SIGIA-L Mail Archives: Re: understanding metadata , was Re: SIG
Re: understanding metadata , was Re: SIGIA-L: future directions for IA
From: Peter Merholz (peterme_at_peterme.com)
Date: Thu Aug 30 2001 - 18:42:30 EDT
James "Weizenheimer" wrote:
> > I would emphasize the word: principles. When this word is
> > emphasized, then yes, I think that they are a magic potion,
> > provided they are used correctly. Why do I think that? Because
> > centuries of experience have proven that they work.
Um. No it hasn't. No such proof exists. You can't make any such claim,
anyway, until you specify for whom such systems work. Because there's a lot
of evidence to suggest that they don't work for "info-civilians."
Don't foolishly assume that longevity = success or proof of "work"-ness.
There's a lot of evidence that suggests that maps by-and-large don't "work",
yet they've been around for centuries, too. Sure, they work for some, but I
wouldn't be comfortable with the claim that they "work."
Karl wrote:
> 4. OPACs have many positives, but are widely considered hard to use.
> This despite their rich data and having been part of libraries
> for a decade, often much more (ie: first mover advantage).
>
> 5. Amazon is based on library principles. BUT, they use publisher
> data, not AACR2; BISAC codes instead of LCSH; and (I suspect) has
> little in the way of authority control.
>
> 6. Google isn't based on any of this, but instead tries to extract
> inherent structure.
>
> 7. Anecdotally at least, Google and Amazon provide a better search
> experience than OPACs, despite being based on structured, but
> inferior, metadata (Amazon) or minimal and wildly varying
> metadata (Google).
>
>
> Why?
Because, largely, the claims James stakes are untrue. Perhaps such library
classification schemes worked for smaller bodies of information. Perhaps
it's simply an issue of scale, and the LIS field hasn't really been able to
come to terms with the utterly massive amounts of information out there.
Also, look at the tasks at hand. Amazon has an amazing constrained
task--shopping. The entire search experience is focused on products, which
is an extremely limited set, compared to the universe of things you could
search on at large.
Also, one could likely argue that, for a trained individual, Amazon is quite
likely rather impoverished when compared to OPACs.
Google works because it takes advantage of an *extremely* powerful
notion--socially situated intelligence. Libraries would be tons smarter,
too, if they captured and utilized usage data and reference citations in
it's indexing and presentation of the search experience.
This makes me think of somethng else... In my experience, library searches
gauge relevance solely on tagged metadata. If you search for "Stephen King",
you will receive a list of authors named "Stephen King", likely sorted in
some uninteresting way (date of birth, etc.). A search for "Stephen King" on
Amazon or Google will take you right where you (likely) want to go--to the
best-selling author--because both systems are infused with popularity
rankings. If you were interested in some other Stephen King, you'll also be
able to find him, it'll just be harder. That trade-off is worthwhile. Also,
I think the positive feedback loop of such systems (best-sellers are more
prominent, leading them to be sold more, keeping them as best-sellers) is
more than outweighed by the fact that most people want to know about what
most other people want to know about.
--peter
This archive was generated by hypermail 2.1.2
: Sun Nov 23 2003 - 22:54:48 EST
|
