Re: Labelled graphs

On 16-04-2012 14:14, Sandro Hawke wrote:
> On Mon, 2012-04-16 at 13:07 +0100, Andy Seaborne wrote:
>> On 15/04/12 15:57, Sandro Hawke wrote:
>>> On Sat, 2012-04-14 at 20:35 +0100, Andy Seaborne wrote:
>>>> On 13/04/12 11:33, Steve Harris wrote:
>>>>>> Read that as rdf:aaa to avoid natural language meaning.
>>>>> I don't get that, but…
>>>> Natural language can be confusing.
>>>> <u>   rdf:label G .
>>>> so is<u>   labelled by G ?
>>>>       "<u>   has a label of G"
>>>>       "G is a label for<u>"
>>>> or is<u>   labelling G ?
>>>>       "<u>   is a label for G"
>>>>       "G has a label of<u>"
>>>> (sometimes it does not matter -- e.g. sameValue).
>>>> "labelled graphs" is more of the style"<u>   is a label of G".
>>>> All that is needed as the most basic building block is a neutral
>>>> association of (<u>, G) and that within a dataset (<ui>, Gi) is
>>>> functional from<ui>   to Gi : given a label, find the associated graph.
>>> With these looser semantics, I don't think it needs to be functional.
>>> Rather, I think we can define it as labeling the associated graph and
>>> each of its subgraphs.
>>> There would be some kind of deep equivalence between these two datasets:
>>> D1:
>>>      <u>   {<a>   <b>   1 }
>>>      <u>   {<a>   <b>   2 }
>>> and
>>> D2:
>>>      <u>   {<a>   <b>   1,2 }
>>> The SPARQL semantics might written in terms of a normal form, where a
>>> term is only used as a label once, but the other forms would be valid in
>>> trig and in the RDF Semantics.
>> In earlier threads, the proposal was that a TriG document had one graph
>> per label but (syntax) it may be split into separate sections for
>> practical reasons.
> Yes, it was.  I'm suggesting it makes sense to put that in the semantics
> of datasets, instead, so we can use this for merging datasets.  It makes
> merging datasets the same as concatenating trig files (give or take
> @prefix declarations and blank-node labels).

An attempt at formulating a possible conclusion/consensus from this thread:

* Non-typed labels are simply associations, no special semantics

* Typed labels: we do not seem to come up with more than these two:

   1. The label is typed as a rdf:StaticGraphContainer:
       the graph labeled is a g-snap of the label URL
   2. The label is types as a rdf:Graph:
       the label denotes the graph it labels

* When the same label is used multiple times in the same dataset, the 
graph is
    assumed to be the union of the graphs labeled with it

The appears to be in line to the 6.1 design, with some 

>      -- Sandro

Received on Tuesday, 17 April 2012 15:59:51 UTC