am sure people on the Sem Web mailing list would much rather have their inbox filled with concrete problems than with high browed nonsense. I won't be able to answer this immediately though, as I have a tight deadline to meet at work. ---------------------------- Hi Henry, I am answering this off list, because I am not sure everyone wants their mail box filled with this discussion :) First, thanks for your reply, and for putting some effort into it. I always find these discussions fruitful. It is true that it is possible to identify some correspondences between these ontologies, which you have done. Of course, it is also true that I could have arranged it so there were no correspondences at all. Specifically, if Class1 is A union B and Class2 is A union C you cannot take an instance of Class1 and tell me whether or not it is in Class2, or vice-versa. I suppose the point you are making is that given Peter is o1:Married and Paul is o3:Married, as a worst case you can "merge" the ontologies by simply keeping the triples distinct. I suppose the point I am making is that it is not very satisfactory to have a "merged" ontology with two different concepts of what it is to be married or unmarried, and that in hindsight it would have been better to have placed the instances in a more general ontology which included the concepts of "now", divorce and death. So I think I am saying that * It is quite legitimate to construct your own ontology, and collaborate with other agents (people) over the web to populate it. * You don't have to refer to anyone else's ontology to do this. It is enough that it is useful. * I can construct a different ontology in the same area, and "may the best agent win". * But you are most unlikely to be able to merge the ontologies in a meaningful way unless both are based on a common parent. (I don't count simply whacking all the triples into a single database as meaningful merging). Tim.
Received on Monday, 10 April 2006 07:08:33 UTC