Joho the Blog
|
|
|
May 18, 2006
Sean Coon muses about how Silly String completes trees. Well, actually, he muses about the shape of language, wrapped in an homage to his felicitously-named mentor, Bill Readings. I like a lot what Sean says and the people he quotes from. And it makes clear just how un-tree-like is the structure of language. Ferdinad de Saussure, whom Sean quotes, talks about words not as leaves on forking conceptual branches — a picture Aristotle might have liked and that WordNet assumes — but words as standing in distinction from other words. Saussure's view does not resolve into anything like a tree. At least as far as I remember. Likewise, when the British philosopher John Austin says that the word "real" usually doesn't signify some positive quality but merely flags a distinction in mode — we only talk about a real gun if we need to distinguish it from a toy gun, a fake gun carved from soap, or a pretend gun made by pointing a finger — the meaning of "real" does not consist of its position in a tree. Language, it seems to me, generally lacks the basic properties that make a tree a tree, with the one important exception that sometimes concepts contain other concepts. But there's lots more to trees than that. E.g., a tree structure has a top and bottom. The elements are discrete. Each element hangs from one branch. All the branches signify the same basic relationship. The branches inherit essential characteristics from the branches they're attached to. Branches have essential characteristics. Meanings can be traced and paths can be followed. The organization is neat, not messy. And even the basic notion of containment is a metaphor and way too general: Does "color" contain "red" the way "nation" contains "city" and the way "actor" contains "David Caruso" ? And, by the way, "yard" does not contain "dog" even if your dog is in your yard and "stomach" does not contain "peanut" even if you've just eaten one. Sean's post doesn't get stuck in the tree metaphor. On the contrary. He uses Silly String to remind us that the tree of language has messy connections among its leaves. He points out that language isn't a single tree, the same for all. He refers usefully to Saussure and Barthes. So why stick with the tree metaphor at all? It's gotten in the way of understanding for about 2,000 years now. (Porphyry is usually credited with being the first to draw categories in the shape of a tree.) Except in the limited domains where we carefully structure language into a tree, I think we ought to drop it. I tried to get at this, or at least hint at it, in my reply to Julian Bond's comment on a post of mine a few days ago. Or, as a certain book puts it, everything is miscellaneous...although that phrase by itself is misleading unless we immediately ask: Then why didn't it stay that way? Posted
by D. Weinberger at May 18, 2006 09:21 AM
|
Comments
Does "color" contain "red" the way "nation" contains "city" and the way "actor" contains "David Caruso"?
No. But Semantic Web people have been working on this stuff for years, and AI people were working on it for a couple of decades before that - a paper that's still cited (called "What IS-A is and isn't") was published in 1983. There's even a philosophical discipline devoted to part/whole constructs - 'mereology'. I've only scratched the surface of this stuff myself, but it's given me several lightbulb moments.
DAGs aren't (necessarily) mathematics, by the way. A DAG consists of bubbles connected up by lines and arrows; the arrow that goes from bubble A to bubble B says "A is a B", and the arrow that goes from A to C says "A is a C". That's why it's directed; it's acyclic because it doesn't loop round (you never end up saying 'A is a B which is a D which is a K which is an A'). What a DAG isn't is monolithic (or mono-hierarchical): there's no problem defining heroin as a pharmaceutical, an illegal narcotic and a herbal extract, say. The underlying logic is set-based, and consequently supports a single item belonging to multiple sets (all blue things, all round things, all Chinese things...)
Posted by: Phil | May 18, 2006 12:28 PM
hey phil, so bubbles... possibly similar to this subset?
Posted by: sean coon | May 18, 2006 12:47 PM
And philosophers have been working on it for even longer than that, of course. You could trace it back to Aristotle's essentialism (which is a wrong answer to the right question).
Ultimately, none of these representations are as rich as language. That's not an objection to any of them, unless they're being put forward as the be-all and end-all of meaning.
Posted by: David Weinberger![[TypeKey Profile Page]](http://www.hyperorg.com/blogger/nav-commenters.gif)
|
May 18, 2006 01:10 PM
agreed, david.
maybe we can raise the bar of the conversation by simply dropping the overused analogy of a tree of knowledge to another physical metaphor... like an ocean.
it's not a be-all and end-all of meaning, but it's much more dynamic than a tree.
and for phil, there are plenty of bubbles ;)
Posted by: sean coon | May 18, 2006 04:55 PM
Isn't a tree just a point of view on relationships?
It's like: if I start from this point here, set some boundaries, ignore anything outside those bounds, and look in one direction, I see a tree.
A typical recording of a "family tree" is a good example: I might see my family tree as starting from the point of my paternal great grandparents, and then ignore various in-law branches, and, voila, I see a tree. It's not a real tree unless I look at it that way. And, I can see it as not-really-a-tree, too.
I think a good question is: can people see / use X as a tree, and if so, do they see / use the tree to gain insight into, or ignore, the unbounded and multidimensional structures of X.
Posted by: Jay Fienberg | May 18, 2006 05:59 PM
I think a good question is: can people see / use X as a tree, and if so, do they see / use the tree to gain insight into, or ignore, the unbounded and multidimensional structures of X.
I think this works in the "opposite" way too, e.g.:
Can people see / use X as a set of tags, and if so, do they see / use the tags to gain insight into, or ignore, the tree structure of X.
;-)
Posted by: Jay Fienberg | May 19, 2006 02:45 AM