Re: doi xorxes. do ponse xo tanxe

[Jorge Llambias <jorge@PHYAST.PITT.EDU>]

Mark A Biggar:
> Another way to state Zipf's Law is that in a lot of "natural" distributions
> the nth most frequent item appears with a probability proportional to 1/n.
> But I always thought that was obvious because most things are Normally
> distributed and if you fold a normal distirbution in half around then mean
> line you get something that looks alot like the 1/n curve.

Are you sure? I think they are significantly different. To start with, the
1/n has a much longer tail, which is probably the most interesting part of
Zipf's law (I imagine). If there was no difference with the normal
distribution it wouldn't be named after Zipf.

> Where the
> lingusitic version of Zipf's law comes from is that Zipf originally
> developed his distribution law by studing word frequencies in natural
> languages and observed that the most frequent words tended to be shorter.

And to think that he didn't have computers to count the words.

> Zipf (probability putting the cart before the horse) theorized that first
> observation was a natural law that was responsible for the second.

Did he really make some sort of connection between the word length and
the distribution, or just observed that most short words were very
frequent?

> Personaly, I always though laziness explained it better.

Yes, I'd say a combination of laziness and impatience.

> Both of Zifp's observations have become know collectivly as Zipf's Law.


Jorge