this (probably related) experiment would be of your interest. The
outline is as follows:

extract (keywords) RDF via GRDDL (1)
--> map key'word' to an object (URI) (2)
   --> look up corresponding Wikipedia data (3)
    --> relate object (2) to PSId if exact match found (4)
    --> or to disambiguation obj. of disambiguation page found

Step 1 uses a GRDDL profile [1] and associated XSLT [2]. Step 2 uses a
vocabulary to map 'word' to an object [3]. Step 3 uses an experimental
mapping web service [4]. Resulting PSId uses an Wikipedia page as PSI.

An example to relate an ambiguous keyword 'opera' to the most likely
concept and other candidate objects will be found at [5]. Note URIs
for (2) and (4) are hash type, not 303 redirect, so that the base URIs
for those object can behave as sort of PSIs.



@prefix : <> . <> :from [:name
"KANZAKI Masahide"; :nick "masaka"; :email ""].

Received on Monday, 21 January 2008 08:02:32 UTC