Making a Drupal Folksonomy Tag Cloud

I created a tag cloud for my website, and I'd like to see this feature added as a dynamic Drupal module. I thought I'd briefly go through the steps that I went through to give a leg up for anyone who wants to code this up in PHP.

The hardest part is the algorithm that automatically determines the distribution of font sizes based upon the frequency distribution of tags.

Below is my distribution that I used to determine the font sizes:

Notice that my tag distribution exhibits some Power Law behavior of the Long-Tail of the Internet

More technical details below...

The first step to creating a tag cloud in Drupal is to use Morbus Iff's free-tagging patch to go through and tag all of your archived blog posts. This saves your entered folksonomy free tags as regular taxonomy terms.

UPDATE: Morbus Iff writes, "My freetagging is currently in Drupal HEAD (along with AJAX support, which makes it amazingly useful) - and I backported it to Drupal 4.6 (but not 4.6.1; that's what you've linked to in your HOWTO)."

I then went into PHPMyAdmin to hack into two MySQL databases within Drupal: "term_data" & "term_node"

I exported all of the data into a CSV file so that I could import it into an XL spreadsheet.

The data in "term_data" are used to correlate the folksonomy tag "name" with the "tid"

The "vid" variable also in "term_data" is the vocabulary id that can be used to isolate groups of terms into separate tag clouds. In my case, my "Folksonomy Tags" vocabulary "vid" was 4.

The data in "term_node" are used to count the total number of occurences of a folksonomy tag "tid" across the entire site. I order the data according to tid, and created a counter in XL using four total columns

Columns
A = "nid" = data from term_node
B = "tid" = data from term_node
C = "0" = I copied "=IF(B2=B3,1,0)" into the entire column to give a flag of 0 at the last value of each tid
D = Count = I copied "=IF(C1=0,1,D1+1)" to count the total number of occurences of each tid

I imagine these counts could easily be coded up in PHP.

I copied the values of 0 (Column D), tid (Column B) & Count (Column D) into a separate column and sorted in the order of 0, tid, & Count

This gives you the total number of occurences of each folksonomy tag "tid."

This frequency will determine the relative font sizes of the tag "name" in the tag cloud

The next step is to correlated the tid number with the tag "name" using the data from term_data.

The hardest step for dynamically automating this into a Drupal module is determining how to automate the font size distribution based upon the frequency of tags.

I just plotted my tag distribution in XL and eyed it

I qualitatively determined that I could break up the distribution by dividing the tag frequency total by 10 -- round down -- and then add 1.

For example, the New Media tag occurs 53 times.
FONT SIZE = round(53/10) + 1
FONT SIZE = 6

Collaboration = 41 times = font size 5
Transparency = 30 times = font size 4
Open Source = 23 times = font size 3
Folksonomy = 15 times = font size 2
del.icio.us = 6 times = font size 1

This algorithm doesn't work in all cases, but it works for now.

UPDATE: I figured out a more universal way to break up the distribution. I'll post more about it after I finish the tweaking and graphing it.

The Drupal font size range seems to be from 1 to 7 which gives 7 possible sections.

I'll have to see if any solution pops to mind, but I think I'll pass the baton to someone with a computer science background to figure it out and code it up in PHP for the whole Drupal Community.

UPDATE 6-20-05: I've come up with an algorithm for a more even distribution of font sizes. More tomorrow.

This type of dynamic tag cloud aggregator could provide a very helpful organizational and navigational tool for Drupal sites.

UPDATE: Morbus Iff pointed me towards tagadelic, and I included an update in the previous post, but not this one.

It also appears as though Bèr Kessels, who is listed as the developer of the tagadelic module, just left an anonymous comment down below pointing me towards tagadelic, and I will definitely follow-up with him to see if we can't update the algorithm to show a dynamic range of font sizes correlated to the frequency -- and get a more even distribution by taking the logarithm of the frequency since it seems as though most folksonomy distributions exhibit Power Law behavior.