Tags Suck

I feel a little schadenfreude about the potential demise of Delicious. Not for popularizing social bookmarking, I like that a lot. What I resent them for is popularizing tags. Tags since infest services like Flickr, many blogs and Mendeley. If you listen to their advocates, you will be told that tags are one of the cornerstones of the Web 2.0, and even more so of the coming Web 3.0. To me, it seems that tags are an essentially useless technology that has somehow managed to become a remarkably long-living fad.

Sure, you can tag things (webpages, videos, books, research papers…) with keywords descriptive of their content. That is a 20th century librarian’s approach to making sure stuff can be easily retrieved. We live in the age of fulltext search, machine translation and the AwesomeBar (I myself stopped using Delicious the very moment Firefox 3 came out), at the dawn of semantic search and object-class detection in images. A Web service that relies on users sitting down and tagging stuff must be failing to use and further develop this technological potential.

So as long as you just want to retrieve stuff from a big bag of documents, search is your friend and tagging is slave labor. But what, the song of praise for tags goes on, if you want to order things, to structure them, to gain a deeper understanding of their relations and to improve your overview over the knowledge they represent? For this purpose, I hear, 20th century people used folders and put everything into a tree structure. The 21st century digital native social media Übermensch, on the other hand, uses tags. The obvious reason is that folders are so one-dimensional and rigid and you can classify things by only one dimension (or you have to first classify them by what to classify them by). Ordering schemes for pre-enlightened world views, those trees! The earth is a disc and built around heavenly Jerusalem, and everything has its one place and purpose, that’s the kind of insight that fits in folders, isn’t it?

I beg to differ. The “ordering schemes” you get by tagging stuff are aptly depicted by one of the most popular visualizations for tags: a cloud. A vapor. Nothing the human mind can really make sense of. Nothing that represents relations or insights. The only structured views on this cloud are filters, showing all items bearing a certain tag or combination of tags. Lists which are in no visible relation to each other. Venn bubbles bubbling around in the cloud. An excessively primitive database that does little more than a search engine but involves a hell of a lot more tedious human work. Concerning the collective knowledge aspect of this work, here’s what: in the time it would take you to tag a collection of bookmarks concerning a topic, rather write a blog article or a new paragraph for a Wikipedia article and let Google’s algorithms do the keyword extraction part for you. I’m sure you’re doing humanity a greater favor this way.

What you get by carefully organizing things into a tree structure, on the other hand, really is an ordering scheme. In the process, you weigh dimensions, you pick one as the one to order by, and discard other dimensions that are less useful for organizing your knowledge, honoring the fact that unlearning is an important part of learning. You think hard about things in order to find the best ordering scheme. Often you will find that there is no single dimension by which all things can usefully be ordered, and you will split them up into collections in each of which a different dimension is more relevant. These splits will likely be driven by your needs for your current project, and your information bookkeeping becomes goal-oriented. And isn’t this what dealing with the stuff and organizing it is all about: learning something from it so you can do something with it? And with the way the human mind works, I claim boldly, you haven’t really “got” something (information doesn’t turn into knowledge) until you have succeeded in putting it into a tree shape. Folders rock.

5 thoughts on “Tags Suck

  1. Arnymenos

    [Dieser Beitrag wurde als lesenswert getaggt.*]

    Einspruch, Euer Ehren: Worauf trainiere man seine Extraktionsalgorithmen, wenn nicht auf Tagging-Sklavenarbeit? Oder willst du auf regelbasierte Systeme hinaus? Dann: Worauf evaluierst du sie?

    Das großartige, volltextdurchsuchte Web 2.0 aufzubauen erforderte wie jede große Zivilisation unmenschliche Arbeitsbedingungen. Vielleicht gilt das im Web 3.0 dann tatsächlich als Schinderei; als Kopfrechnen, wo man Taschenrechnen könnte. Und Taggen ist dann „so 2.0“. Bis dahin vergehen aber noch viele Tags.

    *) Und zwar von einem heuristischen Algorithmus, der 26 (!!) verschiedene Inputmerkmale auswertet.

  2. Eugê

    I think you have some valid points, but I think that you see tags too one-dimensional. Look at your blog for example, even this very article. What I read at the bottom is Dieser Beitrag wurde unter English, Geistesleben, Internet veröffentlicht.

    The article is tagged with different categories. And — even though the categories themselves are structured — contrary to the last paragraph you don’t have to choose one category. You will find this article in the categories Sprachen, Welt and Internet and not only in Geistesleben.

    Thus, you can see categories as structured tags and I think that they are quite handy. Maybe you can see that tags are not only an infestation if used correctly and in many contexts they can lead to an information gain.

  3. ke Beitragsautor

    Arnymenos, du meinst also, es wird an Suchsystemen gearbeitet, die sich auf die von Benutzern vergebenen Tags stützen? Kann das wirklich sehr viel besser sein als das, was Googles Linkanalyse schon macht und schon im Web 1.0 machte? Ich bezweifle es.

    Aber wenn doch, dann bleibt meine Kritik daran bestehen, dass seitens der Tag-Evangelisten der Eindruck erweckt wird, Tags wären für den einzelnen Benutzer eine wahnsinnig nützliche Möglichkeit, seine Bookmarks, Papers usw. zu organisieren. Dann sehe ich in emsigen Taggern unbezahlte Annotatoren, die unter Vorspiegelung von mir für falsch gehaltener Tatsachen zu – immerhin – gemeinnütziger Arbeit gebracht wurden.

  4. ke Beitragsautor

    Eugê, I agree, there are jobs for which tags are the right tool. I just don’t think building collective knowledge resources is one of these jobs, and for structuring data, tags can only be a first step. In short, tags are not what they are made out to be by some (see the enthusiastic articles my post links to). If it wasn’t for this hype, I think we wouldn’t even be using the word “tags” and just talk about “categories”, something very basic and hardly remarkable. By “essentially useless technology”, of course I don’t mean the concept of categorizing, but its less than user-friendly implementation and use in many Web 2.0 applications.

    Thanks for your comment, it gave me the opportunity to add a new blog to my feedreader and tag it file it under “Computerlinguistenblogs”. :D

  5. ke Beitragsautor

    An article similar in spirit to this one: http://al3x.net/2009/01/31/against-everything-buckets.html

    I just think he’s a little hard on search. To reiterate my point, search is great if implemented well. If it doesn’t suffice for a purpose, then some half-ass pseudo-structuring using tags or an “Everything Buckets” app probably won’t help much either. Then you need real structuring.

Schreibe einen Kommentar

Deine E-Mail-Adresse wird nicht veröffentlicht. Erforderliche Felder sind mit * markiert.