- From: Mork, Peter D.S. <pmork@mitre.org>
- Date: Tue, 22 Aug 2006 10:13:09 -0400
- To: "John Barkley" <jbarkley@nist.gov>, "Ivan Herman" <ivan@w3.org>, "Matthias Samwald" <samwald@gmx.at>
- Cc: <public-semweb-lifesci@w3.org>
>>just a small datapoint to what you say. Although I essentially agree, we >>should not forget that even OWL-Lite requires a Description Logic >>reasoning engine (if one wants to use reasoning, that is). > >I also think we should consider a recommended practice of using OWL Lite or >DL where possible (i.e., when the knowledge base can be expressed in no >more >than DL). > >With regard to the point "if one wants to use reasoning", unfortunately, >those who are only interested in querying must be aware of reasoning. For >example, the property "sameAs" is symmetric. Thus, if your knowledge base >has "A sameAs B", one would fully expect the query "?x sameAs A" to return >"B". However, query engines without reasoners won't. Therefore, one must >put >"A sameAs B" and "B sameAs A" into their knowledge base in order to get the >expected results. > Entering "A sameAs B" and "B sameAs A" into the knowledge base (to get the expected results) just means that you are performing forward-chaining rather than backwards-chaining. For smaller knowledge-bases expected to handle a large number of queries, forward-chaining makes sense. My point is that pre-computing a result doesn't allow you to avoid reasoning. You're just making a trade-off between space and time. (Note that once you introduce certain constructs forward-chaining may not be possible because of an undefined fixpoint.) Peter
Received on Tuesday, 22 August 2006 14:16:39 UTC