Would the real source of metadata please stand up?

In my local library last week, I was surprised to find Robert Pirsig's

Zen And The Art of Motorcycle Maintenance filed under Travel. Whenever I

encounter interesting classification events such as this, I find myself

asking the question 'what was the shelf-stuffer thinking?'. Here are

some possibilities I see.

1) The shelf-stuffer may have had no knowledge of the book, read the

back cover, encountered words like 'motorcycle' and 'road trip' and

concluded it was a travelogue. Fair enough.

2) The shelf-stuffer may have known a lot about the book and concluded

that, although the main journey in the book is an abstract one in search

of a meaning for the term 'Quality', perhaps 'Travel' is as good a shelf

as any. Fair enough.

3) The shelf-stuffer may have known a lot about the book and decided to

inject some humor or surprise into someone's life by planting the book

on the 'Travel' shelf. Fair enough.

My favorite example of shelf-stuffer type 1 came many years ago in a

Dublin bookshop where I watched a person file Douglas Hofstadter's

classic Godel, Escher, Bach under 'Religion'.

My favorite example of shelf-stuff type 3 came from the same bookstore

where they filed Dublin's Bus and Train timetables under 'General

fiction'. The Bus and Train company was not amused.

The physical world puts constraints on content classification that do

not occur in the electronic world. In the real world, we need to pick a

shelf for a book because both shelves and books are physical things.

There is no way that the same copy of a book can be on multiple shelves

at the same time.

In the electronic world, this constraint disappears. On the Web for

example, things have names (URLs) and these names have a built-in

retrieval mechanism i.e. you can click on them. Consequently, it is the

most natural thing in the world to create multiple pages (shelves!) that

contains the same book by referencing it (or indeed transcluding[1] it)

via a URL.

On the Web, we would think nothing of a virtual library that listed

Pirsig's book under multiple headings - 'Travel','Zen' and 'Philosophy'

perhaps. On the Web, we think nothing of being able to locate Pirsig's

book by trying various likely words or phrases in a search engine.

The ease with which we have taken to this virtualization of content

classification and effortless text search tells us a lot about how we,

as humans, have chosen to resolve the shelf-stuffers dilemma. Instead of

spending a lot of time (and money) deciding on a perfect classification

for a piece of content, we prefer not to bother with classification too

much when we create content. Instead, we use external, ex post facto

methods - most notably, pages of links and search engines - to allow us

to retrieve content.

From a theory of information classification perspective, this is a bit

troubling perhaps. Proponents of metadata would argue that information

can and should describe itself. Authors can and should take the time to

classify the content they produce, preferably at the time they produce

it, for ease of retrieval later on.

Although I have sympathies with this view, the evidence would suggest

that metadata-in-content strategies end up being more honored in the

breach than in the observance. As a social phenomenon - embedded

metadata as classically formulated - has been, and continues to be a


On the other hand, search engines like Google have shown what a powerful

source of metadata the link structure of the Web can be. As a social

phenomenon, Google (and of course The web of links on which it depends)

has been a runaway success.

Such non-classical approaches to classification metadata look set to

expand greatly over the next few years. Perhaps the biggest buzz at the

moment surrounds the possibility of using relationships between

like-minded bloggers as a sort of 'collective intelligence'[2] to tune

search results based on results that their peers think are relevant.

One might ask - is this sort of thing really metadata? Some would say

'no'. I would say 'yes'. It is metadata all right, metadata of the most

potent kind, organically grown and self modifying based on how

information is actually used in the real world.

It is metadata, but it has not been filed where we expected to find it.

The irony of that, really appeals to me.

[1] http://www.usemod.com/cgi-bin/mb.pl?TransClusion

[2] http://c2.com/cgi/wiki?CollectiveIntelligence

ITWorld DealPost: The best in tech deals and discounts.
Shop Tech Products at Amazon