Re: RDF Graphs

On Oct 27, 2014, at 5:40 PM, Hugh Glaser <hugh@glasers.org> wrote:

> Thanks to all.
> So here’s my problem, and I still sort of have it.
> The meaning of the word “Graph” in “RDF Graph” is not something that people normally recognise - but it doesn’t say that anywhere that I have found.
> To be honest, it all feels rather sloppy.

If by sloppy you mean, it is not written for mathematicians to read, then yes, it is sloppy. 
> 
> The documents and everywhere use the term “RDF Graph”.
> And in fact they use terms like nodes and edges (as do many of the responses here).
> 
> But no matter how many times people say the set of triples are a Graph, it doesn’t make it so!

It depends what you mean by "graph". This word does have a meaning outside of mathematical graph theory. A set of RDF triples can be viewed as a  a network of nodes and arcs on a page, an actual diagram. Calling this a graph does not seem to me to be unreasonable. Calling it a colored pseudograph would not be helpful to most readers. 

> That is, unless you define what you mean by Graph to be consistent with that view.
> And nowhere can I find (in the documents or in any postings) what is the formal definition of a Graph, as the term is used in the phrase RDF Graph.

The term that gets defined is "RDF graph". 

> (Yes, Pat, I do try to read the normative standards documents where I can - and in fact I search as well, so for example I foundhttp://answers.semanticweb.com/questions/8726/rdf-graph-is-not-defined-as-the-mathematical-concept-of-graph
> :-) )
> http://www.w3.org/TR/rdf11-concepts/#section-rdf-graph certainly doesn’t define anything that looks like a mathematical graph to me.

It defines the term 'RDF graph', which is the term you were using. What more would you expect it to define? 

> Now, I am usually happy to be fuzzy, but I am a bit surprised that this list is - I chose the SemWeb list rather than LOD because I thought that I would just get back some maths that would explain it! :-)
> I probably expected someone to tell me what is the isomorphism between sets of triples and a particular sort of Graph.

Why would you expect mathematics? Did you think that RDF was a branch of topology? What you should reasonably expect, and I believe you actually get, is precision. 

> All the doc really says is "An RDF graph is a set of RDF triples.”.
> Now, that seems problematic to me.

To me it seems as clear as it could possibly be. It takes a previously undefined technical term and gives it a crisp, unambiguous, definition. 

> Because what was meant to be a good intuitive description for people - that RDF is a graph of sim[ply nodes and edges, doesn’t really cut it.

Surely it is obvious to even the most basic intuition that RDF does not correspond to a simple graph in the sense of graph theory, which has no labels or other structure. But then most things called "graphs" are not simple graphs in the sense of graph theory. In fact, graph theory is a rather arcane topic having very little to do with the totality of things commonly called "graphs", just as string theory has very little to do with things commonly called "string" and knot theory has almost nothing at all to do with things commonly called "knots". You might also think about whether mathematical groups, rings and fields have very much to do with what are commonly called groups, rings and fields.  

> Of course, it goes on to say "It is possible for a predicate IRI to also occur as a node in the same graph.” is rather strange.
> And this confusion has an effect, because I find that people who don’t read the source material don’t realise they can do that - even though every significant ontology does that.

Frankly, I don't really give a damn about people who don't read the source material. Similarly for people who have fatal accidents with chain saws because they don't read the user manuals. 

> And another problem is that I then find a risk of confusion over the scope spaces of the IRIs as node or edge labels. If I have nodes with labels, clearly nodes with the same labels are the same node in a graph. But what of IRIs that are both node and edge labels?Can you consider thing that have the same label, but different type, to be the same?

Well, the IRIs themselves are, I would venture, *obviously* the same. 

> I have searched quite hard, and can’t find any other context in which such a set of triples is considered to be something called a Graph.

It is not called a Graph here either. The term is "RDF graph". 

> Can anyone find such a thing I can refer to please?
> 
> Isn’t it bad to appeal to an intuition that breaks down very quickly?
> Especially when it is using a very common term in something other than the normal sense?

If you believe that the normal sense of "graph" is provided by the branch of finite mathematics called 'graph theory', then you are part of a very small minority of readers. Even technical dictionaries do not give this as the standard meaning. (Even in mathematics, 'graph' is also used to mean a diagrammatic representation of a function, and this is apparently considered to be a more 'normal' sense than the graph-theoretical one.)

> 
> There seems to be an agreement (not surprisingly) that "an RDF graph is a set”.
> So that sort of begs a very simple question:- if it isn’t a graph, and it’s a set, why isn’t it called an "RDF Set”?

Because the usage "RDF graph" was already widespread and in common use when the documents were written, and there seemed to be no good reason to change it. 

Pat

> 
>> On 27 Oct 2014, at 19:10, Melvin Carvalho <melvincarvalho@gmail.com> wrote:
>> 
>> 
>> On 27 October 2014 19:06, Pat Hayes <phayes@ihmc.us> wrote:
>> The use of the word "graph" in RDF is motivated by the illustrative diagrams rather than by the definitions of mathematical graph theory. In fact, mathematical graph theory gets in the way, since RDF graphs are not graphs in the sense of graph theory.
>> 
>> Formally, an RDF graph is a *set* of RDF triples, and an RDF triple is a 3-element *sequence* comprising a subject which is an IRI or a blank node, a property which is a IRI and an object which is an IRI, a blank node or a literal; for definitions of 'blank node' and 'literal', see http://www.w3.org/TR/rdf11-concepts/#section-rdf-graph
>> 
>> So, long story short: an RDF graph is a set.
>> 
>> On Oct 27, 2014, at 10:26 AM, Hugh Glaser <hugh@glasers.org> wrote:
>> 
>>> Everyone talks about RDF Graphs, and I have sort of puzzled over what an RDF Graph is - so I thought I would ask.
>>> 
>>> Sorry if you just need to point me at some W3C resource somewhere.
>> 
>> The formal definitions are given in full detail in the RDF concepts document, part of the normative standard. http://www.w3.org/TR/rdf11-concepts/#section-rdf-graph.  By and large, it is usually good policy to read the normative standards documents to find the formal definitions of concepts, for any standard.
>> 
>>> 
>>> "This linking structure forms a directed, labeled graph, where the edges represent the named link between two resources, represented by the graph nodes. This graph view is the easiest possible mental model for RDF and is often used in easy-to-understand visual explanations.”
>>> (http://www.w3.org/RDF/ )
>>> (I strongly agree with the second sentence, by the way!)
>>> 
>>> Simple Graphs are usually G = (V, E) comprising a set V of vertices together with a set E of edges, but that doesn’t seem to describe RDF Graphs for me.
>> 
>> Indeed, it does not.
>> 
>>> The sort of thing that I am considering is an RDF Graph such as:
>>> 
>>> rdfs:label rdfs:label “Label” .
>>> 
>>> Is it G = ({rdfs:label, “Label”}, {(rdfs:label, “Label”)} with edge-labelling function (rdfs:label, “Label”) => rdfs:label ?
>> 
>> No. It is in fact
>> 
>> { < www.w3.org/TR/rdf-schema/label, www.w3.org/TR/rdf-schema/label, < "Label", http://www.w3.org/2001/XMLSchema#string > > }
> Specific on this bit, please Pat:
> So I think you are saying that the graph is a set of triples, in this case one triple.
> And that you are separately saying that my graph concrete syntax should not be allowed to use “simple literals", as the RDF was?
> Also, is there any reason that you moved to (partial) URIs for the Graph syntax?
> 
>> 
>>> So we need to have both a vertex and an edge label with value rdfs:label, and they don’t really have a logical connection.
>>> Sort of worrying?
>>> 
>>> Is that the sort of graph an RDF Graph is, and is that how it is formally defined?
>> 
>> No, see above.
>> 
>>> 
>>> Also, a "labeled graph” usually refers to the vertices being labelled; should it not say that RDF is a “directed, edge-labelled graph”?
>>> 
>>> Not exactly my forte this, so I am hoping I will be able to understand any answers!
>> 
>> Hope this helps.
> It does, thanks.
> 
> Best
> Hugh
> 
>> 
>> Pat is right (as usual).  You may want to add the term "directed" graph, as I vaguely remember in graph theory edges are bidirectional by default.
>> 
>> 
>> Pat
>> 
>>> Best
>>> Hugh
>>> --
>>> Hugh Glaser
>>>  20 Portchester Rise
>>>  Eastleigh
>>>  SO50 4QS
>>> Mobile: +44 75 9533 4155, Home: +44 23 8061 5652
>>> 
>>> 
>>> 
>>> 
>> 
>> ------------------------------------------------------------
>> IHMC                                     (850)434 8903 home
>> 40 South Alcaniz St.            (850)202 4416   office
>> Pensacola                            (850)202 4440   fax
>> FL 32502                              (850)291 0667   mobile (preferred)
>> phayes@ihmc.us       http://www.ihmc.us/users/phayes
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>> 
> 
> -- 
> Hugh Glaser
>   20 Portchester Rise
>   Eastleigh
>   SO50 4QS
> Mobile: +44 75 9533 4155, Home: +44 23 8061 5652

------------------------------------------------------------
IHMC                                     (850)434 8903 home
40 South Alcaniz St.            (850)202 4416   office
Pensacola                            (850)202 4440   fax
FL 32502                              (850)291 0667   mobile (preferred)
phayes@ihmc.us       http://www.ihmc.us/users/phayes

Received on Tuesday, 28 October 2014 10:57:37 UTC