Re: [lod2] World Bank Linked Data

On 12-03-19 12:35 AM, Hugh Glaser wrote:
>
> On 19 Mar 2012, at 00:05, Sarven Capadisli wrote:
>
>> On 12-03-16 07:35 PM, Hugh Glaser wrote:
>>>
>>> On 16 Mar 2012, at 16:22, Sarven Capadisli wrote:
>>>
>>>> On 12-03-16 02:19 PM, Hugh Glaser wrote:
>>>>> Hi Sarven,
>>>>> That's great.
>>>>> I have added your owl:sameAs links to sameAs.org
>>>>> (I read your Terms of Use, which seemed to a allow it, but if you want I can remove them.)
>>>>> (http://sameas.org/?uri=http://worldbank.270a.info/classification/country/AD)
>>>>> Really gives a sense of the linkage in the LD cloud on countries.
>>>>
>>>> What can I say other than, thank you :)
>>> Great :-)
>>>>
>>>> If I were to stick to a super restricted view on owl:sameAs and the fact that the country resources are just concepts (as I claim them to be skos:Concepts), it should use skos:exactMatch to DBpedia's country resource. To keep the LOD crowd happy and slightly loosen the view on the relationship between the two resources, I went ahead with owl:sameAs. Either way has its pluses and minuses and that burden is unfortunately dumped on the consumer. If there is an intelligent way to go at it while keeping a great number of people happy, I'd be glad to update things accordingly from this end.
>>> As far as the exact predicate, I avoid venturing into what is best.
>>> I would have done the same harvest if you had put skos:exactMatch or even skos:closeMatch.
>>> sameAs.org is a discovery mechanism - it helps people/agents find things (I hope), and then they can choose whether they believe it!
>>>
>>> Hmmm.
>>> Now I notice you *do* have some skos:exactMatch and even three skis:closeMatch.
>>> So they have gone in as well :-D
>>> (I missed them earlier as they were not in your nice voiD file, which is the place I looked, of course :-) )
>>
>> I've just updated the VoID file. However, I didn't put in skos:exactMatch and skos:closeMatch this time around in void:Linkset because there are only a few of them in the whole dataset. I will however revisit this when there is more stuff going on.
>>
>>> By the way, you have the 2 letter skos:exactMatch with the 3 letter (e.g. AW = ABW).
>>> But the 3 letter (e.g. http://worldbank.270a.info/classification/country/ABW) resolution has very little information.
>>> Is this intentional?
>>
>> Yes, that is intentional. I don't have a perfect solution to my intentions at this time. Perhaps you can chime in on that.
>>
>> What I was thinking is that, the URI for the three letter code is just there to allow consumers which are aware of the three letter code (as opposed to the two letter code) to easily jump to the two letter URI. I suppose that could have been also achieved with a literal for the two literal resource as well. In the name of automatic discovery, of course more triples can be introduced for the three letter resource (e.g., rdf:type), but I think that unnecessarily bloats the data. The three letter resource is only there as a hook for the two letter resource. In the end, the consumer will have to comb through the data any way to find out about the three letter code. I'm not sure if there is any real benefit of one way over the other i.e., to use a skos:exactMatch to three letter URI vs. the three letter literal value for the two letter URI. I've created a URI for the three letter code because I /sort of/ wanted to give both of the codes equal importance even though only one of them cont
ains the data. Perhaps owl:sameAs would be more appropriate here.
>>
>> -Sarven
> I (still) won't comment on which predicates you use to describe equivalence :-)
> My observation was that it was hard (impossible unless you use sameAs.org :-) ?) to get from your 3-letter URIs to the 2-letter ones, which seemed strange.
> To be more specific:
> If someone gives me the URI http://worldbank.270a.info/classification/country/AND I can find nothing about it, not even http://worldbank.270a.info/classification/country/AD.
> In fact this actually contradicts principle 3 of Linked Data - you don't Provide useful information, as far as I can tell.
> Not a big deal - just a link (of some kind) to the 2-letter one would help.

You are absolutely right about this. This was the exact reason why I 
thought it may be sufficient to just provide the three letter code for 
the two letter code description - I think the main difference is that, 
iterating over a base URI is slightly more "hackable" than iterating 
over a triples.  As you say, if someone lands on the three letter URI, 
there is no connection to the two letter URI.

I've added:

country:ABW skos:exactMatch country:AB .

> In fact, I think, from what you say, you think the link is there from 3-letter to 2-letter.
> But I couldn't find it.
> I am looking at http://worldbank.270a.info/classification/country/AND.turtle

No, I'm saying the opposite. There is:

country:AD skos:exactMatch country:AND .

Just a side note on the formats (e.g., .turtle, .rdf) of a resource: 
they are currently generated versions of the SPARQL query i.e., it is 
what the framework is spitting out as opposed to what's really in the 
RDF store for a given resource i.e., it contains some extra triples to 
make the HTML output more human readable. Technically, it should not be 
there for non-HTML formats. I plan to fix this.

-Sarven

Received on Monday, 19 March 2012 17:26:10 UTC