W3C home > Mailing lists > Public > public-owl-dev@w3.org > July to September 2006

Re: Using RDB for a large OWL-DL database.

From: Bijan Parsia <bparsia@isr.umd.edu>
Date: Thu, 28 Sep 2006 14:10:08 +0100
Message-Id: <824A8578-3118-46D2-97C4-D103EBEB04BB@isr.umd.edu>
Cc: <public-owl-dev@w3.org>
To: "Jinsung Seol" <sjs98ed@blue.knue.ac.kr>

On Sep 28, 2006, at 1:58 PM, Jinsung Seol wrote:

> I cannot help augmenting for my work.
> I tried to transform OWL TBOX scheme into Relational Model that can  
> represent Interpretations one by one.

I don't know how to make this clearer: To the degree that I can  
understand what you wrote, what you wrote indicates that you are  
almost certainly not performing any sort of correct reduction.

>  I think my implementation a little can represent TBOX scheme easily.

This strengthens my certainty.

> ABOX representation is harder than TBOX one,

This even more so.

> because creating tables for each class or individual makes RDB full  
> of tables.

This indicates that you don't understand the ABox/TBox distinction,  
among other things.

> That is my consideration that should be solved.

While it's clear that there is a language difficulty between us, it  
is also clear that you are almost entirely on the wrong track. I  
strongly recommend immersing yourself in the literature for a while.

One indication of a problem is that, for example, in the KAON2  
reduction to *disjunctive datalog* (which is far more expressive than  
the relational algebra), there is the possibility for and exponential  
blowup in the size of the TBox (IIRC). I.e., you might generate an  
exponential number of rules. That, plus the complexity of the  
reduction (including the need to skolemize and saturate) indicates  
that there is not going to be a straightforward reduction of TBoxes  
to relational *tables*.

> You mentioned a large ABOX database in your message. I will try to  
> find a more easily constructing algorithm as much as possible.

  I mentioned many things. But I think you need a better grounding in  
Description Logics first. And databases.

Good luck.

Received on Thursday, 28 September 2006 13:10:31 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:58:14 UTC