Re: freebase parallax: user interface for browsing graphs of data

Tim Berners-Lee wrote:
>
>
> On 2008-08 -19, at 21:27, David Huynh wrote:
>
>>>
>> The trouble is of course when the whole web is the database, it's 
>> hard to suggest those relationships (connections) for a set of entities.
>> How might one solve that problem? I suppose something like Swoogle 
>> can help. Is that what Tabulator uses to know what data is on the SW?
>
> Swoogle? Centralized index? No, not at all. The Tabulator is a browser 
> for linked data.
> The convention for linked data is that if I do a GET on the URI 
> identifying something, the returned information will include the 
> incoming and outgoing links that a reasonable person might be 
> interested in following.  Linked data clients dereference any URIs 
> they have not come across before, and pick up either data directly or 
> clues as to where to find what more data.  So at each point, you know 
> what the options are leading on.
>
> When you pick up a set of related things, of course, there will be 
> some presidents who have children and some who don't.  And there will 
> be some presidents which a have bunch of obscure properties. 
> Especially once you have overlaid data from many sources.  So then 
> there may be a case for having lenses linked from ontologies to allow 
> one for example to focus on geneology or political career.
>
> It gets more complicated when you try to automatically build an 
> interface for allowing people to input data about a president, chose 
> which fields to offer.
>
> I'd like to see the Freebase data as linked data on the web ... then 
> we could try all our other UIs on the same data!  What would it take 
> to put in some shim interface to Freebase?
>
> Tim
>
> PS:  There are places where a centralized index will help though, like 
> finding people's FOAF page from their email address.
>
>
>
>
>
>
>
David,

Which brings us back to the initial point: expose the URIs (you've done 
this partially) to Linked Data aware agents :-)  This, as I've already 
indicated, comes down to:

1. Replace all "javascript:{}" with their actual Freebase URIs (which 
already exist)
2. Use an information resource to collate all the URIs that are in scope 
at a given point in time in you UI, and then expose the resource URI via 
<link rel="dc:source" ../> (you can use RSS, Atom, OPML for this, the 
key thing is to list the URIs)

By doing the above, your Visualization oriented information resource and 
the data sources behind remain Linked Data Web accessible.

Anyway, you've already met me half way (re. yesterday's comments), so 
I'll crack on with my bit: showing how your interface can provide a 
beachead for launching a SPARQL Full Text query (without the user 
writing any SPARQL or seeing any RDF in plain sight).


Tim et al: 
There is a "first cut" of the last but one Freebase dump in Linked Data 
form (to the degree attainable) at:
http://linkeddata.openlinksw.com:8891/sparql (the graph URI for SPARQL 
FROM is : <http://linkeddata.openlinksw.com/freebase#> )

We are working with Francois Belleau (of http://bio2rdf.org) to produce 
an updated release, but even that release is going to require some work 
in order for the graph to be totally coherent (*a long story*).

Also note, we are RDFizing Freebase on the fly and producing slightly 
better Linked Data, so if I go to 
<http://mqlx.com/~david/parallax/browse.html?type=%2Fgovernment%2Fus_president> 
and simply do a "View | Linked Data Sources" via the ODE XPI's context 
menu against a President's Freebase URI I get: <http://tinyurl.com/6pxetc> .


To conclude, we have Linked Data in two forms for the Freebase Data Space:

1. Their Quad Dumps transformed into RDF and then loaded into Virtuoso 
resulting in a SPARQL endpoint and a Linked Data Space
2. Linked Data generated "on the Fly" via our RDFization cartridges with 
proxy URIs used as mechanism for endowing the resulting Linked Data 
Space with dereferencable URIs

Tim: Tabulator should work with either.

-- 


Regards,

Kingsley Idehen	      Weblog: http://www.openlinksw.com/blog/~kidehen
President & CEO 
OpenLink Software     Web: http://www.openlinksw.com

Received on Wednesday, 20 August 2008 13:36:18 UTC