Re: Review of Semantics Documnt

From: Peter F. Patel-Schneider <pfps@research.bell-labs.com>
Date: Tue, 14 Jan 2003 03:08:43 -0500 (EST)
Message-Id: <20030114.030843.125112962.pfps@research.bell-labs.com>


Reviewer Name: Jeff Z. Pan
> Affiliation: Computer Science Dept., University of Manchester
>
> Document title:
> Web Ontology Language (OWL) Abstract Syntax and Semantics (Editor's Draft 2
> January 2003)
>
> Editors:
> Peter F. Patel-Schneider, Bell Labs Research
> Patrick Hayes, IHMC, University of West Florida
> Ian Horrocks, Department of Computer Science, University of Manchester
>
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>
>  Summary:
>
>  This document provides a high-level, abstract syntax for OWL/Lite and
> OWL/DL, and gives a clear direct model-theoretic semantics to the languages.
> Since RDF/XML is selected as the normative exchange syntax for OWL, a
> mapping from the abstract syntax to RDF graphs is also provided, together
> with the RDFS-compatible model-theoretic semantics.
>
>
>
>  This document was generally of very good quality. The structure is very
> clear, and the work is solid and important. Most parts of the document are
> relatively easy to understand.
>
>  The only part that is not so easy to follow is section 5.2. Most of the
>
>
> **1. Something I don't quite understand:
>
> - section 5.2, why does owl:complementOf have only-if, but not iff
> characterization, while sameClassAs and disjointWith etc have iff
> characterizations. I become wondering what will happen if we change e.g.
> sameClassAs to only-if characterization, does the reasoning rely on the
>
The OWL vocabulary that corresponds to DL constructors is given a
different treatment from the OWL vocabulary that corresponds to
(other) semantic relationships.  The former have only-if semantic
conditions and comprehension principles and ; the latter have iff
semantic conditions.  The only-if semantic conditions for the former
are needed to prevent semantic paradoxes and other problems with the
semantics.  The comprehension principles for the former and the iff
semantic conditions for the latter are needed so that useful
entailments are valid.

I've added the above wording to the document.

>
> **2. some typos:
>
> - section 5.2, I(E) should be S_I(E), or atleast introduce/define I(E)
> somehow like that in RDF MT.

Changed to S_I.

> - section 5.2, "RDFS domains and ranges" part, column 3 of row 1 in the
> table, "CEXT" should be "EXT".

Fixed.

> - section 5.2, "R_I contains elements corresponding to all possible OWL
> descriptions and data ranges" part, in the sentence "The first three
> conditions ... ", why "three", aren't there two? (It should be three if we
> consider the third item of part **3.)

There used to be three.  Because of a change, the third one was
removed.  I've fixed the wording.

> - Appendix B.1, first example line 4, is "value" missed in front of
> "(ex:author ..."?

Correct, and also missing before ex:name.  This was a recent change
to the abstract syntax.  There are other places in the document that
needed a similar fix.

> - Appendix B.1, second example line 2, is "intersectionOf" missed?

No.  There is an implicit intersectionOf here.

> **3. minor bugs:
>
> - section 5.2, "Some OWL properties have only-if characterizations", the
> last rule of the second table, "if E is owl:oneOf ...", this is not common
> portion of the semantics of OWL/DL and OWL/FULL, thus should be moved to
> section 5.4. Also in this rule, it seems that "RI OR LV" should be "RI" or
> "IOT" (in OWL/FULL they are the same), since LV is already a subset of RI.

This condition was placed in the common section so that OWL/DL and
OWL/Full would be very small changes.  It does add more to OWL/DL
than might be expected.  The change from R_I to IOT is needed,
however, to make all this work.

Because of the recent changes to the RDF model theory, a better
version of this can be formulated - one that does not add extra
oneOfs to OWL/DL while still working for OWL/Full.  I have made this
change.

> - section 5.2, "Some OWL properties have only-if characterizations",the
> fourth table, "y\in IOC" should be "y\in IOC OR IDC".

Correct. Changed.

> - section 5.2, "R_I contains elements corresponding to all possible OWL
> descriptions and data ranges", if there exists l, a sequence of x1,...xn
> over IOT OR LV, then there exist y with y\in CEXT(I(rdfs:Class)), <y.l>\in
> EXT(I(owl:oneof)): I think this rule should be moved to section 5.4, and in
> here (section 5.2) should be replaced by two, one for IOT and the other for
> LV, otherwise if x1,...,xn are mixed objects/values, it will cause trouble
> to OWL/DL.

I believe that there is no problem here. These elements of
rdfs:Class would not cause a problem, as they cannot particate in
most OWL/DL constructs.  However, the recent changes to the RDF
model theory make a better version possible.

> - Appendix B.2, the three antecedent triples don't entail Susan is an
> object, and friend is an individual property.

You and I misread this section of the document.  I've made it
clearer.

>
> **4. Some suggestions:
>
> - section 5.2 "Relationships between OWL classes", it is kind of weird to
> have  two (or more than two) tables as conditions without any explanations.
> Instead, it might be better if we explain that the relationships between OWL
> classes may be divided into subset relationships (the first table) and
> membership relationships (the second table).

This has all been reorganized.

> - section 5.2 "Characteristics of members of OWL classes": it might be a
> good idea to have some descriptions before the table. In general, some more
> texts are needed to make section 5.2 easier to understand.

I've reorganized the tables and changed the text.

> - section 5.3, second paragraph, "V'" includes VRDFS, while in section
> 5.3.2, the last sentence of second paragraph, "V'" is disjoint from VRDFS.
> It might be a good idea to change the first one to "V".

Done.