Thursday, August 11

How to classify

Today I read an essay by Clay Shirky about the problems of classification that we face in nowadays digital world. I really liked that article! If there is a librarian living somewhere in your heart, please click on the link hidden in the topic for this entry and read his text, it is entertaining and informative both at the same time.

In essence, he tells that a hierarchical classification like the one that we find at DMOZ or even at Yahoo proves to be inferior to Google, and even more to what gives us. The main reason seems to be that most classification schemas are based on the assumption of a shelve and that a book can be in one place only, while it can be written about multiple topics at the same time. (It may even mean different things to different readers, I might add.)

Reading his text, one might come to the conclusion, that those classifications are to be avoided and bad things. I do not agree to this conclusion fully, though. A hierarchical system may not be perfect, but it still basically documents the contents of the information system. A user can, while browsing the system, see what kind of information she may expect to get out of it. A system as broad as Google, which covers probably every possible topic one can think of, may not benefit much of it. But as soon as the system is smaller and the information stored in it less diverse, it pays off to know what is in there and what is not.

A hierarchy easily allows to see to the boundaries of the information system. If the system knows only links, then even after hours of browsing through documents about, say ... cars, a user can never know for sure, if one of those other links somewhere points to a totally different world of non-car-topics. If the user has a hierarchy in front of her which says Vehicles at the top, then she knows right from the start, that it would not pay off to look for information about mp3 players or other non-vehicles.

In smaller setups it may also be impossible to benefit from the community effects that drive, forcing the system architect to look for other solutions. Like a classification.

In practice I guess, if one thinks in smaller terms than Google, a classification may still be helpful. The perfect information system probably provides the user with everything - a classification, fulltext search as well as user-provided labels.

So far my thoughts for tonight :)


PS: (By the way: I agree that implements a very innovative and interesting idea, but I still... I don't really use any bookmarks. I have Google, why do I need bookmarks :-) How about you? Are you actively using services?).


Post a Comment

<< Home