« Information Card Foundation - second attempt | Main | Doin' the Museum Mash »

June 26, 2008

What makes a good tag?

Yonks ago (that's... like.. you know... quite a long time ago) I suggested to the web-support@jiscmail.ac.uk mailing list that we needed an agreed set of tags for labelling UK universities and colleges in Web 2.0 tagging services.  I'd raised the issue because we had just agreed with John Kirriemuir that he would create a Flickr pool in order to collect images of UK HE and FE activity in Second Life, as part of the series of snapshots that we are funding, and we wanted a way that people could consistently tag images according to which institution they represented.

I don't recall the details of what I suggested at the time but I think it was to use tags of the form 'universityofbath' (based on the list of names used by the HERO service).  Whatever... the specifics aren't important.  What happened was that I got deluged by replies offering different and conflicting advice about what makes a good set of tags - from totally unique but not very memorable UCAS codes, thru DNS domain names (bath.ac.uk), to human-readable but ultimately rather long strings such as the form I'd originally suggested (with or without hyphens and/or using camel-case).

Some useful points came out of the discussion, like the fact that unique but incomprehensible tags based on codes of one kind or another aren't very useful because no-one would ever 1) think of searching for them, or 2) remember them.  Unfortunately, nothing approaching consensus was reached.

We had a brief but rather similar exchange on Twitter yesterday because Brian Kelly suggested that the JISC Emerge project had got their tag strategy wrong by using 'em0608' for their current online conference, largely (I think) on the basis that Americans might get confused as to whether it meant June 2008 or August 2006. I responded along the lines of, "who cares, a tag isn't meant to be parsed anyway", to which Brian, rightly, responded that parsability and memorability are intertwined.

To cut a long(ish) story short, two things have emerged (excuse the pun) from this exchange:

  • firstly, having a conversation in bursts of 140 characters isn't ideal - and is probably annoying for those people not interested in the discussion in the first place, and
  • secondly, there is still little consensus about what makes a good tag!

I suggested that tags (particularly in the context of Twitter) need to be relatively short, relatively unique and relatively memorable.  But as Brian noted, there is a significant tension between shortness and memorableness (is that a word?).  Further, Steven Warburton questioned the value of uniqueness in the context of a relatively short-term forum like Twitter (i.e. it probably doesn't matter too much if your tag gets re-used a year later because the Twitter context has moved on).  However, it's important to remember that tags get shared across all kinds of Web 2.0 services (del.icio.us, Flickr, blogs, YouTube, Slideshare and so on) in order that applications like Hashtags and Onetag can pull everything together and that persistence requirements in those other services may be very different than they are in Twitter.

David Harrison asked a practical question concerning an upcoming UCISA conference - what did we think of 'ucisa-usc2008' as a tag? (Though it subsequently turns out that he meant 'ucisa-usc08'.)

I said that I thought it was too long - 14 characters (you need to prefix the tag with a '#' in Twitter) is 10% of the available bandwidth in a Twitter tweet.  I think that's too wasteful.  I suggested dropping the hyphens and using something like 'ucisausc08' or 'uusc2008' as alternatives but Brian commented that the hyphens were important to improve the tag's 'recitability'.

I'm not totally convinced... though I concede that our use of 'efsym2008' for our symposium earlier this year may have had less impact than it might because people didn't find it easy to remember (either because they didn't know what the 'ef' and 'sym' bits meant or because they got confused about whether it was '2008' or '08').

Ho hum... as I say, and this is basically the whole point of this rather long-winded post, there doesn't seem to be much in the way of agreed best practices around what makes a good tag.  And perhaps that's right and proper - we are talking about user-generated content after all and, in the case of the tags for universities, folksonomies are supposed to grow organically rather than be prescribed (though this isn't true for meeting tags which necessarily have to be prescribed by the organisers in advance of the meeting).

FWIW (which probably isn't much given the apparent level of disagreement) my current feeling is that brevity trumps clarity (at least assuming a desire to use the tags in Twitter), which means that 2-digit years are better than 4, hyphens are usually superfluous, and other strings should be kept as short as possible - but, as always, I reserve the right to change my mind at any point in the future.

TrackBack

TrackBack URL for this entry:
http://www.typepad.com/services/trackback/6a00d8345203ba69e200e55372c1f78833

Listed below are links to weblogs that reference What makes a good tag?:

Comments

Hi Andy - Thanks for this post. My view is that we are placing too great an emphasis on one particular application (Twitter). The majority of users who will either create or make use of tags will be in the desktop PC world (e.g. when uploading photos to Flickr), where the 140 character limit imposed by Twitter isn't an issues.

And, in addition, the tag completion implemented (using AJAX) means that users may not need to type the full tag on subsequent occasions.

So I would probably say the opposite to most of your comments in your final paragraph - withe the exception of reserving the right to change my mind! And I should add that I had a private IM chat with Mike Ellis during yesterday's Twitter discussion, with Mike pointing out that users of mobile devices would agree on the merits of brevity.

I should have added that one idea Mike and I discussed was the possible need for a tinytag service (along the lines of tinyurl) so that we can make use of both long and tiny tags.

In addition, how would you tag a JISC project with a (very) common acronym? Use the common acronym and include 'JISC' as another tag? Or 'JISCproject'? Or 'JISCproject-'?

As long as issues are decided in advance, I think as simple as possible: Andy, 'edsym08' would have been my logical offering for the event.

@Brian Yes, there's a focus on Twitter - but as an exemplar of microblogging, not as a specific in itself. So the kind of requirements we see in Twitter will remain, even if Twitter dies (and I'm hoping that the recent investment by the likes of Jeff Bezos makes that less likely anyway).

@Neil Without seeing an example it is hard to comment, but to take the specific example of the Emerge project conference - creating a tag like 'em0608' seems to me to have got it just about right in terms of brevity vs. meaningfulness - so, if prefixing with 'jisc' helps with meaningfulness and uniqueness without compromising too much on length it is probably OK.

Re: 'edsym08' - I tend to agree. 2-digit year rather than 4 = good. 'ed' more understandable than 'ef' = good. Ah, well - there's always next year... look out for #edsym09 !!

As I've said, the problem with two digits for the year is it can be confused with the conference count. Is edsym08 the 8th edsym conference, or the edsym conference in 2008? (Plus which, the tags will all collide in 3008.)

http://scilib.typepad.com/science_library_pad/2006/07/conference_tag_.html

I don't think the original post was especially 'long winded' - there was a point to it and (like tags) there never seems to be enough words to describe the issue.

Just sigining up for TypeKey (so I could respond properly) - made me bastardise my own name as no spaces are allowed.

That's the problem with Tags - how does one person think of the thing you have represented (on Flickr, on Jaiku, on Twitter, on Blogger on ....). I've described this in workshops before by asking delegates to tag an image from my Flickr acccount. There is never 100% concensus - even from me. I have posted two views of the same (kissing) statue at St Pancras Station (stpancras; st_pancras; st-pancras etc) and on one I use London as a tag but not on the other. So if I can't be consistant - how can I expect viewer to be so?

I agree that the debate should be widened to rraise awareness of the issue. Thanks for the post.

David

ah, the faithful librarian in all of us trying to come up with static application profiles. Have we not learned that there is no controlling the mass, or rather there is only pattern recognition witin the mass that allows us to determine the next layer in the stack that will provide the solution. For example in regard to DTF should we not be relying on te web to give us that data (http header DTF + tag). We should not try and cram all te world into a tweet, but utilise the cloud that the tweet and tag sit within. Also, with JISC projects, the world is just easier if you pick a project name that is a unique tag. Stop being clever with acronyms and go for unique. This post has parellels with the Lisp vs. C debate / MIT vs Newark layers of imperfection. on mob phone posting so can't ite (but that is what the cloud is for!)

The comments to this entry are closed.

About

Search

Loading
eFoundations is powered by TypePad