[Data-modeling] Library of Congress and Dewey Classifications

Jeff Prucher jeff at metaweb.com
Tue Mar 18 20:54:12 UTC 2008


 

Ed Laurent wrote:
> 
> My initial thought is to list the full codes as machine 
> readable strings with library disambiguation. These numbers 
> are essentially unique codes in unique but purposefully 
> similar classification systems. Because the topics (i.e., 
> books) that the codes refer to are themselves unique and have 
> multiple codes, some form of cross reference is needed to 
> link multiple codes from multiple classification systems to 
> the same topics. My Classification code 
> <http://www.freebase.com/view/schema/user/spatialed/default_do
> main/classification_code>  type and Motorcycle part number 
> types (both manufacturer 
> <http://www.freebase.com/view/schema/user/spatialed/motorcycle
> /motorcycle_part_number_manufacturer>  and retailer 
> <http://www.freebase.com/view/schema/user/spatialed/motorcycle
> /motorcycle_part_number_retailer> ) provide examples of how 
> I'm handling these kinds of situations. I don't think the 
> Dewey Decimal system is very different. It is unfortunate to 
> hear that the Library of Congress codes are not unique for 
> each publication and edition. I thought that was the point of 
> the L of C system!

After poking around a bit more, it does seem like the LoC data is less
completely weird than the Dewey data. What variation I've found (and from an
admittedly small sample) is that sometimes the entire code is different at
different libraries, but within that variation, the codes are stable at the
edition level.  E.g., for a 1976 edition of "The Wealth of Nations", one
library has "AC7 .S59 1976" and another has "HB161 .S65 1976", but I didn't
see any with something like "HB161 .S64". So maybe LC classification should
stay on the book edition.

The codes, regardless of how we do this, should definitely be
machine-readable strings, rather than text strings (as they are now).  I'm
not sure about modeling the entirety of Dewey/LoC classifications as topics
-- we could do it (modulo any copyright issues -- I haven't looked into
that), but I wonder which would be more useful -- a simple string or a
topic?

Jeff




More information about the Data-modeling mailing list