Re: Question about doing RDF analysis in Ruby

I thought of a couple more questions while I was out and about:

1)  It's fairly common to design applications with an RDBMS (and now
document-based and key-value) data stores, and then serving that data in
various formats (html, xml, json, csv, etc.).  Obviously RDF could just be
another of these formats.  Or you could use RDF as your primary data model
(and data store).  Are there some guidelines or "rules of thumb" that are
out there to help decide which path to take?

2)  Everywhere I read about the semantic web and RDF, it talks about how you
can combine different data sets and derive new data (inferences) - has
anyone had experience with building systems that can answer questions that
users come up with at runtime (using SPARQL queries?), or is this mostly
done by writing programs?

Just looking for some general info as there's a lot of academic articles out
there, and a lot of books, and I want to know what I should be trying to
learn.

Thanks!

Greg


On Mon, Jan 31, 2011 at 3:26 PM, Greg Lappen <greg@lapcominc.com> wrote:

> Hi all,
>
> Thanks for the informative replies.  I have a couple followup questions:
>
> Is the ability to query multiple repositories something that is inherent in
> SPARQL, or in the Ruby implementation of it?  Also, is the Ruby SPARQL
> implementation that you're referencing going to layer on top of RDF.rb
> Repositories?
>
> We're starting to use RDF.rb to store our data natively in RDF, but we're
> just checking that we'll be able to use it in this format before we go too
> much further.  It seems like our choices at this point are to use SPARQL
> against our RDF.rb repositories, or use an external system to do the queries
> (Hadoop or SPARQL server like 4store, etc.).
>
> Thanks for all the help!
>
> Greg
>
>
> On Mon, Jan 31, 2011 at 2:57 PM, Gregg Kellogg <gregg@kellogg-assoc.com>wrote:
>
>> That said, you could use SPARQL to query multiple repositories.
>>
>> There is work proceeding on native SPARQL support in Ruby, stay tuned...
>>
>> Gregg Kellogg
>> Sent from my iPhone
>>
>> On Jan 31, 2011, at 2:13 PM, "Ben Lavender" <blavender@gmail.com> wrote:
>>
>> > Greg,
>> >
>> > I'm understanding this as you want a single RDF::Repository that
>> > represents multiple RDF::Repositories in a few different couchDB
>> > installations. In this case, SPARQL won't help you meet that API, as
>> > that API has functions that SPARQL does not (certainly not SPARQL 1.0
>> > anyway).  You'd need to just make an RDF::Repository that sends out
>> > each operation to the sub-repositories and merges the results. This
>> > has been on my personal roadmap and I just haven't made the time.
>> > Shouldn't be a big deal, and I'd push to include such a repository in
>> > the linkeddata gem if you wanted to release it.
>> >
>> > Ben
>> >
>> > On Mon, Jan 31, 2011 at 7:34 PM, Greg Lappen <greg@lapcominc.com>
>> wrote:
>> >> Hello,
>> >> We are using RDF.rb now for persisting RDF data into our CouchDB
>> >> RDF::Repository.  We believe that we will be splitting the data into
>> >> different repositories in order to shard and distribute the data
>> between
>> >> different locations, but we are definitely interested in doing analysis
>> and
>> >> inferences on the data spanning multiple repositories.  What is the
>> general
>> >> approach for doing this kind of thing with RDF.rb?  Is there a SPARQL
>> >> implementation for RDF.rb that would help with this?  Or would we have
>> to
>> >> write ruby code to walk the repos and do the queries programatically?
>> >> Thanks!
>> >> Greg
>> >
>>
>>
>