Re: semantic data models

Thanks a lot Danny
(leaving the pragmatic side of thing behind for a moment, but it would be 
interesting also to see
how a pragmatic data model would compare to a semantic data model )
I am reading through mails superficially now, just a few quick comments
P

----->
> ...if you discount Semantic Web technologies ;-)

yes - we are still trying to understand the practical applications of those 
(semijoke)

>
>> questions:
>>
>> - other than SIM, prototypes like Hstore and some semantic data wrapper,
>> what tools can I use to implement semantic data?
>
> Correct me if I'm wrong, but I think the original notion of semantic
> data was the overlay of ontology kind of ideas over an
> entity-relationship kind of model. There's been a great deal of work
> done since around Description Logics (which offer formalisations of
> ontologies) and the Resource Description Framework offers an
> entity-relationship model designed for the Web. Put these together,
> with a few other bits and pieces and you have Semantic Web
> technologies.

I am trying to understand what work I need to look up... still reading

>
>> - when modelling data, usually we follow standard 'data modelling' 
>> methods
>> based on relational data model. what parts of our methods do we have to
>> upldate to develop semantic data model?
>
> Depends how semantic you want to get.

How semantic do I need to get? Assume I want to get totally semantic (even 
pragmatic, after we know what that is exactly)

 Down at the RDF level, you can
> just normalise the data down to binary relations (and name the
> relations with URIs). This can be done as a view of the relational
> data. The SPARQL query language can be used on such data, and because
> of the use of URIs as identifiers it's inherently Web-friendly.

sounds good -
assume I have all my enterprise data stored in a standard RDBMS
which tool can I use to 'normalise the data to binary relations' (and what 
form would that be?)
Also, I am under the impression that SDM require a degree of 
de-normalisation (allowing for derived data)
but normalisation down to binary sounds like highly normalised
can you explain that


Do you have access to this paper? I cant seem to get it from my ACM 
connection here:

Normal Forms and Reduction for Theories of Binary Relations


I think from what I understand so far that SDM extends RDM with additional 
properties that are also shared with OODM, but
I have not studied how semantic technologies should be used to be applicable 
with normal datasets


> If you want more sophisticated reasoning over the data, you'll need to
> find or create OWL ontologies in which your data fits.

I see. I think that semantic data models requires an ontology to be 
deployed, but I would not want to rescrict ontology to owl
as we have it now, as I think owl may be extended at some stage
Id like to have an ontology that is owl independent.  but I guess thats what 
you mean (I guess you say ontology=owl while I say
ontology= any KRL)

>
>> - As a project manager in charge of making sure that my organisation's 
>> data
>> model is 'semantic enabled' what should I absolutely make sure of?
>
> Depends what you want to do with the data. I'd suggest exposing an RDF
> view of the data, reusing existing RDF schemas/ontologies wherever
> possible. See the specs listed at:
>
> http://www.w3.org/2001/sw/
>
> also:
>
> http://sites.wiwiss.fu-berlin.de/suhl/bizer/pub/LinkedDataTutorial/
>
> To make it useful internally, you may want to check out some of the
> available tools, e.g. at:
>
> http://www.mkbergman.com/?page_id=325
> http://sites.wiwiss.fu-berlin.de/suhl/bizer/toolkits/
> http://planetrdf.com/guide/


thanks a lot Danny - I am collecting points of view at the moment

most useful
best
pdm




> Cheers,
> Danny.
>
> -- 
>
> http://dannyayers.com

Received on Friday, 28 December 2007 06:03:03 UTC