From: Peter F. Patel-Schneider (pfps@research.bell-labs.com)
Date: 02/15/05
From: "梅婧" <mayyam@is.pku.edu.cn> Subject: Re: SWRL expression in a property instance? Date: Tue, 15 Feb 2005 15:24:04 +0800 > hello, all, > > sorry for bother you all, but, being a friendly reader, i have some points > as > follows: [...] > (2) as Peter replied: > > If you want to write SWRL then you should have separate documents for OWL ^^^ IN RDF > information and SWRL rules. > If you want to embed SWRL in something else, then you should probably not > use the RDF syntax. > > ??? why ??? !!! if it is true, why SWRL has been proposed??? a SWRL document > should contain the OWL part and the rule part together, as we all expected! [I misspoke a bit, as indicated by my addition above.] Well, having combined documents is possible, and non-problematic, using the abstract or XML syntaxes for SWRL (and OWL DL). There would be neither syntactic nor semantic problems using this approach. However, using the RDF syntax introduces problems on both the syntactic and the semantic fronts. On the syntactic front, it is not yet clear how to pick out the SWRL rules from a collection of triples. On the semantic front, there is no RDF-compatible semantic account for SWRL rules written as triples. It may be the case that no such account can be given. Why should it be the case then that the situation for the abstract and XML syntaxes is better? Well, the abstract and XML syntaxes do not let you do a number of things, including using SWRL rules as data. The abstract and XML syntaxes also don't have to treat everything as triples. > and Sesame actually recognized such SWRL documents, since we have tested > one from Protege ontology library: > http://protege.stanford.edu/plugins/owl/owl-library/family.swrl.owl, > meanwhile added the recursive rules Def-hasDescendent, it works!!! we can > draw out the conclusions of hasUncle, hasDescendent etc. from SWRL RDF > triples, based on our coding :) > so --- personally, i do not agree with the separation of the OWL part and > the rule part in a SWRL document! because, it has actually worked > sucessfully! I do not believe that it has. Yes, there may be a tool that parses combined SWRL documents written in RDF. Yes, that tool may actually do something with the SWRL rules. However, does that tool fully implement SWRL? What does it mean to fully implement SWRL documents written in RDF? For example, is the tool able to infer the truth of (the RDF version of) owl:Nothing(?x) => owl:Thing(?x) from an empty document? Should it? If not, why not? If so, why so? You need to come up with a comprehensive treatment of these and other related issues before you claim success. How do you know whether the system has actually done the (or even a) right thing? > Thank you very much! > > Best wishes, > Jing Peter F. Patel-Schneider Bell Labs Research
This archive was generated by hypermail 2.1.4 : 02/15/05 EST