Tuesday, February 21, 2006
Generative Methodology Glass Bead Games
On the limits of the OWL standard à [184]
Reading material [1]
Reading material [2]
Reading material [3]
Summary of the discussion up to this point à [186]
Dialogue about natural ontology and mapping to Information Models
Communication from Andrea
Hello everybody.
I think that talking about "converting OWL to RDBMS" is improper.
First, I would rather say "converting an OWL document to an instance of a relational schema", which seems to be what you desire.
Second, the real translation you need is not just a syntactical encoding of statements as tuples, of classes as tables, or whatever, but also a 1-1 mapping (if you want an invertible translation) between *(SPARQL?) queries over an OWL document*, and *SQL queries over its relational encoding*, keeping into account the semantics of OWL and the semantics of the relational model (Closed World Assumption) in answering the two queries in both contexts.
Let X be an OWL document. Let "enc" be an encoding function such that enc(X) is the relational encoding of X as an instance of a relational schema. Informally, what you want is a semantics-preserving function translating queries over X into queries over enc(X). A formal investigation of the necessary and sufficient conditions to ensure that such a function is "semantics-preserving" [4] is far from being addressable in a short post.
Moreover, I think that a *common* denotation semantics for OWL and the relational model on the one hand, and of SPARQL and SQL on the other hand, should be devises in order to define what does "semantic-preserving" exactly mean. Not considering the CWA/OWA issue, both an OWL document and an instance of a relational schema can be seen as first-order theories, for example, and both SQL and SPARQL can be seen as mechanism for selecting some theorems which are deducible from the axioms of those theories (i'm clearly simplifying, the CWA/OWA issue can not be neglected). That could be a starting point, but I am not able to go further right now.
However, a lot of interesting work has been done about this problem. I can give you some references to the most relevant literature. First, here is a series of slides from Enrico Franconi introducing at a high level the problem of answering queries placed on an ontology by reformulating them into queries over an underlying database. There is no mention of OWL, but the discussion applies to DLs in general, of which OWL(-DL, -Lite) is a syntactic variant. The problem of the LAV (Local-As-View) and GAV (Global-As-View) approaches to query translation is also addressed.
http://trinity.dit.unitn.it/vikef/swap2005/using-ontologies.pdf
The following is a more detailed paper by Calvanese et al. addressing the query rewriting problem. Basically, a DL with relatively low expressive power is defined (a proper subset of OWL-Lite, which the authors - quite misleadingly, in my opinion- called "DL-Lite") that ensures the feasibility of answering a query over a DL KB by rewriting it into a set of SQL queries over a relational encoding of that KB.
http://www.dis.uniroma1.it/~quonto/articoli/calv-etal-AAAI-2005.pdf
Hope this could help :-) Regards,
Andrea
[1] http://dip.semanticweb.org/documents/ECIS2005-A-Methodology-for-Deriving-OWL-Ontologies-from-Products-and-Services-Categorization.pdf
[4] Comment from Paul Prueitt: I feel that the phrase “information-preserving” is better here than the phrase “semantic-preserving”.