# Map of 64 Tags from Physics.SE

I've just made a map of tags for Physics.SE.

In short: tag size is related to tag popularity and edges are related to tag co-occurrences in questions.

For me it looks as a "snapshot" of topics and scope of this SE site. So if you want to you it in any way to promote Physics.SE - feel free!

Also, if you have comments how to improve its usefulness or niceness to our community, I would appreciate them (but bear in mind that I have no color esthetics).

-
Nice :-) I might use it in a (personal) blog post later this month to promote your work and the site. –  David Z Nov 1 '12 at 17:29
@DavidZaslavsky Nice to hear that! :) So just let me know (e.g. "+" in G+) when it happens. –  Piotr Migdal Nov 1 '12 at 17:59
This is a cool representation of our community and it would make a good motive for a T-shirt :-) Can it be made dynamic such that it automatically adaptes the size and the links among the tags to include the newly incoming questions? –  Dilaton Nov 1 '12 at 20:07
@Dilaton Dynamic version (most likely with d3js.org + SE API) is in my plans. However - depending on my energy & time, it may be sooner or later. But remember that dynamic, interactive plots may have some compatibility issues with T-Shirts ;). –  Piotr Migdal Nov 1 '12 at 20:17
I like it a lot. Now what I want to add is some way to overlay a particular user's footprint (assuming the data can be obtained) - are they more of a particle physicist or a cosmologist? Maybe I'll play with the code. Odds of success: $\sim\varepsilon$. –  Chris White Nov 2 '12 at 8:24
Ha ha @ChrisWhite :-D, I thought about a similar idea about having the possibility to let each user draw such a diagram for the tags he uses (with the size given by # each tag is used or the points obtained from it). I personally should probably better refrain from using it, if such a feature existed, because it would make me look like a really really really bad geek :-P –  Dilaton Nov 2 '12 at 10:44
Entropy being exactly in the middle, that doesn't make sense ;) –  Bernhard Nov 2 '12 at 11:47
@Chris not too hard to do that, in fact. If you look at the instructions on github, you first have to download a CSV of a certain data.SE query. It's not too hard to fork the query and add a WHERE P.OwnerUserId=##UserId## clause in a few places. I've tried it out, see data.stackexchange.com/physics/query/84235/…. Just use the CSV from that and follo the rest of the procedure. –  Manishearth Nov 4 '12 at 5:32
@Manishearth Just a small warning - here statistics matters. That is, I plotted 64 tags for the whole physics.SE, as for 128 tags there was too much noise (e.g. edges for only 1 co-occurrence). Instead, I was thinking about making a heat-map over the full graph. Anyway - did you try plotting it for any particular user? –  Piotr Migdal Nov 4 '12 at 10:27
@PiotrMigdal: Not yet, I've not got access to my Ubuntu laptop atm. I'll try it later, and I'll edit the query to take care of the 64-tag restriction. –  Manishearth Nov 4 '12 at 10:35
@PiotrMigdal: How did you get them grouped like that? Is there some procedure? TH one on the GH README isn't doing the trick.. –  Manishearth Dec 5 '12 at 19:10
@Manishearth You mean - colored or positioned? (And what's "TH"?) –  Piotr Migdal Dec 5 '12 at 19:16
@PiotrMigdal: Colored AND positioned. The Phys.SE one is pretty, I can't manage to make it that pretty for chem and Linguistics :\ And the grouping is meaningful (it looks meaningful, at any rate), so that helps. "TH" was me mistyping "The" :P (and GH=GitHub) –  Manishearth Dec 5 '12 at 19:21
@Manishearth Strange, it is there Layout -> ARF -> Run for an example Layout (you can try others), Statistics -> Modularity -> Run then Partition -> Nodes -> Refresh -> Modularity Class -> Apply are for the coloring. Which part isn't working for you (I know that the Gephi UI is not the intuitive or even logical). –  Piotr Migdal Dec 5 '12 at 19:25
@Dilaton Anyway, to get graph I need to use all data, data from 3 months would not suffice for correlations (but very likely is perfectly enough for just graph counts). BTW: If you just want to see tag popularity over time, it's enough to use data.stackexchange.com. E.g. data.stackexchange.com/stackoverflow/query/83404/… plus a filter on post dates. –  Piotr Migdal Feb 8 '13 at 12:13

Is short - links are only for very specific connections. If a given topic os connected with many things, but roughly in the same way, then there are not many links from it. It's an effect of taking $P(\text{tag1}\cap\text{tag2})/[P(\text{tag1})P(\text{tag2})]$ as measure of being connected. See also discussions here on "homework" and "linear algebra". Otherwise it would be hard to plot it (tried, failed) as some topics would be connected to almost everything (e.g. quantum mechanics) while many other - to nothing. –  Piotr Migdal Nov 5 '12 at 20:04