Re: Ontology Wars? Concerned

Hi Nathan

I think you have to try harder than that to cause offence!

I think an attempt at standardisation on the one 'true' set of  
ontologies is futile, not scalable and ultimately a dead end.   
However, using suitable existing ontologies in a sensible way leads to  
a kind of lingua franca of properties and classes that supports a  
useful interchange.

If you are both creating the markup and writing the SPARQL then you  
don't really have a problem.  The difficulty I suppose is trying to  
maximise the chance of independently created query tools and data sets  
talking to each other meaningfully.  I don't think there is a miracle  
cure.

As Kingsley explained, you can define mappings between ontologies -  
and these mappings might be for a specific purpose, maybe not  
universally valid.  Such a mapping  combined with some reasoning can  
increase the capabilities of automatic tools.  Ultimately, for maximum  
accuracy, some human input is probably going to be required for data  
normalisation and/or query design.  But whatever happens, by using  
linked data your properties are uniquely identified and someone can  
track down what the properties mean and how they are used, so  
interoperability is possible, if not always 'for free'.

Bill


On 18 Nov 2009, at 20:23, Nathan wrote:

> Hi All,
>
> Many thanks so far for the invaluable input I've been getting from the
> community; I may be about to commit the cardinal sin here, but I'm a  
> bit
> concerned and only saying this with the best intentions.
>
> Before I start, if I can be considered an early adopter then please do
> disregard the rest of this mail.
>
> I'm finding the path to entry in to the linked open data world rather
> difficult and confusing, and only for one specific reason -  
> ontologies;
> it /feels/ like there are some kind of ontology wars going on and I  
> can
> never get a definitive clear answer.
>
> Perhaps I'm missing something, but the primary focus for me is to use
> ontologies that people will be using in SPARQL (or alternative  
> language)
> queries. Anything else appears to be a waste of time.
>
> Multiple properties in multiple languages that appear to describe the
> same thing make no logical sense to me whatsoever, and questioning my
> own programming capabilities here, I don't see how they will to a
> machine either.
>
> To put it in real terms, all I need to do is describe the relations
> between a URI and multiple other URIs, yet this is the blocker in my
> current project? - and it's really nothing complex (or shouldn't be);
> I've managed to get a grip on all the various concepts and formats,
> software, tools, methods, get everything set up, start consuming lod  
> and
> correlating plain text entries up to URIs, yet still choosing which
> properties / ontologies to use is the blocker :(
>
> Please do tell me if this is just me missing something simple, if I  
> can
> simply write up my own ontology and everybody else will be able to
> consume the data without any input from me; or me informing the  
> world of
> the new ontology so they can consume and handle it, then great; but if
> not then surely this is a problem?
>
> My concern is on two levels here;
>
> 1: that my own path to entry is being slowed and indeed blocked by
> confusion over ontologies.
>
> 2: that many other people will find the same problems (or worse) and  
> it
> could potentially be a show stopper for something so important.
>
>
> Just to clarify, I'm not dealing with big custom data sets here, I'm
> coming at LOD from the "normal" developer angle, writing systems that
> publish articles and such like; and whilst describing things like
> titles, authors, publish dates etc is all nice and simple ontology  
> wise;
> I'm finding that describing what the content is about is virtually
> impossible - tags and subjects just don't cut it & the level of
> description of relations needs to be somewhat more fine-grained to  
> be of
> any use.
>
> Many Regards & do hope I've caused no offence;
>
> Nathan
>

Received on Wednesday, 18 November 2009 19:55:33 UTC