Re: Blank Nodes Re: Toward easier RDF: a proposal

Thanks for bearing with me/us, William.

> On 3 Dec 2018, at 23:34, William Waites <wwaites@tardis.ed.ac.uk> wrote:
> 
> 
>> Having an oracle that defines what an unambiguous Thing looks like is one
>> organisational structure… However I (and perhaps David Booth) am after
>> something more anarchic,
> 
> I’ll admit to being a bit confused about what you’re getting at. Surely if someone
> defines a type for a thing, and gives it a URI, then they are in a position to say
> what kind of thing it is and what the type means. Hopefully they’ll convey their 
> idea of the meaning of the type (including things like equality) in a machine 
> readable way. And maybe people who have data of that kind will like it, or they’ll
> use some other type that they think fits better. Some kinds of things are harder
> to define than others — I would not want to be the person to work out the details
> of street address equality — but that’s ok and not even surprising. 
> 
> How does that connote a monolithic oracle? Someone has to mint URIs for types
> of things and stand behind them, and everybody is free to choose whatever 
> types they like best, or make their own. Sounds like the original anarchic idea.
Fair enough.
The reason I was talking about an oracle is because someone was suggesting that schema.org might drive the Thing equivalence world in general.

And yes, URI minting is the original anarchic idea.
But what I am after is trying to get web benefits without always going through all the stuff in your longer first paragraph.
Essentially, if we happen to independently choose the same way of generating similar URIs for Things, then developers can work more efficiently, and still get the benefit of integration with other datasets that means that they benefit from the SW/LD world they have put the work into inhabiting.

Another way of looking at it is that a developer can sign up to the idea that if they choose to generate a URI from input data so that it colludes (or whatever) with the URIs from the same input data from other datasets, then that choice captures a context for type etc (as you describe) and does the search/lookup without even looking outside the subject dataset, never mind having to go to the time and expense of looking at the machine readable meanings.
Dangerous? In general, yes. But often pragmatically useful, in my experience.
For example.
If I use common algorithms to normalise citations so that I make statements about
https://example.org/paper/the-semantic-web-revisited---n-shadbolt-t-berners-lee-w-hall (or a hashed version)
Then there is a decent chance that others will use the same URI for their reference to the same paper.

When I compare that with the process we usually need to go through, of finding source datasets, seeing if they offer an API to do lookup, reading the ontology to try to understand what the properties mean, etc., I think that maybe it is useful.
"You mean I have to do that to get the benefit of the Web bit of Semantic Web?!"
Perhaps DBLP et al could join this world by using such URIs.
They are already close - that paper is "ShadboltBH06" on DBLP.
After all, Wikipedia (and hence DBpedia) does it for us that way, and we are used to guessing the URILs for Wikipedia pages.
IN SW we can go that much further and better because when constructing a URI we have all the RDF goodness in the graph around to choose good inputs to the URI construction process.

Perhaps it is trivial, and certainly not a big technical challenge - but I think it may be a significantly different way of doing stuff (that quite a lot of us are already doing), and different from the way I often see things argued and supported.

> 
> I get the feeling that I must be missing something very basic here…
I hope I have done a bit better here, without just repeating myself.

Best
> 
> -w
> 
> 

-- 
Hugh
023 8061 5652

Received on Tuesday, 4 December 2018 23:42:56 UTC