Re: RDFTM survey comments

Dear Nikita,

thank you a lot for your comments.

My best regard,
Valentina

>
> I am a bit confused regarding the purpose of this work: Is  
> interoperability being sought on syntax level, data model level,  
> semantic level, etc.?
>

We don't care about syntax. The aim is to obtain interoperability at  
the data level preserving the semantic.

> Even English and Russian languages (those that I somewhat know) are  
> not fully interoperable. There are limitless examples when  
> roundtrips lead to larger and different text when translators try  
> to express idioms of one language with idioms understandable to  
> speakers of other languages. Imagine Shakespeare translated by  
> Boris Pasternak to Russian and then by Walt Whitman back to  
> English. How close will it be to the original?
>

This is a good issue!
The "default" mapping criteria we are defining should work under  
specific hypothesis. That is, considering your example, we don't  
pretend to obtain the original text but to have the same concepts  
expressed. We are defining also a set of use cases that should help  
to understand the usefulness of the mapping and its restrictions of  
applicability, if any.

> And, if (surprise, surprise) the roundtrip result is different from  
> original – does it mean that translators did a poor job?
>

If you refer to semantic differences, perhaps yes!
Consider that a translation performed by an expert user is done in a  
custom way.
The mapping we are defining is in the form of a set of guidelines. It  
gives the possibility to an expert user to customize the translation  
to the extent of such elements whose handling can vary.
If the user performs a bad work.... the mapping will be wrong.

> Topic Maps and RDF authoring is a creative work even if one uses  
> some ready to use generic TM/RDF software that claims to do  
> anything-one-could-ever-ask-for. (And translation as well – is  
> creative) Thinking otherwise is similar to the idea of communism in  
> one separate country.
> (By the way, [Ogievetsky 02] represents the least verbose XTM to  
> RDF translation of all generic translations considered in the  
> survey - IMHO, of course)
>
> I doubt that even a simpler task of translating a topic map  
> expressed in one vocabulary to a topic map in another vocabulary  
> will ever be "perfect" in the sense that authors of the survey  
> imply. (Same apply to RDF documents with different ontology/schemas  
> behind). Syntax level merging is fine, data model level is  
> arguable, semantic - that is where the real problem hides. PSI may  
> partially mitigate it but only in the part which is most obvious.  
> Creative authors will violate any fixed set of PSI, unless  
> restricted by totalitarian rules).
>
>

What we are trying to obtain is not a "perfect" translation, but a  
"correct" one under given assumptions.
We prefer to correctly translate a subset of all possible RDF/TM  
documents instead of approximately translate all possible RDF/TM  
documents.




Valentina Presutti

Phd Student of Computer Science
Department of Computer Science
University of Bologna
Mura anteo Zamboni 7
Ph. +39 (0)51 20 94871
Fax. +39 (0)51 20 94510
home page: http://www.cs.unibo.it/~presutti