http://www.osnews.com/story/19993/Scalable_Highly_Expressive_Reasoner Exploring the Future of Computing en-us Copyright 2001-2010, David Adams adam+nospam@osnews.com Fri, 19 Mar 2010 19:06:03 GMT http://www.osnews.com/images/osnews.gif http://www.osnews.com http://osnews.com/thread?321985 http://osnews.com/thread?321985 This article makes me feel stupid Tue, 08 Jul 2008 18:40:00 GMT donotreply@osnews.com (righard) Comments http://osnews.com/thread?321996 http://osnews.com/thread?321996 Yeah, I'm with you there. I thought about posting it in the humor category - just to test whether anyone who does understand it has a sense of one. Tue, 08 Jul 2008 19:43:00 GMT donotreply@osnews.com (LostAirman) Comments http://osnews.com/thread?322097 http://osnews.com/thread?322097 Don't feel stupid . The problem is just that the paper is using a lot of technical jargon without explaining it properly. For a much clearer overview of what is going on, see:
http://en.wikipedia.org/wiki/Web_Ontology_Language
or
http://www.w3.org/TR/owl-guide/

Basically, it's an automated theasaurus on steroids. You build a database of things(classifiers) that you know about objects, then the computer can use it to infer things about other objects. So, for example. If you tell it that:
Chardonnay is an alcoholic drink
Chardonnay is made from grapes
Wine is an alcoholic drink
Wine is made from grapes

Then, if you tell your semantic-web engine to search for wines, when it comes across a Chardonnay listed, it will be able to infer that it is a wine without you actually telling it that. Of course that is a very basic example. Another good example is searching a drug database to infer wether 2 drugs might have side-effects when taken together, based on incomplete data.

As far as I can tell, what is special about the IBM approach, is that it uses 2 algorithms to resolve relationships.
1. A fast algorithm quickly tries to eliminate whole sections of the classifier database by asking special 'aggregations' about the query. so, if the results of asking the Ciders class 'Is a Wine?' and 'Is not a Wine?' are identical, then all Ciders can be eliminated from the query.
2. Then a full, traditional algorithm is used on the subset of classes that cannot be eliminated. Wed, 09 Jul 2008 10:05:00 GMT donotreply@osnews.com (stestagg) Comments