On 8/23/06, Neil Harris neil@tonal.clara.co.uk wrote:
The problem is different senses of 'subcategory': there's [1] "the set of objects in B are a strict subset of the set of objects in A", which is loop-proof, and [2] 'B is a topic that is usually discussed in the context of A'
For example:
London -[2]-> Thames Valley -[1]-> London, comes to mind
Russia -[2]-> Soviet Union -[1]-> Russia, as well.
Good explanation!
See [[WordNet]] for a serious attempt to tease out the relationships between concepts in detail: there are _lots_ of possible ways that one thing can be related to another.
One major complication not to forget is that our categories are at the *article* level, not at the *topic* or *concept* level. Often one article = one topic, but not always: for example, [[Beamish and Crawford]] is about both a brewery and a beer. [[Controlled Impact Demonstration]] covers the topics of an air safety program, a type of jet fuel, and a specific event.
If anyone has some schemas or diagrams or whatever that could attempt to make sense of this, I'd love to see them. Clearly a strict hierarchy as the only classificational structure is inappropriate. However, we can also do better than unstructured tagging. What's a good compromise? How can we get the benefits of a hierarchy where it works, and the flexibility of tagging when it doesn't?
Steve