categorising or relationships

Thanks to Johnnie Moore and his links to Stuart Henshall referring to a keynote by Dave Snowdon I was yet again reminded about the importance of ditching the tags.

Yep, the tags, a popular categorising tool still spreading, as an alternative to the age old taxonomy tree structures which I find no better. (But as I supported until doubt set in!)

My favourite take away was Dave's request to the audience:

"What's the odd one out from chicken, cow and grass?"

What's your answer? Pretty standard IQ test stuff such...

"Here in the West most of us would say grass but in much of the world they'd say chicken. That's because we're trained to filter by categories; elsewhere they filter for relationships."

says Dave, and I admit to the same fault. So much for western dominated IQ tests... heh.

Thing about categorising is that it creates relationships between objects in a indirect way, and thus leaves precision out.

Chicken and Cow are not directly related other than both being members of same groups: "Domesticated animals" or both "farm dwellers" or both being "staple food in many countries", when dead and parted of course.

Hardly precise those relationships, and thus not very valuable as knowledge enhancers.

Adding knowledge to the objects is what it's all about.

If precise and fulfilling you and a system can find the right object at the right time - enhancing productivity, learning, speed, precision and minimising errors and waste of time.

Whatever we're doing to software systems or ways of running anything - using the best possible knowledge enhancers is of utmost importance. Without the best the rest is kind of moot.

Back to the cow.

Cows eats grass. That's useful and pretty simple or what?

Cows lives on farms.

Chicken lives on farms.

All relationships. Semantic N-triples - readable by you and me, and a proper system as well.

Allowing queries like "what eats grass?", "what animals lives in same locations as cows?" or even "what is the most popular car brand among owners of locations with grass eating animals and egg laying animals?".

Nifty eh? And not easy to do with categorising...

Comments

[phil jones]

It's the old story.

No one doubts it would be *useful* to have the world conveniently modelled in a bunch of relationships.

The issue is going to be *who* is going to do the analysing of the world into just the right model that it will turn out to be useful to you in future.

What's convenient about tags is that, by piggy-backing on a bunch of common cultural norms, (yes, we in the west think cows and chickens can both be tagged "farm" and "animals") it means that a bunch of people can help each other out without having to work too much at it. The communication is perhaps a little rough, but it's useful enough to do some work with.

They don't need to sit down in a committee and work out what kind of relationship a cow has with a farm. Is "cow" a "component-of" a "farm"? Part of the "inventory" of a farm? Part part of the "productive machinery" of a farm? Maybe merely "resident-in" a farm? Etc.

[sig]

Phil,

did some more "highly scientific" research during the week, asked the question to a whole bunch of people at a UK educators conference and found that it's really not a west vs. east thing: It's a gender thing, or rather "structured male" (who used the category approach) vs. the rest including all women (who went the relationship way).

So much for a common cultural understanding among a homogeneous sample of UK educators :)

I have two problems with categories, first it requires training, I have to know the system otherwise I'll never "find" things. Who knew that a marmotte (marmotta marmotta) was of the squirrel family?
Secondly it does not really add direct knowledge - as Plato defined it (agree with him or not, I like that one): Knowledge is how objects relate to other objects. Categorising is indirect cow and chicken both belong to same group, then you need to have more on how that group was defined and how that affects it's members.

One might say history went this way:

1. We amassed knowledge in our heads and transferred them by word of mouth - all as a graph, multidimensional and in accordance with Plato 2400 years ago.
2. We needed to capture and distribute "knowledge" by pen and paper - a two dimensional method that did not allow a "graph". Thus Carl von Linnae won over Comte de Buffon and tree-structured taxonomies came to life, categorising became the norm in certain cultures (and parts thereof).
3. Now with IT we have the opportunity to capture and distribute multidimensional graphs of knowledge and our brains are beginning to be rewired back again (see my experiment this week ;)). So why not go all the way and use the better method that deliver higher precision, more depth and requires less up front training?

Ah, a theme for much discussion - we have a Skype session yet to happen!