Fw: Why Literals should be unique and why this is a serious issue

----- Original Message ----- 
From: Jeremy Wong 黃泓量 
To: Andreas Andreakis 
Sent: Sunday, November 20, 2005 10:34 PM
Subject: Re: Why Literals should be unique and why this is a serious issue


If you want to apply such a unique constraint, you should define your own vocabulary.

OWL is to confine your search space when the search space of RDF is so unlimited. An inconsistent reasoning of OWL means that the search space is empty. I am not really sure if augmenting the UNIQUE constraint would help in confining search space.


Jeremy Wong
  ----- Original Message ----- 
  From: Andreas Andreakis 
  To: Jeroen van der Ham 
  Cc: semantic-web@w3c.org 
  Sent: Sunday, November 20, 2005 8:46 PM
  Subject: Re: Why Literals should be unique and why this is a serious issue



  Fine, you are pointing the SemanticWeb Stack, OWL over RDF, which is actually very comparable to TCP over IP (for instance IP does not garanty delivery, while TCP does)

  So IP is very flexible, just like RDF right ? So you say to use OWL if it suites you to cover the flexibilty of RDF. Im fine with that, and I absolutely agree that we need flexibility in the down layers and so in RDF.

  But there is a dramatic error. In the SemanticWeb, what counts is the Knowledge you define (in an Ontology). On top of your Knowledge Base you are free to choose a specific inferencing so you can produce more knowledge - in example transitifity of subClassOf produces additional triples to cover all over-under classrelations. But (and here is the error) you are NOT able to extract additional information / knowledge out of an Ontology, if the ontology is not suitable. 

  Try to get additional subClassOf Information out of an Ontology, which does not contain any Class (and subClassOf) definitions. Try to to identify Resoures on Values in an Ontology, which does not specifiy InverseFunctionalProperties. The required results you will get will be 0, simlpy because the Information is not there (and therefore no inferencingrules can be applied). You are not able to cover the simplicity of an RDF Ontology using OWL Inferencing.

  I know that (in your reply) you said: " in which case RDF is not for you, but OWL might be something more akin to what you want". This is very correct. But what are you going to do, with ontologies which are already described using "super simple" and "super flexible" pure RDF ? How are you going to cooperate RDF Ontologies with Ontologies, which are well defined using lets say OWL Full (incl. InverFunctProperties).

  Before hearing people telling: Of course you can cooperate in some way - Yes. Dont forget that we are basicly talking about resource-dublication (see first email-post).

  We feel very comfortable "now" (and me as well), but currently everythink is still very simple. Will you keep the overview if thausands of Ontologies (multi) interconnect to each ?? 
  Will you tell people who described their resources using RDF: "Hey you guys, you must use OWL now, we not more precision now !" ? 

  Without wanting to disrespect any of you guys, I think that same-concept dublication is is not taken serious enough and IMHO this can become very tricky in the long-term.


  If you mean that the SemanticWeb Stack is not supposted, for this kind of circumstances -  I think you might be correct!!! The question is, if this is a good solution.
  We will be forced to cover such circumstances in an higher level of abstraction...someday. If you seek for appropriate contributions on this field (such as research-papers) Im prommising you that this is not fun, since automious merging of equiavalent knowledge is a very difficult topic. The problems which could arise in this field, might even not be solvable.

  So is there a chance to prevent dublication-hell earlier ? Do you think I´m a panic-maker ? Maybe you are, but I still have not heard constructive suggestions on how to solve this problems. I´m still looking forward to it. 


  Do I think there is a solution ? In my IMHO yes, but this might also include a refactoring of the SemanticWeb Stack and its definition. (you can also see previous posts on this topics for some suggestions)


  best regards,
  Andreas



  Jeroen van der Ham schrieb: 

    Andreas Andreakis wrote: 

      The question is how can we prevent same-concept dublication ? 



    Excuse me for being blunt, but I really don't see why you would want or need this in RDF. 

    If you really want to prevent duplication, you have to use OWL Full, as someone already explained. I don't see why or how you would prevent this in RDF. 

    IMO RDF was designed to be a loosely formatted way of describing resources. There is no limit to what you can describe or how. This also means that you are almost completely agnostic of what is being described. And this means that there is no way to really prevent a 100% fault-proof method of preventing concept duplication, unless you are really meticulously defining what you are talking about, in which case RDF is not for you, but OWL might be something more akin to what you want. 

    And about rdf:id, it has been posted on this list before, but it really doesn't mean much to declare something rdf:id. If you use it and convert it to triplets, then the fact that it should be a unique id is lost. 
    It only gets in your way if you want to extract things from the XML/RDF notation, because you have to account for both rdf:id and rdf:about. 

    Jeroen.