- From: Jeremy Carroll <jjc@hplb.hpl.hp.com>
- Date: Sat, 19 Jan 2002 12:58:56 -0000
- To: "Patrick Stickler" <patrick.stickler@nokia.com>, "ext Sergey Melnik" <melnik@db.stanford.edu>
- Cc: <www-archive@w3.org>
Sergey, Patrick, this discussion presupposes a Model Theory for TDL which Patrick has explicitly not furnished. Of course I am going to solve the issues with a stroke of a magic wand. When the magic happens either be amazed and applaud or (more likely) offer criticism at that point! Jeremy > -----Original Message----- > From: Patrick Stickler [mailto:patrick.stickler@nokia.com] > Sent: 19 January 2002 10:16 > To: ext Sergey Melnik > Cc: www-archive@w3.org; Jeremy Carroll > Subject: Re: Problem with TDL (EVEN BIGGER ONE) > > > On 2002-01-18 22:41, "ext Sergey Melnik" <melnik@db.stanford.edu> wrote: > > > Patrick, > > > > thank you for the clarification. In fact, your approach suggests a third > > original idiom, in which typed elements are interpreted as pairs (value, > > lexical representation). I'm going to accommodate this idiom in the > > datatyping document that I'm editing. > > > > However, under the above interpretation, your document [1] is > > inconsistent. The figures in 3.1 and 3.2 can only be consistent with > > each other, if literals (rectangles) are interpreted as pairs. This is > > fine, but you run into the issue that the interpretation of literals is > > context-sensitive (i.e. not fixed, but depends on the property that the > > literals are handing off). So, for example, the literals hanging off > > rdf:value are now also interpreted as some kind of pairs. > > I'm not sure I follow you. The idiom requires that rdf:value has > a literal value. It is not a recursive idiom. You can't have > an anonymous node with an rdf:value that is itself an anonymous > node with an rdf:value.... > > Perhaps this is not specified clearly enough in the discussion > of the idiom. I will try to make that clearer. > > Note though, that this is an issue with the idiom, not the TDL > model. > > The separation from the TDL model and the idioms employed is > intentionally strict -- so that each idiom is simply a means > of expressing a pairing, and the pairing itself is the foundation > of interpretation. > > > Furthermore, your scheme *does* require modifications the the present > > RDF graph model: it is based on untidy graphs, quite in contrast to what > > you claim! In other words, you cannot just merge two literals with the > > same content together (they might represent two different pairs, > > depending on the properties they are hanging on). > > My understanding of graph tidying is that nodes with literal labels > and anonymous nodes are never merged. Only nodes with URIref labels. > > Thus, you would never merge two literals. Every literal node (every > instance of a lexical form) is unique in the graph. > > If this is not correct, then that's *major* news to me, and I'll > have to go back and re-re-re-re-read the MT and other relevant > discussions. > > Thus, per my present understanding, TDL is based on tidy graphs, > and both of the defined idioms are valid for tidy graphs without > the problems you indicate. > > Also, again, if it turns out that I am mistaken about the tidying, > it is only an issue of the particular idiom, and not of the TDL > model itself. > > Regards, > > Patrick > > > All in all, given your clarification, your scheme is a specialized > > variant of Patel-Schneider's/Hayes' one, but with some false claims... > > > > Sergey > > > > [1] http://www-nrc.nokia.com/sw/TDL.html > > > > > > Patrick Stickler wrote: > >> > >> On 2002-01-18 4:42, "ext Sergey Melnik" <melnik@db.stanford.edu> wrote: > >> > >>> Patrick, > >>> > >>> there is a well-known problem with the rdf:type/rdf:value > representation > >>> that you described in > >>> > >>> http://www-nrc.nokia.com/sw/TDL.html > >>> > >>> In essence, the interpretations of nodes _1 and _2 in > >>> > >>> _1 rdf:value "3.0" > >>> _1 rdf:type us-double > >>> _2 rdf:value "3,0" > >>> _2 rdf:type german-double > >>> > >>> fall together (_1 and _2 represent the same real value), and > the pairing > >>> between the lexical representation and the type gets corrupted. > >>> Specifically, the semantics of the above four statements is equivalent > >>> to that of > >>> > >>> _3 rdf:value "3.0" > >>> _3 rdf:value "3,0" > >>> _3 rdf:type us-double > >>> _3 rdf:type german-double > >>> > >>> How do you suggest to work around this problem? > >> > >> There is no problem. > >> > >> The nodes _1 and _2 would *never* be merged in that fashion. > >> > >> 1. They do not denote the same value space member of the same > >> datatype as us-double is not the same datatype as german-double. > >> > >> 2. Neither anonymous nodes nor literal labeled nodes participate > >> in graph tidying. > >> > >> The fact that the TDLs ("3.0",us-double) and ("3,0",german-double) > >> may correlate to the equivalent member in each other's value spaces or > >> in the value space of e.g. xsd:double is a matter > >> of interpretation -- not of representation. > >> > >> E.g. > >> > >> <rdfl:LexicalDatatype rdf:about="xsd:double"/> > >> > >> <rdfl:LexicalDatatype rdf:about="us:double"> > >> <rdfl:lexicalSubClassOf rdf:resource="xsd:double"/> > >> </rdfl:LexicalDatatype> > >> > >> <rdfl:LexicalDatatype rdf:about="german:double"> > >> <rdfs:subClassOf rdf:resource="xsd:double"/> > >> </rdfl:LexicalDatatype> > >> > >> which defines, as we would expect, that the value > >> space of both us:double and german:double are subsets > >> of the value space of xsd:double but only the lexical > >> space of us:double is a subset of the lexical space > >> of xsd:double. > >> > >> (note the difference between rdfs:subClassOf and > >> rdfl:lexicalSubClassOf!) > >> > >> And if we know that the value spaces of us:double > >> and german:double are in fact a perfect intersection, > >> then we can also say > >> > >> <rdfl:LexicalDatatype rdf:about="us:double"> > >> <rdfs:subClassOf rdf:resource="german:double"/> > >> </rdfl:LexicalDatatype> > >> > >> <rdfl:LexicalDatatype rdf:about="german:double"> > >> <rdfs:subClassOf rdf:resource="us:double"/> > >> </rdfl:LexicalDatatype> > >> > >> which says that any member of the value space of > >> us:double is also a member of the value space of > >> german:double and visa versa. > >> > >> Thus, there is no problem. > >> > >> The nodes in your example would never be merged, and > >> the relationships between the lexical and value spaces > >> of the datatypes in question are explicitly defined, > >> and interpretation about whether the two TDLs > >> ("3.0",us-double) and ("3,0",german-double) denote > >> the same value happen outside the RDF-space, yet > >> the RDF representation -- using the TDL scheme -- is > >> fully sufficient for making that interpretation. > >> > >> Cheers, > >> > >> Patrick > >> > >> -- > >> > >> Patrick Stickler Phone: +358 50 483 9453 > >> Senior Research Scientist Fax: +358 7180 35409 > >> Nokia Research Center Email: patrick.stickler@nokia.com > > -- > > Patrick Stickler Phone: +358 50 483 9453 > Senior Research Scientist Fax: +358 7180 35409 > Nokia Research Center Email: patrick.stickler@nokia.com > > >
Received on Monday, 21 January 2002 05:05:29 UTC