Re: lack of clear motivation for transforming from CSV to JSON or XML

Hi Anastasia, the point I was suggesting (I understand this thread is about
requirements, not solutions! :-) is that if we think about a general form
for CSV to JSON we should expect to have only one json, which is not true
(and yet is definetly not true for CSV or other tabular representation...
:-)
Adopting RDF is good in this direction not only for enriching with
metadata, but at a more simple level for describing the model in a simple
and yet explicit way: this model could be then easily transformed and
enriched in order to obtain a more "proper" RDF, as tarql or similar tools
does, for example. In this particular case, talking about json, I suggest
using json-ld. If we talk about going CSV->XML I'll probably suggest using
RDF/XML, and so on, just that.
Sorry for naive comments maybe, but following the discussion I'd like to
contribute :-)

A.


2014-03-12 18:59 GMT+01:00 Anastasia Dimou <anastasia.dimou@ugent.be>:

>
> I would further argue that it is impossible to define a mapping for every
> format. Why would we stop to "CSV to JSON" and "CSV to XML"?
>
> On 03/12/2014 05:15 PM, Alfredo Serafini wrote:
>
> Hi what about adopting json-d for this?
> I mean: json-ld can be seen as a good compromise because it offers a
> "natural" RDF conversion, while it is actually  a specific JSON dialect. As
> it is almost impossible to provide mapping for every kind of json dialect
> (and CSV formats too!), the usage of a specific json syntax as a reference
> might simplify things, and json-ld already embeds RDF logic.
>
>
>  Alfredo
>
> There is a clear requirement to transform from CSV to RDF<http://w3c.github.io/csvw/use-cases-and-requirements/#R-CsvToRdfTransformation>– which implies that is should be possible to convert CSV to one or more of
> the RDF encodings (incl. TTL, RDF/XML and JSON-LD) …
>
>
>
> But there’s a risk that this RDF-centric approach misses a concern simply
> about, say, converting CSV to simple JSON.
>
> The difference with RDF (and any of its encodings) is that we don't purely
> "convert" to another format just because there are different reasons that
> one might prefer to use the data in this format. It is that we enrich the
> data; we "map" them to their metadata providing additionally their meaning.
> In order to do that we need another structure (thus different
> format/serialisation). Those annotation are given because we expect that
> the manipulation of the data (not only in RDF but in other formats too)
> would be facilitated, as we know then the context of the data.
> Why "CSV to RDF to JSON-LD" or "CSV-LD (to RDF) to JSON-LD" (and in
> general CSV to RDF to any_format), would not be sufficient if it is needed
> to do so?
>
> Kind regards,
> Anastasia
> --
>
> Anastasia Dimou
> @natadimou | mmlab.be | iminds.be
> Semantic Web - Linked Open Data Researcher
> Ghent University, Belgium - Multimedia Lab - iMinds
> Gaston Crommenlaan 8 bus 201, B-9050 Ledeberg-Ghent, Belgium
>