After seeing this, the good moderator @Alenanno requested me to make a similar map for Linguistics.

I present to you:

enter link description here

Tag size is proportional to tag popularity, and the size of links between tags is proportional to the amount of times they are used together. The tags are approximately clustered according to frequency of use together, and the colors help visualize the clustering.

Here are some previous attempts. The clustering/coloring means nothing here.

Recipe, code and images at GitHub (e.g. tweak the plot to your taste; feel invited to co-develop)

Credit Piotr Migdal for the scripts and instructions.

  • Btw, if anyone's good at using Gephi, I have the .gephi files for this if you want to experiment. Ping Alenanno in chat if you want them. Commented Dec 6, 2012 at 15:04
  • I agree it's pretty, and pretty cool. But what does "frequency of use together" mean?
    – jlawler
    Commented Dec 10, 2012 at 15:03
  • @jlawler: It's the likeliness of question X having tag B given that question X has tag A. So, [online-resources] is used often with [resources]. Similarly with [transcription] and [ipa]. A less obvious one (to me) is [parts-of-speech] and [word-classes] Commented Dec 10, 2012 at 15:05
  • Yes, but the tag set is a mishmosh, and randomly applied, mostly by people who don't understand the terminology. So I'm not clear on what's being measured here; same with "popularity", I spose. We don't monitor our membership -- indeed we appear to be advertising at random for more members -- and linguistics (like EL&U, but quite unlike physics) is not a field with a well-defined and commonly-understood set of concepts, let alone abbreviated tags.
    – jlawler
    Commented Dec 10, 2012 at 15:10
  • @jlawler: Actually they're using something called the observed/expected ratio, so it's weighted for popular tags. I don't really understand the stats behind it, you may want to ask Piotr on his original post. It's better to invite him to a chatroom or something, don't clutter up that space. Commented Dec 10, 2012 at 15:13
  • If it's am observed/expected ratio, the "expected" part comes from unspoken assumptions about the distribution of a fairly random list of words. Those are what need explicating.
    – jlawler
    Commented Dec 26, 2012 at 22:59

1 Answer 1


Great job!

I have a request... Is it possible to make the circles less transparent? I think that if you couldn't see the lines behind them, it would yet improve it further. :D

  • 1
    Will do later, got to go soon. Commented Dec 6, 2012 at 15:04

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .