Taxonomy of Books

Newsgroups: talk.origins
Subject: Book Taxonomy II (long)

  Due to thunderous demand (one email), I'll outline the essence of 
my attempt at a better taxonomy for books than either the LC or the
Dewey cataloging systems.  The relevance to normal t.o. content is 
minimal.  The only bearing being that which I noted already about
the difficulty of book taxonomy versus biological taxonomy being a
sign of the former being applied to a created system and the latter
applied to an evolved system.


  Any taxonomy must meet some standard of 'goodness'.  In biology, 
one would like to declare genus's which are collections of 'similar'
species.  Further, the species within a genus should be more like
each other than they are like species in any other genus.  Repeat
this in your route up the taxonomic hierarchy.  The classification
of a species as to its genus, family, ... should also be repeatable
by independant observers, and disagreements should be decreasingly
common as we move up the taxonomic line.

  For book classification, my measures of 'goodness' are governed by
the fact that I'm a browser.  That is, I like to go to the stacks
looking for one thing, and find related, interesting books, near to
the book I'm looking for.  Repeatability is also important, in that
both as a browser and in thinking of the catalogers, I want to be
able to predict in which part of the library (_without_ reference
to a computer search or a several hundred page manual) a given 
subject will be in.

  A notion which influenced me was the classification that X (who was
it?) applied to his US geography books.  He arranged them on bookshelves
according to where their state was.  Hence Illinois came out near the
middle; Maine was at the top right, etc.  The key is that the ordering
of the books was not one dimensional.

  Two important notions so far: 'nearness' in coding should be equatable
to 'nearness' in content, and that the measure and coding should be 
multidimensional.

  The principal axes in the coordinate system should be things which are:
independant (to a large extent, to preserve the 'dimensionality'), 
continuous (to preserve the 'nearness' property), and fairly easily 
identifiable (reproducibility).  The first three identifiers I thought of
were: level of the book, subject matter studied/discussed/described, 
and method of study.  

  Level should be a descriptor of how much background a person needs
to understand the book, ranged from 0 to 9.  9 would be professional
journals, while 0 would be an 'intelligent layman's' introduction.  This
partly finesses the matter of whether you file journals separately from
books.  Most books would be classed lower than a 9 (they have the
explanatory material that journals leave out), but some (conference 
proceedings, for example) would be in there with journals.

  Subject matter: I arranged from the abstract, inorganic, to the organic,
to human activity.  1 would be mathematics, 2 physics, 3 inorganic chemistry,
... to B - politics  C - religion D - philosophy.  (This wound up with
about 25 divisions, and what is listed above is only suggestive of the
one I arrived at before.)  This is the part that both Dewey and LC use
as their only major divider.  Dewey made some attempt to keep a continuum
in his classification, LC pretty much ignored it. 

  Method of study or use.  Oceanography as a subject of study was split
between several libraries on my campus.  The division was (loosely) based
on the method of study.  Mathematical oceanography in the math library,
chemical oceanography in the chemistry library, observational oceanography 
in the main library (where geography was kept), and the residue in the
science library.  In reflecting on the books, I decided that method of
study was as much an identifier as the subject being studied.  Mathematical
physics has much in common with mathematical ecology, for instance.
Again, I arranged them in a continuum from the mathematical to, in this
case, I think historic: 1 Mathematical ... B historical.

  Arranged as level, method, subject, it should be relatively easy to 
code 'advanced philosophy of science', 'introductory mathematical physics', 
'intermediate physical oceanography', and the like.  And one can have 
'history of history', or 'mathematical study of mathematics' (this 
latter would be more typically called foundations of mathematics, I think).  

  After that, I added scale (length, number of elements/people, area, volume,
... as appropriate to the major index), and location (4 codes for space 
and time).  These are less generally applicable, so there might be some
specialization of the 5 reserved spaces for some of the major codes.
Something like this does preserve the scheme for US geography books
mentioned above.  One can quite easily deduce the code for historical 
geography (major code) of the eastern US (location and scale) 
during the 1800's (time).  Or, for that matter, Velikovskian catastrophism
(mythological study of the solar system over the last few thousand years).

  So 8 codes for the book, plus however many (3-5 in LC typically) are
needed for the author.  So 0A1.13456.A31 would be something like Asimov's
book on the history of mathematics in the middle east during 700-1200.
(I don't think there is such a book, but you get my drift.) I don't wind 
up with a shorter coding that LC, and typically am longer than Dewey, but 
that wasn't my complaint about either system.  My system does allow for
new fields (mathematical history), which would also avoid some of the
uglier problems I've had with LC.