Can perennial AskMe questions be grouped for easier reference? January 23, 2007 11:34 AM   Subscribe

Ask Metafilter: some intuitive interface to Time-Periodic Frequently Asked Questions...(more inside)
posted by felix to Feature Requests at 11:34 AM (73 comments total) 1 user marked this as a favorite

For instance:

* Christmas posts ("what should I get my xyz for Christmas") are usually mostly back-referrable ("you should get her a diamond car", "try hipster t-shirts from www.example.com") no matter what year you're in, but also gain new information yearly (in 2006 it was elmo extreme or whatever).

* The question of how to get video on the web is mostly the same answers as it was last year, but may include new information related to codecs as of this year.

* How do I deal with winter depression, lose some weight for the swimsuit season, etc.

* How do I buy a car, etc.

Many of the questions out there involve some form of basic understanding which persists from year to year, but may also include adornment from this year. Although it it possible to search through every post from every period related to the issue, it would be nice if there were some mechanism by which they were automatically grouped, both to limit the profusion in the future and also to keep all of the related information in the same place. Please make the mane brown and the hooves white.
posted by felix at 11:41 AM on January 23, 2007


This is the sort of thing that could be accomplished by a volunteer with some time on their hands—review the archives post-by-post and compile an organized list of topic/time relevant threads. It'd be manual, and require manual maintenance, but for a limited set of domains would be practically doable.
posted by cortex at 11:46 AM on January 23, 2007


cortex, did you have your hand raised while you wrote that?
posted by goodnewsfortheinsane at 11:55 AM on January 23, 2007


Good idea. The Ask/Tell section of the SomethingAwful forums has somethign similar: a 'sticky' thread with links to commonly asked questions.
posted by PercussivePaul at 11:57 AM on January 23, 2007


I'm a bit confused about what's being proposed here. I get and like the general idea of gathering similar questions together to create an information repository. I don't get how "how do I buy a car" and "how do I get video on the web" are 'time-periodic' questions -- can someone explain how they fit into this?
posted by chrismear at 12:14 PM on January 23, 2007


USE THE TAG SYSTEM PROPERLY, FRIEND(S)
posted by dorisfromregopark at 12:22 PM on January 23, 2007


I think the word is perennial.
posted by OmieWise at 12:33 PM on January 23, 2007


I think it's more that felix is saying there should be a way of labelling (and easily finding) old-but-still-relevent info.

I think that's true -- I think the archives are under-used -- but I'm not sure what would be done with it after that?
posted by occhiblu at 12:36 PM on January 23, 2007


USE THE TAG SYSTEM PROPERLY, FRIEND(S)

A solution as elegant and apt as it is unimplementable. The question is, how do we compensate for the observed failure of tagging* as a rigorous description/organization mechanism. One clear answer: single or small-group dedicated human agents with explicit tasks.

*failure in a strict sense. I get a lot of use out of the tags, regularly, the just aren't thorough.
posted by cortex at 12:39 PM on January 23, 2007


And you are as attractive as you are correct, Cortex (hint: very). This no-content producing lurker encourages all attempts to make efficient use of AskMe's large database of questions. I recall some brief talk a few months ago of organizing a group of volunteers to go through AskMe and label questions with proper tags. felix's idea is in the same vein and I acknowledge it with an e-thumbs up of limited value.
posted by dorisfromregopark at 12:57 PM on January 23, 2007


One clear answer: single or small-group dedicated human agents with explicit tasks.

What ever happened to the idea of a archive-tagging-super-friends volunteer corps?
posted by joshuaconner at 12:59 PM on January 23, 2007


We were building tag urls just a few posts down; it would be cool if one could integrate the archival year into that system somehow. The most obvious way coming to mind would be to tag each question with its respective year. Are auto-tags an offensive bauble? Would auto-retro-fitting be difficult?
2007
posted by carsonb at 1:46 PM on January 23, 2007


One clear answer: single or small-group dedicated human agents with explicit tasks.

Draft some specifications, then maybe we can all pitch in to Mechanical Turk it.
posted by rkent at 1:58 PM on January 23, 2007


We need community tagging. Seriously. Have two lists of tags in the sidebar for each post: poster tags, and the top 5 or 10 tags by other community members.
posted by monju_bosatsu at 2:00 PM on January 23, 2007 [4 favorites]


Seconding monju_bosatsu. Community tagging is a brilliant idea.
posted by vacapinta at 2:05 PM on January 23, 2007


Are auto-tags an offensive bauble?

They'll offend some. monju has the right idea: make an explicit distinction between author tags and community tags. Autotagger3000 could be a member of the community for communal tagging purposes.
posted by cortex at 2:07 PM on January 23, 2007


monju is as attractive and correct as cortext is. Community tagging for ask and mefi, plz.
posted by boo_radley at 2:13 PM on January 23, 2007


Community tagging

Tags:
* obesity
* fast_food
* weight_gain
* statistics

Top 5 Community Tags:
* double
* flagged
* hurf_durf
* this_will_not_wendell
* OMGWTFBBQ!!!!!!
posted by muddgirl at 2:19 PM on January 23, 2007


muddgirl: you forgot "batshitinsane".
posted by matildaben at 2:32 PM on January 23, 2007


having said that, I'll present the easy counterargument of having all of my posts tagged "insufferableDickbag" as less than optimal.
posted by boo_radley at 2:33 PM on January 23, 2007


OMGWTFBBQ!!!!

What if questioner could delete community tags? 'Cause you're right that any community participation here wants moderation.
posted by carsonb at 2:35 PM on January 23, 2007


At the risk of presuming administrative willingness, I'll argue that community tagging is not the same as wholly unchecked and immutable laissez-faire tagging. Shitty tags could be removed; shitty taggers could be cut off.
posted by cortex at 2:37 PM on January 23, 2007


Also, that's precisely why I suggested that only the top 5 or 10 community tags get displayed. Sure, some joker is going to add 'hurf_durf', but unless it's in the top 5 or 10 tags, it won't show up.
posted by monju_bosatsu at 2:48 PM on January 23, 2007


cortex: that idea smacks of effort, though.
posted by boo_radley at 3:01 PM on January 23, 2007


Granted. A dangerous proposal, in that sense.
posted by cortex at 3:04 PM on January 23, 2007


it won't show up.
Do you really think that? Maybe I should apologize for the snarkiness, and be a bit clearer. I don't think we metafilterians (as a community, not individually) are mature enough to handle community tagging. It would cause an astronomical amount of work for the moderators. I can already imagine the MeTa bickering.
posted by muddgirl at 3:04 PM on January 23, 2007


I guess you're more pessimistic than I am. I tend to think that the troublemakers are few and vocal. Community tagging lets the large body of otherwise silent mature users overcome the vocal few.
posted by monju_bosatsu at 3:15 PM on January 23, 2007


Unfortunately, we have this:

<venn name="A">People Who Know How To Search Via Tags</venn>
<venn name="B">People Who Ask Repetitive Questions</venn>

And not this:

<venn name="a">People Who Know How To Search Via Tags
<venn name="b">People Who Ask Repetitive Questions</venn>
</venn>

nonetheless, the idea of community tagging per post is still a good one.
posted by fishfucker at 3:20 PM on January 23, 2007 [1 favorite]


Why not limit the amount of community tags you could hand out to X amount per day or week, and allow community untagging--i.e., the ability to vote against a certain tag--with the same restrictions?
posted by arto at 3:23 PM on January 23, 2007


Tagging needs to be low friction to work.

Perhaps if the users were given guidance for their tags, much like delicious provides. Perhaps the list could be constrained, rather than freeform data entry.
posted by felix at 3:27 PM on January 23, 2007


Seconding monju_bosatsu. Community tagging is a brilliant idea.

I suggested it when tags were first being discussed, repeatedly, before Matt implemented them.

So there. *sniffs to hold back a tear*

My take on it was that users should be able to tag anything, and have an option on their profile to either share all their tags (ie have them accrete into the cloud) or not (ie have them be visible only to themselves). Tag search/related-topic browsing would get better by orders of magnitude.

The db hits would be pretty staggering, though, no doubt, depending on how and when they were displayed.
posted by stavrosthewonderchicken at 3:49 PM on January 23, 2007


(Delicious has for a while offered the option to make saved items private or public. I wonder if tags slapped on private items also remain private... )
posted by stavrosthewonderchicken at 3:51 PM on January 23, 2007


<Svenn name="A">People Whose Name is Sven</Svenn>
<Svenn name="B">People Whose Name is Svenn</Svenn>
posted by It's Raining Florence Henderson at 4:04 PM on January 23, 2007


On AskMe, 2004 - 2006 inclusive


There were 48043 total questions.

Tags were applied 146465 times to 39355 questions, leaving 8688 questions untagged.

Of the 36064 unique tags used:

23910 tags were used 1 time
4379 tags were used 2 times
1969 tags were used 3 times
1112 tags were used 4 times
761 tags were used 5 times
554 tags were used 6 times
420 tags were used 7 times
319 tags were used 8 times
267 tags were used 9 times
223 tags were used 10 times

The remaining 2150 tags were used 11 times or more a piece.
posted by tkolar at 5:35 PM on January 23, 2007


Oh, and because there just has to be a top ten list:

Top ten posts by number of tags:

10. 20 tags
9. 20 tags
8. 20 tags
7. 20 tags
6. 21 tags
5. 23 tags
4. 23 tags
3. 23 tags
2. 24 tags
1. 28 tags
posted by tkolar at 5:58 PM on January 23, 2007


fishfucker has it. The people who ask the moron questions by and large aren't the people whose valiant search efforts are frustrated by inefficient tagging; they're the people who don't search at all [NOT SEARCHIST]. If that is right, then there would be no tangible decrease in moron questions as a result of this policy. So, is the increase in AskMe moderation effort and increase in MeTa posts worth a small efficiency bonus for diligent searchers? I say no.
posted by Kwine at 6:21 PM on January 23, 2007


I wonder how often doing a Yahoo search of the AskMe archives using every word of the person's prospective question would yield a useful "Top 10 questions on this topic" list.
posted by tkolar at 6:32 PM on January 23, 2007


Reading through this post again, I thought for a moment near the end for sure that I was at a Royal Society meeting in the late 17th century. Brute numbers open for interpretation, grand hypotheses on traits and tendencies of The People Who, brilliant and useless bickering adjustments for hot fixes— a beery head of delicious thinking.

Love the age you're living in, I guess. Cheers!
posted by carsonb at 6:52 PM on January 23, 2007


MetaFilter: A Royal Society Meeting In The Late 17th Century
posted by tkolar at 7:17 PM on January 23, 2007


What ever happened to the idea of a archive-tagging-super-friends volunteer corps?

I'm still into it, but it needs some implementation by #1 as well.
posted by jessamyn (staff) at 7:29 PM on January 23, 2007


maybe there should be a tag or way of denoting a particularly good post or website as a "MetaFilter Resource" i..e. a definitive answer that stands the test of time.
posted by pjern at 10:33 PM on January 23, 2007


tkolar's post it the best numeric indictment of folksonomy as a retrieval tool I've ever seen. If nearly half the total tags used are unique, people don't have a clue how to operate in a folksonomy, and it's pointless as a retrieval tool. The meta-noise is completely swamping this folksonomy, and users are reduced to stemmed guessing for the 23,910 unique tags by which wisdom might specifically be filed. Thanks, tkolar, for the numbers. You win the thread, hands down.

So the intelligent, not the clueless behavior, is to never bother with tag search, because it demonstratably won't exhaustively answer searches. Instead, it's inherently broken and untrustworthy, and a piss poor substitute for a taxonomy.
posted by paulsc at 11:07 PM on January 23, 2007


That's incorrect. The dataset is not big enough. With thousands of people all able to tag everything (not merely the very rare threads they personally start), you would see convergence begin to emerge.

Taxonomy has its place. 'Folksonomy' also. They are not mutually exclusive. To suggest that the system as it stands (here, and everywhere is your implication) is 'inherently broken and untrustworthy, and a piss poor substitute' is a bit like saying democracy doesn't ever work, merely because Idiot America elected George Bush a couple of years ago. Sample size is important. It's a very different argument to suggest that democracy never works, based on the results of all elections for which we have data.

With a ratio of total tag instances to questions of about 3:1 (for a swarming multitude of topics), we haven't got enough data yet, I don't think. An order or two of magnitude more might be appropriate.

And with all due respect to tkolar, it seems unlikely that nearly 24,000 unique tags (words or word combinations) were used. That seems to me like an awful lot of uniques, particularly for a total of about 40,000 threads tagged. But regardless, the ratio of total tags:threads is too small by a couple of orders of magnitude to see much convergence.

I don't really mind much, either way. But if we're going to have tags, it seems like a good idea to leverage them as much as we can.
posted by stavrosthewonderchicken at 11:39 PM on January 23, 2007


it's inherently broken and untrustworthy, and a piss poor substitute for a taxonomy.

Well, we do have a little bit of both in AksMe and I think it would be great to be able to combine categories and tags for further searching. The part that I think gets missed from tkolar's list is the way he chops it off at ten uses, noting "The remaining 2150 tags were used 11 times or more a piece." Without this line, which I missed at first, I'd be inclined to agree with you. But having 2,000+ tags which are being used eleven or more times each, that seems to imply that there is at least some coalescing happening. I'm also a little surprised at the high number of unique tags, I'd love to see a partial list of them.

Also, tags are used not just to find previous examples of a question or type of question, but also to have customized RSS feeds so that people can browse questions in their preferred area as well as just idle "I wonder what questions about Seattle have been asked?" speculation. They clearly are not a great tool to create a faq out of, and they require some know-how to search for previous questions [somewhat akin to knowing how to put phrases into quotes in Google, which we've seen a few times lately] but I don't think this means they are are either broken or untrustworthy, just less than perfect, which we've always known.
posted by jessamyn (staff) at 11:49 PM on January 23, 2007


If nearly half the total tags used are unique, people don't have a clue how to operate in a folksonomy, and it's pointless as a retrieval tool.

Depends on what you're looking for. Try the name of a well-known city for travel tips: Boston or London, say. Works for me.

To say its useless because many people have fun and type in useless and unique tags is slightly missing the point. I'll never search on those tags so extra tags do not increase the noise.
posted by vacapinta at 12:54 AM on January 24, 2007


"... Taxonomy has its place. 'Folksonomy' also. They are not mutually exclusive. ..." Huh? Where'd I say they were? In terms of utility, however, the value of a planned, navigable taxonomy that understandably navigates to every node of a knowledge base is so obviously superior to a folksonomy that doesn't even include the 8688 questions that are entirely untagged, as to make an enforced taxonomy worth learning, and folksonomy of the kind we have in AskMe unlearnable.

"... I don't really mind much, either way." Maybe you don't stav, but the AskMe question time out has been raised to 2 weeks, because so many people are popping in to ask questions, many of which have probably been answered sufficiently, and perhaps even better than they will be afresh, if people had an understandable, exhaustive, and reliable means of navigating the knowledge base. Since the folksonomy is none of these things, it's largely unused, and not worth recommending, except as entertainment. A FAQ and a taxonomy or other exhaustive search/retrieval mechanism are the minimums for any site that wants to take itself seriously as a resource.

"... But having 2,000+ tags which are being used eleven or more times each, that seems to imply that there is at least some coalescing happening." That's damning with faint praise, if ever I've such. And it doesn't mean much if half that list are member words of the 1,000 most common words in Flemish, which until we see such a list, we don't know they aren't. In fact, such a list would be the start of a folksonomy recommended taxonmy, which would be an interesting index to build, but hardly exhaustive in a space where 16% of the total items (8688 out of ~48,000) are entirely untagged.

"...Depends on what you're looking for." My point, exactly.

Personally, I just view AskMe as green colored version of Very Trivial Pursuit.
posted by paulsc at 1:14 AM on January 24, 2007


"if ever I've heard such."
posted by paulsc at 1:19 AM on January 24, 2007


The huge hole in your argument, paulsc, is that currently (barring a handful of metafilterhistory tags added by jessamyn) all the tags are added by the poster. Because the same person can't tag more than their handful of posts, it's hard to get the kind of consistency you crave. If tagging brigades were formed, that would change pretty drastically.

The merely big hole in your argument is that tkolar's data set is posts in 2004/2005/2006. Tagging was introduced sometime in early/mid 2005 – nearly all of those posts with no tags at all were posted before tagging was implemented at all!

In fact, there were significantly more than 8688 questions posted between the start of his data set to the date of tagging implementation — meaning that a significant number of questions had their asker go back and tag them long after the fact.
posted by blasdelf at 3:22 AM on January 24, 2007


paulsc writes "it's pointless as a retrieval tool. The meta-noise is completely swamping this folksonomy, and users are reduced to stemmed guessing for the 23,910 unique tags by which wisdom might specifically be filed."

tkolar's numbers are great, but I'm not sure that they say what you think they say. First, as has been pointed out, unique tags add no noise. If I tag a question about Boston with the tag "Boston" and also with "Beantown," and Beantown is unique, it matters not at all. The only problem is if the only tag I use is Beantown. We need numbers that indicated how frequently questions have only one unique tag, that would give us a bit more information, although it would still be limited because some of those questions may well be about discreet subjects.

More to the point, the numbers show that tags are perhaps not as useful as they could be, which is a far cry from "pointless." Certainly there is a threshold for usefulness, although we haven't determined what that is, but I think that many subjects are adequately covered by tags. I've had good success searching by tags, and then, since tags are all links, refining my search from there to other tags. We can argue about how transparent a tag system should be to be useful, but the truth is always going to be that searching for something (whether using tags or not) is a process of winnowing and refining based on progressively selected sets of results. To flip my Boston example around, even if only one post (of the twenty Boston questions) is tagged "Boston", but if that tag shares the "Beantown" tag with 85% of the other posts, I can still easily find that treasure trove of poorly tagged Boston questions. To expect that any solution is going to mimic the way my mind works is to misunderstand the nature of searching for information.
posted by OmieWise at 6:38 AM on January 24, 2007 [1 favorite]


jessamyn wrote...
I'm also a little surprised at the high number of unique tags, I'd love to see a partial list of them.

Here is the full list of tags with the number of times they were used.

stavthewonderchicken wrote...

And with all due respect to tkolar, it seems unlikely that nearly 24,000 unique tags (words or word combinations) were used.


I'll admit that I let a program decide that for me, but you're welcome to read through the list yourself. The one place I'll admit to being to being a bit sloppy is in treating capitalized words as different from their uncapitalized counterparts. If you really believe that would make a big difference in the outcome, I can run the stats again without that.
posted by tkolar at 8:10 AM on January 24, 2007


tkolar- Is it much trouble to re-run the numbers? The tag search seems to ignore case (there's one post tagged americanhistory, and three tagged AmericanHistory, but all four come up in a search for either term. There are a bunch of tags like this (I couldn't believe there was only one post tagged "abortion", for instance, but it turns out to be a case issue, as there are 17).
posted by OmieWise at 8:25 AM on January 24, 2007


Oh, just doing a sampling from the tag list, a really useful number for purposes of discussion would be the number of questions with just one tag, and then a subset where that tag was a tag only occurring once. Since the whole point of tags is to link content with other content, posts with only one tag and a unique tag are the ones most likely to not come up in searches unless one guesses the tag correctly. So far, plugging ~30 single-use tags from the list into the search function, only one "martialartsmovies" was the sole tag on the question. All the other questions had multiple tags, and hence multiple ways to tag-search them.
posted by OmieWise at 8:36 AM on January 24, 2007


The other thing about well-tagged questions, and this really seems to contradict paulsc's post above, is that they often spawn more single-use tags than poorly tagged questions. This single question about a series of children's book has 12 tags, eight of which occur only in this question. One of the remaining four tags only occurs in one other question. One in five questions, and then two in a gajillion questions. But this is as it should be on all levels: the big subject tags (books, fiction) serve to include the question in categories too broad to be useful for answering specific questions, while the intermediate tags (children'sbooks, juvenilefiction) serve to narrow the category for searchers, leaving the specific subject matter tags (betsytacy, etc) to cover this specific author and her series, which have only been asked about once. None of the single-use tags does anything to any search not directed to them, but they are there for the next time someone comes to ask about this series.
posted by OmieWise at 8:46 AM on January 24, 2007


blasdelf wrote...
Tagging was introduced sometime in early/mid 2005 – nearly all of those posts with no tags at all were posted before tagging was implemented at all!

Doh! Good catch.

In fact, over 8500 of the untagged posts were made before the tagging facility was in place.

OmieWise wrote...
tkolar- Is it much trouble to re-run the numbers?

No, but I need to go run some errands so it will be a few hours. Anyone else who would like some stats, now's the time to ask.
posted by tkolar at 9:01 AM on January 24, 2007


paulsc: your insatiable lust for taxonomical correctness will be the death of us all.
posted by boo_radley at 9:38 AM on January 24, 2007


Metafilter: Apparently we're twice as interested in computers as we are in sex.
posted by patricio at 9:38 AM on January 24, 2007


Maybe we just know half as much about computers.

Okay, maybe most of our sex involves computers.
posted by cortex at 9:56 AM on January 24, 2007


Sorry, I forgot a link to my example question.
posted by OmieWise at 10:23 AM on January 24, 2007


OmieWise wrote...
tkolar- Is it much trouble to re-run the numbers?

Okay, ignoring case we get:

32418 total unique tags

21012 tags were used 1 time.
3920 tags were used 2 times.
1867 tags were used 3 times.
1044 tags were used 4 times.
697 tags were used 5 times.
525 tags were used 6 times.
386 tags were used 7 times.
316 tags were used 8 times.
262 tags were used 9 times.
193 tags were used 10 times.

The remaining 2196 tags were used 11 times or more a piece.
posted by tkolar at 11:42 AM on January 24, 2007


A a really useful number for purposes of discussion would be the number of questions with just one tag, and then a subset where that tag was a tag only occurring once.

Actually, given the answer to your first query below, I'm not sure the subset would be that interesting.

I need to take another break, but I'll see if I can come up with an answer for "how many questions only have unique tags applied to them" a little later.

Of the 39355 questions that had tags applied:

2993 questions had 1 tag.
7304 questions had 2 tags.
10375 questions had 3 tags.
8150 questions had 4 tags.
4962 questions had 5 tags.
2572 questions had 6 tags.
1352 questions had 7 tags.
758 questions had 8 tags.
380 questions had 9 tags.
215 questions had 10 tags.

The remaining 294 questions had 11 more tags.
posted by tkolar at 12:00 PM on January 24, 2007


Thanks. There were several thousand fewer single-use tags in that way of counting them.

The second set of numbers it where the good data is, or at least the data that those of us who like tags can speculate with most freely.

According to that, ~8% or questions have only one tag. That's a pretty low number of questions for which you get only one guess at the correct tag to turn it up in the tag sort. That doesn't seem so bad for me.
posted by OmieWise at 12:06 PM on January 24, 2007


tkolar's second set of numbers is interesting, but I see it mostly adding fuel for my central premise, particularly in concert with a perusal of his tag list. The tag list, if you examine it from the most frequently used end down, pretty quickly evidences the common problems with a folksonomy. A n00b to AskMe looking for previous wisdom about a problem with Apple's OS X would probably need to be aware that their search would need to take care of case issues, pluralization choices, brand name alternatives, and God-knows-what-the-hell-were-they-thinking-then issues on the part of original Askers, which is a far cry from the implication of the 150 top tags page. I can see it now, in some future revision of the FAQ:

"Dear n00b:

It's possible that a very relevant answer to your question already exists in the AskMe archives. You might construct a few search strings that incorporate some of the terms (frequency of use in paranthesis) in the following tag string list, to search for such wisdom, about, say, issues with the Apple MacIntosh OS X operating system:

mac (712) + Mac (145) + macintosh (81) + macosx (41) + osx (367) + OSX (136) + apple (344) + Apple (87) + Powerbook (13) + PowerBook (5) + powerbookg4 (1) + powerbook12 (1)

No guarantees, of course that this won't miss relevant answers that were inadvertently tagged 'os (32)' or some such, but tags, as we use them here, are "better than nothing." That'll get you started, any way, and it's entertaining to play with our tag system, if you can possibly find all the tags with which our posts are labeled by their frequently confused creators. Just to confirm your suspicions, 'better than nothing' is the retrieval standard we support with tagging here, and yes, you probably can do as well or better over on Google, where Page Rank with Advanced Search is your beefy pal, if you'd ever read the fine print over there.

Of course, you could just pay your $5, and wail away with your pressing question, once your sign up delay is over. That's what a lot of people who can't be arsed with this tag madness do, any way. Which is why you might be asking a question that has been previously asked 4 times in the last 3 months by others, and why your question may be entirely ignored by those who have most capably addressed it, or very similar topics, 4 times already. Unlike the Blue, here in the Green we loves us some doubles, triples, and nth order rehashes. We're all about fresh approaches to stale topics, and if we're not getting better at them, all the time, we're at least becoming inured.

At the very least, you'll be inflating our use stats, and contributing to endless discussions of post volume regulation over in the Grey, and for this we thank you. n00bies are fun, and we've all been special snowflakes ourselves, at one time or another.

Yours for impractical, if not whimsical archive inquiry,
Your AskMe friends. "
posted by paulsc at 1:33 PM on January 24, 2007


But how do you really feel, paulsc?
posted by cortex at 1:38 PM on January 24, 2007


"But how do you really feel, paulsc?"
posted by cortex at 4:38 PM EST on January 24

With my fingers, usually. Thanks for Asking!
posted by paulsc at 2:49 PM on January 24, 2007


There are 752 questions for whom every tag is completely unique

That number goes up to 3106 questions if you include the tags that were used 10 times or less.

Coming at it from another angle, 20300 questions include at least one tag from the 100 most used tags (tags that were used 149 times or more)



[I'd like to take a moment at this point to reiterate my personal belief that Guido van Rossum is a God. I've had to fall back to C for some of the more CPU intensive queries, but in general Python complements the Unix command line brilliantly. Using good tools is always a pleasure.]
posted by tkolar at 3:09 PM on January 24, 2007


The information isn't in the tag, but in the combination of tags.
posted by Blazecock Pileon at 3:12 PM on January 24, 2007


OmieWise wrote...
There were several thousand fewer single-use tags in that way of counting them.

It's hard to get away from the fact that 94% (93% if you ignore case) of the tags were used 10 times or less. In my mind a tag that is used 10 times or less basically has no effect on the number of re-asked questions AskMe sees.

My take on those tags is that they are largely either a) very obscure, b) uncommon runtogethers of common words, or c) proper nouns. Of those, only proper nouns seem like they would be useful in a tag search, and if they've been used less than 10 times it doesn't seem like they're that popular a topic anyway.

So I would say that 93% of the tags are pretty much only good for entertainment value.

That isn't to say that the tagging system isn't working. It's just that it would probably work just as well as it does right now with 93% fewer tags.
posted by tkolar at 3:35 PM on January 24, 2007


Tags used hundreds of times want to be sub-categories, but we're stuck in single level category heck, and constrained to the Terriffic Twenty when it comes to categories, period. Too bad, as a folksonomy that regularly informed a heirarchally arranged taxonomy (even one so narrow as categories) would be a powerful mechanism for a community site. But this isn't a Drupal ad, or a pony request. I for one, dote on wireless networking questions, and the precious pups who bring 'em in!

Tags used less than 10 times are the nearly unique keys to arcane wisdom that could help us get quickly to the One (or Ten) Truest Answer, if they weren't so damn whimsical and generally hard to guess. Infrequently used tags in AskMe are like vanity license plates, in terms of their utility to folks other than the Asker. They're also clear evidence that people don't use tags to improve the utility of the site 93% of the time, despite the attempts to brand AskMe as a "community." Unless there is a feed back system that rewards people for applying tags effectively, making up new ones haphazardly is acceptable, and perhaps even encouraged behavior.
posted by paulsc at 4:28 PM on January 24, 2007


paulsc, I understand your concern that you cannot click on one tag and have a question (which only you know) answered for you. I still think you're overlooking what BP points out so eloquently above-the power is in the combinations. If you follow tags for many many subjects you can progressively narrow the questions asked and answered. You may not find your question because, among many other possiblities, your question may not have been asked. But if you care about a subject tags can help you to narrow and refine your question, which is an important part of arriving at an answer.

You seem to think that there's a kind of morality attached to tagging that I just don't see. Vanity tags are not even noise, they're non-existent to those searching tags. If you can't guess it, it just doesn't exist; it's a non-issue. If the only tag is a vanity tag, then there's a problem finding that question through the tag search. But most questions I searched under the vanity tags on tkolar's list had other tags associated. In this sense I disagree with tkolar's comment above, as well. If there are other tags associated with a question then it becomes much more findable through tags. The tag search will lead you to it. It won't be one click, but it's still within a general set of ideas.

But, really, what's the harm. There's search, there are tags. There could also be a set of subjects and sub-headings, categories, but I think you'd find similar troubles (although not the same troubles) with user-generated categories. Unless you have mods doing the categorizing you run into the edge cases, which are precisely the ones that seem to be of concern with faulty tagging, too. Is an ipod question Hardware, or is it Music, or is it Apple, or is it MP3, or which of those pre-formed categories does it fit into? If user A asks a question and puts it into every category except Apple, but user B has an ipod question and only looks in category Apple, the question is still missed. Just as with user-generated tags. Or perhaps I'm missing something about the concept of categories?
posted by OmieWise at 5:05 PM on January 24, 2007


So I would say that 93% of the tags are pretty much only good for entertainment value.

Banjo gets 11. Wacom gets 11. LLC 8. That's right around your utility guesstimate, and all of those have been genuinely helpful to me. I can recall anecdotally getting utility out of other tags in the 4-8 thread range, though I can't recall which tags they were now.
posted by cortex at 5:17 PM on January 24, 2007


"... But, really, what's the harm. There's search, there are tags. ..."
posted by OmieWise at 8:05 PM EST on January 24

Not for nothing are you widely linked, OmieWise. And this convinces me entirely.

I used to hate tags, and never used 'em. But between BP and you, the veil is pulled from my eyes.

Now, I love tags, and never use 'em! And I think we can all feel good about that.
posted by paulsc at 5:37 PM on January 24, 2007


Yes, but have we arrived at the brilliant, never thought of before, one-of-a-kind solution to the scaling problem for AskMe, yet?

Not that it matters much if it actually gets implemented; I find system scaling an interesting problem, particular when humans are involved.
posted by tkolar at 6:48 PM on January 24, 2007


paulsc writes "I used to hate tags, and never used 'em. But between BP and you, the veil is pulled from my eyes.

"Now, I
love tags, and never use 'em!"

Hah! This made me actually laugh out loud because I'm in the middle of reading Catch-22, which is filled with this kind of double negative stuff. (I'm not even sure what to call it.)

I was asking a serious question about categories, though, which seem to be you solution. How would they be different?
posted by OmieWise at 5:18 AM on January 25, 2007


« Older Taking money from the bad guys   |   Boston Meetup, try again? Newer »

You are not logged in, either login or create an account to post comments