Visualizing most popular tags

January 23, 2005 9:31 PM

Right now, the top-150 tag page seems like sort of a novelty. "Art", "music" and "flash" are going to be huge, but the vast majority of tags seem to get shunted into the lower mid-range such that there's not a lot of perceived difference tag-to-tag or day-to-day. Over time, I'd imagine we'd see the same sort of homogeneity that characterizes the Flickr tag sizes.

The idea is to add some more dimension to the information display, showing not only popularity, but time-based "freshness" of tags and ordering them alphabetically rather than hierarchically for easier browsing ala the full tags listing. If time could be added as an organizational principle, people who aren't obsessive readers of mefi could use the top-150 page as a viable entry point to the site for finding new posts of interest.

Quick-and-dirty HTML example here. (or screenshot).
posted by Jeff Howard to feature requests at 9:31 PM (19 comments total)

A few other possibilities, inter-tag divisions to help with discrimination at smaller sizes, consistent line-height and an upper limit on the size of tags. When the top-150 have such a huge range, it seems like it might be enough to pick a cutoff point and simply make those top tags bold, rather than overwhelmingly large.
posted by Jeff Howard at 9:32 PM on January 23, 2005

I like it! Having a tag page that is limited, say, to the last month or so would be a great way to browse recent posts.
posted by vacapinta at 9:36 PM on January 23, 2005

Having a tag page that you could pull up by date range would be a great way to browse MeFi's history, too.
posted by caitlinb at 9:45 PM on January 23, 2005

So as the most recent occurrence of a tag recedes into history, the font gets bluer? That's really cool. Very elegant.
posted by me3dia at 10:08 PM on January 23, 2005

Agreed, nice job. How about displaying the number of posts (and latest post date or other freshness measure) on mouseover using the title tag?
posted by fatllama at 10:22 PM on January 23, 2005

Worth mentioning is that the tags are currently sorted by size, and displayed in different sizes, according to size. Considering that there are tags of all sizes, forming a more or less steady continuum, this is kind of redundant. I'd almost rather have them all be one readable size, and count on the sort order to represent popularity. Perhaps a number in parentheses (262) would help. It wouldn't look as pretty, but it would provide more information, in a way that allows the page to be readable.

Really, it should be a list, with a couple columns, sortable by number of entries, date of most recent entry, perhaps other stats people might think up.
posted by scarabic at 10:34 PM on January 23, 2005

I like the sizes. The beauty of having the data available, as I've said so many times before, is that there's any number of really cool ways of presenting it.
posted by stavrosthewonderchicken at 10:35 PM on January 23, 2005

93 posts tagged with quonsar

Hypertroll.
posted by orange clock at 10:41 PM on January 23, 2005

Neato! But 9-11 should be filed under "N"! :-D
posted by shepd at 10:48 PM on January 23, 2005

The sizes thing is going to get worse over time, though. You've got ART and MUSIC taking up half the page, while very popular tags like "government" and "environment," that are in the top 150 for a reason, are in tiny print, dwarfed and shoved down beneath the fold. I know ART and MUSIC are bigger than "government," but I don't need them to take up my entire browser window just to communicate how many dozens of times larger they are. The sizes thing is neat for a managebable number of tags of relatively similar size, but for 150 tags of vastly different sizes, it gets clumsy.
posted by scarabic at 11:21 PM on January 23, 2005

Orders of magnitude of unimportant distinction got you down? A few big outliers giving you the blues in the Blue? Apply logarithms before determining font size!
posted by fatllama at 11:35 PM on January 23, 2005

scarabic, you're right about the redundancy of order+size. That's what got me thinking in the first place. Browsing a list of 150 tags ordered by popularity is doable, but an alphabetical listing makes it easier to find whether a specific tag (say, quonsar) exists or not.

on preview... also scarabic, that's the rationale behind picking a more managable upper size limit for the really popular tags, and using bold to compensate. They should still stand out, but perhaps not dominate to the extent you mention.

me3dia, right, the fading makes it easier to see that, for instance, "election," which had a ton of posts in the fall, is starting to fade away in relevance. It supplements the long-term mapping (size) with short-term mapping (color). The problem is that it ruins the normal link color mapping (yellow fading to blue produces a sickly green) and makes it harder to show visited links.
posted by Jeff Howard at 11:41 PM on January 23, 2005

I put a limit of 50px on font size for both the all tags and 150 top tags list.

Unfortunately, due to the way I'm doing the SQL query, I can't seem to order the top 150 any way besides popularity (vs. alphabetical), but I'll see if I can figure it out.

I'm compelled to make the top 150 tag page the default, with the all tags page being a page away, since it's quickly becoming so large as to be difficult to use.
posted by mathowie at 12:04 AM on January 24, 2005

The color-fading is pretty cool, but I don't think it would be very practical. One obvious disadvantage that I can think of is that visited links could no longer be distinguished. It makes a great visual demo, but in practice would result in too much information conveyed through typography styles alone, and I don't think it could be done without losing some current functionality.

I do like the idea of Top tags lists for the last week/month/year.
posted by stopgap at 1:58 AM on January 24, 2005

I'd like to be able to drill down. Browse art by author, topic, etc. If there were some way of making this a Wiki to harness the member energy - that could be amazing. I know the members here would put a lot of time into organizing and caring for the tags. Also, I know that if the legacy posts were included I would go back and at least tag my own old posts.
posted by xammerboy at 5:59 AM on January 24, 2005

xammerboy: you can tag your old posts right now.
posted by stopgap at 6:22 AM on January 24, 2005

Setting the top 150 as an entry point instead of the full list would be nice. Although it may even itself out over time, the full list has a lot of tags with only one post... and that's not really a category.

Is there a way of finding either list other than clicking through from a tag detail page?
posted by Jeff Howard at 8:49 AM on January 24, 2005

Re. visual display of ranking of top 150. Instead of size, which it looks like you're having to cap at the top end, how about using a colour scheme. This would require something like calculating a relative % value for each of the 150 tags, expressed as a % of the total number of tags in the top 150. You could then calculate a # for the colour of each tag based on this %, perhaps shading (e.g.) from MeFi blue at the top end to MeFi yellow at the bottom.

You'd probably have to ballpark what the possible highest percentage is, and make that blue from the start, as no tag would have a value of 100% of the top 150 tags. However as the distribution is Zipfian, I think this highest % would stay quite steady over time, say (off the top of my head) 10%.

Anyway: in the case of the 150, 'art' and 'music' would be Mefi blue, everything say from 'japanese' on would be pretty much yellow, while stuff in the middle of the distribution (e.g. media) would be greenish.

These are really just my 2c, as I have no idea of how any of this is being achieved.

Re. the 'all tags' page: I'd love to get absolute numbers for these, if only because I like pasting this stuff into Excel and making nice charts. Here's a nice way to do it, although it may be harder to do with the large number of tags on the 'all tag' page.

Final small thing: If I have my prefs. set to 'plain' - that is, white background in MeFi front page and inside pages - I can hardly read the light grey tags.

But as I said - just my 2c - I think this is all way cool.
posted by carter at 9:11 AM on January 24, 2005

xammerboy: you can tag your old posts right now.


(wakes up) Huh? Whuzzah? Oops! Thanks!
posted by xammerboy at 9:52 AM on January 24, 2005