The Popular Tags cloud on ask.me just doesn't

May 12, 2012 5:24 PM

Not Actually Extremely Problematic: The "popular tags" page on the green is a tag cloud, but none of the tags appear much larger than the others.

This is what I'm talking about. Maybe it's just me, but instead of being able to tell at a glance that some tags are just more common than others, the difference between the biggest (ie, most common) tags and the smallest ones just isn't visually obvious. It actually looks more like a list with really sloppy formatting than an actual tag cloud. The size range is from 20 to 30 points; 50% is a big jump in theory, but when it's all run together, It doesn't really impart "This is more common than that" in an obvious intuitive kind of way. Contrast with the tag cloud for the blue, where the range is from 15 to 30, which helps make the size differential a little clearer.

(And, okay, yes, on the How Important is this Really scale, this falls just north of "Obscure bug affecting Internet Explorer for Macintosh," but it's bothering me, okay?)

posted by Tomorrowful to Bugs at 5:24 PM (28 comments total)

Yeah, it's just the result of setting hard numbers on font sizes relative to use, probably 2-3 years ago when we last touched the code. In that span of time, there have been thousands of additional posts, so the tags are relatively all the same size now.

I'm not much of a fan of tag clouds as a wayfinding device anymore (they certainly were all the rage back in 2005 or so) but we'll probably tweak it slightly next week to make it more in line with what was intended.
posted by mathowie (staff) at 5:28 PM on May 12, 2012

I'm surprised the cat tag isn't bigger.
posted by birdherder at 7:18 PM on May 12, 2012 [6 favorites]

Dating depression! Guitar hair health! Legal Linux literature! Marriage math! Security sex shoes!
posted by nebulawindphone at 7:24 PM on May 12, 2012 [8 favorites]

Is this where it's okay for us to voice our dislike of clouded tags-of-many-sizes? Because if it isn't and it turns out to be a jailable offense, then go ahead and lock me the fuck up.
posted by item at 7:45 PM on May 12, 2012 [2 favorites]

Heh. I kind of miss them, and they bring back happy memories of figuring out how to do a crappy Javascript one as a gadget for the crappy social network I was working for,but they were never really that useful for anything.

Would totally like more detailed and useful tag use information though, across the site and per user - beyond the current paltry ten.
posted by Artw at 7:58 PM on May 12, 2012

Happy to see the batshitinsane tag making a good showing.
posted by arcticseal at 8:00 PM on May 12, 2012 [3 favorites]

I'm surprised the cat tag isn't bigger.
Hah, seriously!

I would feed this pretty pony the healthiest pony food!
posted by two lights above the sea at 8:50 PM on May 12, 2012

Speaking of tags, one thing that caught my eye was "xp", way down at the bottom. It seemed to me that that may have been hotter in the past than in the present. So I downloaded the tag file from the infodump, and did some analysis. I pulled the year with each tag in AskMe, and calculated a weighted score for each tag, treating the years like a halflife (i.e. each time a tag was used in 2012* are weighted 1, 2011 are weighted 0.5, 2010 are weighted 0.25,...). This weights tags that are more recently popular higher, and penalizes things people aren't interested in anymore.

Ranking by this metric, instead of total count of tags, doesn't change the top 150 all that much. resolved, music and travel stay the top 3, and the next four are the same, but mac, food, work, books becomes work, food, books, mac. The rest of the top 10 switch in a microcosm of one of the key trends; computer, software, and internet drop out and are replaced by relationships, job, and health.

16 tags of the top 150 would be different, ranked by halflife. The ones that would drop down out of the top 150, in decreasing order, are:
recommendations, math, linux, media, advice, database, radio, mp3, networking, dvd, network, html, microsoft, php, flash, xp (from 96th down to 320th!).

On the other hand, the 16 new tags, in descending order, are:
facebook (80th!), ipad, therapy, android (up from 343rd to 99th!), shoes, friends, interview, friendship, dc, dogs, life, losangeles, macbook, doctor, statistics, gradschool

And while I had the data out, I calculated an index of "hotness" for all tags used at least 40 times; the total counts divided by the halflife count, normalized. These tags in the middle are the timeless ones; books, movie, sanfrancisco, wedding, food, sex.

The old and busted tags read like technology's buzzword graveyard, circa the early to mid 2000s; aol, psp, http, lan, winxp, tiger, www, thunderbird, shareware, weblog, adsense, pda, palm, aim, myspace, p2p, ps2, ibook, divx, im, mozilla, powerbook, movabletype, treo. There are a few nontechnical ones scattered in there as well; iraq, globalwarming, election, bush, katrina. The oldest tag? web2.0.

The new hot tags have lots of current technology references as well; ios, ipad, android, ereader, tumblr, cloud, win7, kindle, droid, app, chrome, etsy, socialmedia, facebook, tablet, twitter. But there's a few new trends bespeaking a more diverse audience; lgbt, transgender, queer, polyamory, kink; trauma and ptsd; preschool, daycare and nanny; awkward and boundaries, psychiatrist and selfesteem, lowcarb, radiation, newjob, catfood, horse, bedbugs, istanbul, introvert, whiskey.

AskMetafilter: it's not just for nerds anymore.
Parsing and analyzing AskMetafilter tags on a Saturday night? Still totally for nerds.
posted by Homeboy Trouble at 10:58 PM on May 12, 2012 [48 favorites]

I'm not much of a fan of tag clouds

You'll never make Product Manager at a startup this way.
posted by drjimmy11 at 12:14 AM on May 13, 2012

I don't see why interest in xp should be decaying - everybody wants to level up.
posted by Dr Dracator at 6:08 AM on May 13, 2012 [3 favorites]

I don't see why interest in xp should be decaying - everybody wants to level up.

Is it worth grinding in MetaTalk or should I take on the Music Challenge instead?
posted by The Whelk at 6:46 AM on May 13, 2012 [1 favorite]

So... it is quite literally:

Metafilter: books, movie, sanfrancisco, wedding, food, sex?
posted by meinvt at 7:36 AM on May 13, 2012

AskMetafilter: your treo is old and dated, no one wants to have sex with you.
posted by arcticseal at 8:08 AM on May 13, 2012

I'm surprised the cat tag isn't bigger.

Come on, both cat and cats are on this list! That's gotta be skewing the numbers.

(Actually, that's making me want to drag out the research I did for Machine Leaning on identifying synonyms in tag clouds - come to think of it, modern Mefi probably would make for a much better corpus than the half-assed de.licio.us scrapes I was using...)
posted by Tomorrowful at 9:17 AM on May 13, 2012

come to think of it, modern Mefi probably would make for a much better corpus than the half-assed de.licio.us scrapes I was using...

Indeed. The modern MeFi corpus has an ass quotient of 0.87 as of the latest update.
posted by FishBike at 9:26 AM on May 13, 2012

books, movie, sanfrancisco, wedding, food, sex

Sounds like a good day to me.
posted by ook at 11:52 AM on May 13, 2012 [6 favorites]

Parsing and analyzing AskMetafilter tags on a Saturday night? Still totally for nerds.

High five, yo.

come to think of it, modern Mefi probably would make for a much better corpus than the half-assed de.licio.us scrapes I was using...

Boy oh boy do you need to take a look at the Metafilter Corpus Project.
posted by cortex (staff) at 12:52 PM on May 13, 2012 [3 favorites]

Is it worth grinding in MetaTalk or should I take on the Music Challenge instead?

So here's the deal.

Due to the archaic "open world" paradigm that Metafilter uses - we're talking the stuff that stretches back to diku and beyond - you're constantly in competition for xp. This was fine back in The Day when the population was small and manageable; now it's a pain in the ass. Music Challenges have some great rewards, but the people there have basically made it their own domain, and breaking in and getting recognition xp (which you can later convert into regular xp) is really difficult. It's doable, but unless you already have a fair bit of either Music or Humor points, you're probably not going to be able to hack it.

So MetaTalk seems like it'd be a good place to grind. Lots of contention xp is available, and occasional threadjacking can give you a boost. There are less posts per day so your actions get you proportionally more xp for the effort you put in.

Another option, though, is doing Ask raids. You're limited to once a week, but a) most people don't bother to keep to a strict schedule, meaning there's room for you; b) a good raid can be talked about and referenced for years, and gets you xp each time it is; and c) most of the raiders as casuals, so if you're gunning for top spot you don't have a lot of competition.

Projects and IRL are endgame content and if you don't already know about those you're probably not really ready for them. IRL doesn't net you anything other than social xp, anyway, and that has a crappy exchange rate.
posted by curious nu at 1:46 PM on May 13, 2012 [5 favorites]

IRC is downloadable content - usually liquid and a pleasant amber color.
posted by Artw at 2:34 PM on May 13, 2012 [1 favorite]

That should be IRL, of course. Do we have IRC? That should be like some kind of bonus content easter egg thing.
posted by Artw at 4:00 PM on May 13, 2012

That's what meMail is for.
posted by The Whelk at 4:20 PM on May 13, 2012

the research I did for Machine Leaning

Did you figure out once and for all why robots are so damned lazy?

If they have time to lean, they have time to kill all humans.
posted by Sys Rq at 6:30 PM on May 13, 2012 [3 favorites]

I am sad, but not surprised, that "delete," "deleted," AND "deletion" all make the biggest of the top 150 MetaTalk tags.

(It's also interesting that MetaTalk's tag cloud is much more differentiated in size than MetaFilter or AskMetafilter. Music, Jobs, and IRL are also much easier to get meaningful information from - I can't help but notice that my city, which has only been used as a tag seven times, makes the IRL list. This suggests that, well, Matt's explanation is so totally the right one.)
posted by SMPA at 7:48 PM on May 13, 2012

This thing that ninemsn uses and calls 'The Flock' has always intrigued me as in interesting way to try and capture traffic to various items in real-time. I can't for the life of me figure out what a use could actually do with it, but it's pretty.
posted by dg at 9:24 PM on May 13, 2012

Hey, Homeboy Trouble's comment got onto the Best of Mefi blog. Cool. I want to see the "new hot" results as a tag cloud.
posted by Pronoiac at 10:55 PM on May 13, 2012

Here's some tag clouds for the most popular* AskMe tags by year, with relatively consistent layout:
2005
2006
2007
2008
2009
2010
2011
2012**

If you look at one at a time, the trend isn't as obvious as when you tab through them quickly. Or follow one word through time; anxiety and mp3 are good ones.

* I stripped out 170 "evergreen" tags, those that were in the top 300 but were the most consistent from year to year (specifically, the total count/halflife weighted ratio was +/- 10% of the average). I made a separate word cloud of these here. I also removed the "resolved" tag, which came into common use in 2009; it would dwarf the words in the cloud.
** The infodump only has January, so this is a small sample size.
posted by Homeboy Trouble at 7:31 PM on May 14, 2012 [1 favorite]

Wordle requires Java. Could you post screenshots?
posted by Pronoiac at 12:01 AM on May 15, 2012

I have some pictures of Homeboy Trouble's tag clouds.

I reinstalled Java just for this, and I thought it might save someone else some effort.
posted by Pronoiac at 2:20 PM on June 2, 2012 [1 favorite]