- From: Rob McCool <robm@robm.com>
- Date: Wed, 23 Jun 2004 16:30:21 -0700
- To: Dan Brickley <danbri@w3.org>
- Cc: public-rdf-tap@w3.org
> Do you think the terms (in TAP, and other larger vocabs) need to be > partitioned? Or can they, with some care, be neutral and be the thing > used to bind together different views/contexts? > > ie. should we be looking for a more neutral, persective-binding set of > terms than 'Terrorist' (eg. Militant, ... whatever), or stick with the > contentious terminology and shuffle things around into value-packs. SO I > might buy into different values much as I use different software > modules. > > Tricky stuff... maybe we could work through an example? I think that it can be contentious any way. For instance, is a Terrorist a kind of Politician? Is a religious organization a political organization? What exactly is the Church of Scientology? I'm not sure there are any purely neutral terms. That being said, I think that there are enough problems to deal with even if you hold the terms (classes and properties) constant. Suppose that you can agree on some set of classes and properties that define the world. Suppose you have Militant, which is defined as a social role or something, and you have MilitantOrganization. More contentious terms could be defined, such as Terrorist and TerrorOrganization, and let's assume they can be put into the class hierarchy without much controversy. The thing that contexts get you here is that "the KB" is no longer a set of triples, it is a set of quads. Each triple has a context it was found in. So in the case of Hamas, it's probable that everybody could agree that it's a MilitantOrganization. But some will classify it as a TerrorOrganization, and others will classify it as a HumanitarianOrganization. The contexts for each would be the site that believes them to be such organizations. Contexts mean that instead of operating in the mode that TAP or a typical search engine does, where everybody sees the same data, you could define a view as being which set of contexts you want to see. So when I look at the data in the KB, I get the data which is from the contexts I want to see, and not from the ones I don't. When I make a query, the information from the contexts I want is aggregated and presented to me as a (hopefully) coherent whole. If I was doing research, I might want to look outside of my point of view, and see how things change. For Hamas, if Bob makes a query for it, he'll see data from sites that consider it a TerrorOrganization. When Alice makes a query for it, she might see data from sites that consider it a HumanitarianOrganization. Since MilitantOrganization can be agreed upon, they might both see data which doesn't take one of the controversial stands. As an interesting aside, here's how Wikipedia treats Hamas: http://en.wikipedia.org/wiki/Hamas > Hamas, acronym of Harakat al-Muqawamah al-Islamiyyah (Arabic: Islamic > Resistance Movement) is a Palestinian Islamist paramilitary and political > organization. > ... > Hamas is considered a terrorist organization by Israel, the United States > and more recently the European Union. A contexts based system is very complex and it will be difficult to manage (for instance, how do I tell the system what my proclivities are? do i copy them from my friends? some sort of web of trust? sets of defaults?) but it seems to be necessary to manage this sort of complexity of data.
Received on Wednesday, 23 June 2004 19:30:26 UTC