Tag cloud: linear dependence vs logarithmic

This post made me write two things. Autokadabra and someone under the nickname napisal , who actively disagreed with my post ...

It was like this: I entered the map of Avtokadabra clubs and realized that she did not speak to me about anything. The impression was created that they write only in three clubs. After studying the cloud, I came to the conclusion that the linear dependence of the tag size (in this case, the blog) on ​​the number of publications is completely not informative. Clubs with one publication look the same as with the 15th .

The first thing that came to mind was to replace the linear dependence of the logarithmic one , which I did. I liked the result so much that I decided to do the same with the tag cloud . Then wrote a post on Mmm ... tasty .

So the tag cloud has changed in a compact form (it is evident in the form of a cloud of clubs, but not so large):
')

Linear dependence ( before )	Logarithmic dependence ( after )

Actually, why did I decide to write about it all the same on Habré? Because, in the process of a dispute with napisal , I googled such an interesting article . It is curious in her that in the form of a cloud, it shows not tags, not blogs, but statistics on the use of first-level domains. What led me to deep philosophical reflections on the logarithms of statistics and space :-)

Here is a picture from the article:


UPD from Wikipedia: The normal distribution , also called the Gauss distribution , is the probability distribution, which plays a crucial role in many areas of knowledge, especially in physics. The physical quantity is subject to normal distribution when it is influenced by a huge number of random noise. It is clear that such a situation is extremely common, so we can say that of all the distributions in nature, the normal distribution is most often encountered - hence one of its names.

Actually, tags (as well as domain names), it turns out, obey the same distribution law. From this and logarithms ...