Re: Google API -> Jena

On Fri, 19 Apr 2002, R.V.Guha wrote:

>  
>
> <shameless_promotion_of_own_project>
>
>    You might want to take a look at  this url for an example
> of combining web services from different places 
> with google search results based on the semantics of the query.
> All the code and the kb is available for free ...
>
> </shameless_promotion_of_own_project>

<shameless_promotion_of_guhas_project>

http://www.alpiri.com/ss/ is pretty cool, especially the knowledge base
(or 'database' for the KR-averse).

I do think we often worry too much about technologies to support sweeping
generalisations (reasoning about class hierarchies, rule systems etc)
while not paying enough attention to the individual resources those
hierarchies are describing. So the TAP/Alpiri KB is nice as it is pretty
focussed on specific things, rather than on capturing a worldview that
generalises about those things.

...and this is where network effects and the (blah blah semantic) Web come in.

knowing that XML file A and XML file B are describing the selfsame thingy
is extraordinarily useful. We can answer questions that require
information from both A and B. Seems so simple, but makes such a difference.

'Guha's big list of useful things to describe' (aka the KB) makes it easier
for us to merge scattered descriptions of them. Even if we don't take much
notice of the worldview implicit in his ontology.

 </shameless_promotion_of_guhas_project>

<shameless_trailer_for_own_project>

...this turns out to be useful for something I'm working on with some
friends.

We're writing RDF to describe digital images, especially who is in which
picture, along with wordnet classifications and SVG path outlines. Most
people in the photos we've catalogued can be picked out usign a mailbox or
homepage property. But then we wanted to describe a photo of JFK, of Bill
Clinton, and of Frank Sinatra. We didn't know their email addresses;
bummer. Fortunately enough, Guha and co have (in the
big-list-of-useful-things-to-describe) assigned Alpiri/TAP identifiers to
those 3 people. So we can start describing them in a somewhat more
realistic manner.

More on all this eventaully. If you rummage around http://rdfweb.org/
you'll see some of what we've been up to. The photo stuff is at
http://rdfweb.org/2002/01/photo/ and nearby, with Libby's Java
implementation at
http://swordfish.rdfweb.org/discovery/2001/08/codepict/allpeople.jsp
...along with links to a paths-thru-the-image-data tool that Damian and
Libby implemented recently.

An example query, finding paths thru images connecting the person with
mailbox mailto:guha@guha.com to the person with mailbox mailto:president-35@whitehouse.gov

see:
http://swordfish.rdfweb.org/discovery/2002/02/paths/allabout.jsp?mbox=mailto%3Aguha%40guha.com&mbox2=mailto%3Apresident-35%40whitehouse.gov

This is an example of the value of (a) common data model (b) common
mechanisms for identification that go beyond the fiction that 'everything
has a URI'.

more demos, writeups etc after WWW2002...

 </shameless_trailer_for_own_project>

dan

>
> guha
>

Received on Friday, 19 April 2002 11:49:58 UTC