Taxonomy Recapitulates Folksonomy

Back when I worked at the Globecon Group - a personality-driven boutique that trained wholesale bankers - one of my jobs was to come up with keywords for the thousands of books, articles and PowerPoints that comprised our library of training material. Some people didn't like the keywords I chose; they wanted new ones that made more sense to them. "Can't have that," I thought. I used the power of the sysadmin to guard those keywords like a paranoid dictator, leading to even more dysfunction in a company that was already pretty screwy.

That's why I was so interested to read Tagging vs. Cataloging - What It's All About by the lovely Chiara Fox, who looks like Winona Ryder without the anorexia. Chiara describes the difference between taxonomies - collections of keyword pick-lists like the Library of Congress Subject Headings - and folksonomies, such as the tags assigned by users of Flickr or

Librarians have traditionally lived in a controlled world filled with the ultimate of “keyword pick-lists”: the Library of Congress Subject Headings, the Medical Subject Headings (MESH) or even the Dewey Decimal System. A cataloger looks at an item to be cataloged, determines its “aboutness” and then finds the word (or words) from the list of accepted subject headings that best describes it.

These keyword pick-lists aren’t perfect, nor are they all-inclusive. But they do help control the terms that are used to describe items, which, as any cataloger worth her salt will tell you, is of the utmost importance. Without a way to account for the ambiguities of language and meaning, there is chaos. Terms and values must be controlled in order for the system to maintain a high level of recall and precision in results. This is something that every cataloger is taught in library school.

Tagging, on the other hand, is a bottom-up, wisdom-of-crowds approach with no standardization whatsoever (though the auto-suggest and type-ahead features of can elicit some order). And this lack of standardization is one of the biggest criticisms of tagging.

Take the classificiation problem I faced with Globecon's teaching material, for instance. Almost every piece of content in the system could easily be tagged "corporate finance" - whether the subject was bonds, valuation, yield calculations or even commercial lending. When all content has the same tag, the tag becomes useless. That's what I was trying to avoid by coming up with a controlled set of keywords.

In any case, I can't wait for Chiara's next essay, where she promises to tell us about "the dark side of tagging."