The Persistent of Tags

About the persistence over time of tags, from an academic perspective.

The Skeptical Librarian: 

One of the problems with tags and folksonomy is that they will not preserve a terribly useful legacy for access.  Terms come and go in usage faster than ever. To rely only on tag clouds will mean information will be harder to find in any kind of precise way in 10 years.

Tagger:

For an individual, the terms are whatever you tag them as and the terms will change for you as you grow, but they do reach a pretty quickly diminishing limit (as Marlow, Naaman, et al have found) as a tagger reaches a mature level in tagging practice.

Of course, between any two random taggers, their tag sets may vary markedly, but overall, an object reaches a stable proportion of tag representation after only about 100 markings (Golder and Huberman 2006), so it doesn't take many users tagging to get a very useful and stable core tag set.

So I don't see tag distribution change as a negative but a huge positive: objects get additionally tagged as the audience conceptualizations of them change, so they will remain pertinent over time, making objects EASIER to find.

But they will also have historical depth, unlike say, the Dewey decimal or Library of Congress time fixed taxonomies, which have historical stasis (rather than depth), and hence pertinence drift.

However, there is discussion about "tag decay," the obsolescence of tags. But that is simply a property of the type of data it is and should be handled with standard data management practices.

Pruning old or irrelevant tags is (well, should be/will be) a standard data management practice. Golder and Huberman's findings about the stability of the central core of tags is very important for our situation -- cleaning out one, a hundred, or even a thousand graduating students' personal tags will not adversely impact the central tag core for an item.

For example: As a student, I would tag a library journal article with tags for the class, term and Professor I was doing the research for. If the article was used in that class by that Prof year after year, more students might take this way and so the tags will stay. If not, then the tags will diminish as the old student data is cleaned out and disappear as they should.

Finally, I don't believe anyone is seriously advocating the replacement of standard controlled vocabularies by tagging and folksonomy.  In fact, the argument that tagging will replace standard taxonomic metadata is a straw man in such arguments. 

Rather, the purpose of tagging and the reason for its incredibly swift adoption and growth  is that it is  filling the big niche between controlled conceptual taxonomies and brute force lexical search. All three are complementary tools.

Sources:

Marlow, Naaman, et al. 2006
http://www.danah.org/papers/Hypertext2006.pdf

Golder and Huberman 2006
http://jis.sagepub.com/cgi/content/abstract/32/2/198

Comments

[joelamantia]

Another strong analysis - placing tagging on the larger continuum of information retrieval / management solution types - nice work!