tags: an autocomplete tool

Ok, it's been a while since I've talked about tags. Yes, I messed around with a tag tool for a while, and that went nowhere. But it did point me in the right direction.

I got a tag dump from ct (thanks, ct), and processed that into a db which is queried by an Ajax autocomplete routine. In plain English, this means I've made a page where you can start typing in a desired tag, and after you've entered three characters the database of all tags is queried, and you're given a list of matches from which to choose. You can then select from those matches. So if you can't ever remember how Condi spells here first name, this'll do it for you. Sorta.

[more...]

Once you've picked the right tag, you can continue to add tags -- the autocomplete sticks in a comma and then works on the next entry. Right now you can then select this string and paste it into a diary tag entry box, but eventually it should make it into the regular diary entry page. Remember, while right now it's a bit more than a proof of concept, it's still not really ready for prime time.

You'll notice that some entries are bold, and some are not. The idea is that the bold entries are the "correct" ones to use, but I'm not real happy with my algorithm for choosing "valid' tags at the moment, so take these with a grain of salt. (Briefly, I've chosen a set of "tag librarians" (those who have contributed most to the dKosopedia tag project pages) and set any tag  associated with their user id (50017 4242 49281 33535 17760 71723 4703 52910 38103 3541 93685 15940) to be a valid tag. It's not perfect, but it's a start. My next big task is to provide a form which tag librarians can use to manage the tag db.)

Anyway, please play with it and let me know what you think. It's right here.

(Yes, I've attached the "meta" tag to this diary, but don't get all excited. This isn't the fun type of meta, wherein we get to simultaneously wank, navel gaze, and insult friend and foe alike, just for the hell of it. No, this is the boring type of meta. This actually has to do with the mechanics of the site. Sorry.)

•  you mean the best kind of meta

  by buddabelly on Thu Feb 15, 2007 at 04:26:40 PM PDT

•  Tried it -- super!

  I expected a slow response, but it brings up the tags right away.  So much better than that dreadful tag cloud.

  by A Mad Mad World on Thu Feb 15, 2007 at 04:30:37 PM PDT

•  I think it is cool, but...

  I like it, but do we really want all of the misspelling s of George Bush in there? :)

  Thanks for reading,

  Steven Joseph :)

  by StevenJoseph on Thu Feb 15, 2007 at 04:30:58 PM PDT

  •  Yes, a great tool

    for the tag police

    by A Mad Mad World on Thu Feb 15, 2007 at 04:32:17 PM PDT

    •  Looks like it'd put the police out of a job!

      by LibertyGuard on Thu Feb 15, 2007 at 04:36:46 PM PDT

  •  like I said

    this is little more than proof of concept. I'm returning all tags that match because I'm not confident in the filtering yet. Give me time...

    by The Centerfielder on Thu Feb 15, 2007 at 04:50:26 PM PDT

•  Woo Hoo!

  This is cool.  I have been flogging:
  http://meta.dkosopedia.com/...
  when I change tags but I will add this to my toolkit and suggest it for the tag impaired!

  Thanks.

  by i dunno on Thu Feb 15, 2007 at 04:36:07 PM PDT

  •  Seconded

    This is a good toy.

    by monkeybiz on Thu Feb 15, 2007 at 05:34:25 PM PDT

•  I hereby dub thee....

                ♣ dKosuggest

  Pretty slick.

  by adios on Thu Feb 15, 2007 at 04:43:06 PM PDT

•  Pop-up version?

  I haven't tried this yet, but from your description, it sounds like a separate page.  I know that there's no good reason people should not be willing to open and then shut a separate page, but they often seem to hesitate to do so.  This strikes me that it would be really useful if yoked to the "enter tags" button to function as a popup window which people could click off when done with it.  Just a suggestion.  Good work!

  by Major Danby on Thu Feb 15, 2007 at 04:56:23 PM PDT

  •  Try It

    It's an inline thing, like Google Suggest.

    by Scipio on Thu Feb 15, 2007 at 05:14:01 PM PDT

•  very nice. I added it to my blogroll for easy

  access. I wanted to copy all the impeac* choices to show how ardent we all are about it, but you are too good at making the list a select one list :-)

  by Wee Mama on Thu Feb 15, 2007 at 05:04:46 PM PDT

  •  Great idea! n/t

    by littlesky on Thu Feb 15, 2007 at 05:21:54 PM PDT

•  Bug Report

  I know it's incomplete and all, but if you've got a scrolling bug.

  1. Type in a few letters (like "Geo)

  2. After results come up, start pressing the down key a  bunch

  This will have the yellow bar start to move down. When it reaches the bottom of the scrolling window, however, it keeps going down, but the window doesn't scroll to keep it in view. In effect, you can't scroll without using the mouse.

  Also - what do you think of the Yahoo! UI library? I use the events one a whole lot, but haven't tried their ones for visual chrome yet.

  by Scipio on Thu Feb 15, 2007 at 05:17:52 PM PDT

•  Pretty cool tool!

  I like this misspelling in the tags - Condoleeza Rick.  heh.

  Thanks!  Good meta!

  by littlesky on Thu Feb 15, 2007 at 05:19:01 PM PDT

•  I tried the following tags

  Iraq, WMD, Scott Ritter, and Baker Report

  and found a Lot of tags on Iraq, a number on WMD, Scott Ritter came up almost immediately, and Baker report gave several choices, including Baker-Hamilton Report.

    My feedback is I'm not sure, as a casual user of tags, when exactly I'm finished (is there an edit review?), and in what way my choices might influence the data-base.

    For instance, Iraq War in different time periods might be more useful, especially for any historical research. But the term "war" is itself a bit controversial, in that some call it an invasion, some imperialism, others occupation, etc. Not sure how you sort this out, other than making the choice of tags available in the data-base.

   The other thing is how does one use the information to improve accuracy in the site? For instance, there are conspiracy theories, which are nominally banned from the site, which will keep popping up--and thus it would seem that it would be a good preventive action to have some "tags" pre-loaded, to deflect known falsehoods. One of the more prominent is "holocaust deniers." Yet I have seen one poster try to make a valid point about the Armenian genocide, yet get lumped in with Holocaust deniers, by people who seemed only partially acquainted with the facts.
    An many posts would benefit by an acquaintance with Wikipedia. One hopes that at least the front-pagers have their posts run through a Wikipedia/fact-checking process, and thus some of the tags used in Wikipedia are probable candidates for inclusion in the DKos tag data-base. Some of the D-Kos community are pretty inventive with their word choices, and I'm pretty sure that there are some genuine neologisms coined here, and I would think that the tag database might be compared against the Oxford English Dictionary (O.E.D.) so D-Kos can claim bragging rights to the new words coined on the site. Attention to this detail alone would further secure D-Kos a more prominent position in the English Language.

  by jfdunphy on Thu Feb 15, 2007 at 07:34:13 PM PDT

•  Are you aware that there are no 'Steven's?

  Lots of Stephens, including some that I think are really spelled with the 'v'.

  Good start, especially with the bolding, though it'd be better, eventually to just exclude all of the one-offs from this database. The temptation is too strong to use them, simply because they appear. Ideally there'd be a map from the unbold to a bolded approved version. Heh.

  by Halcyon on Fri Feb 16, 2007 at 11:42:29 AM PDT

  •  hmmm

    yes, eliminating one-offs should be pretty easy. Good idea.

    And the mapping you mention is crucial to the design. Odd that you mention it, of course, because I was about to email you about just that thing...

    by The Centerfielder on Sat Feb 17, 2007 at 07:05:16 PM PDT

  •  one offs excluded; see my reply below

    by The Centerfielder on Sat Feb 17, 2007 at 09:16:52 PM PDT

•  Nice work!

  One thought, though -- would it be possible to sort the list by frequency, instead of alphabetically?

  by Abou Ben Adhem on Fri Feb 16, 2007 at 02:47:48 PM PDT

  •  another hmmm

    yes, this is possible too. Give me a day or two.

    by The Centerfielder on Sat Feb 17, 2007 at 07:06:03 PM PDT

    •  ok

      didn't take me as long as I thought it would. Results are now sorted descending by frequency, and the one-offs are excluded. Note that this data is static as of the tag dump I got on Feb 10th or so.

      by The Centerfielder on Sat Feb 17, 2007 at 09:16:08 PM PDT