Re: pre-coordination

Kelley McGrath wrote a nice article on LCSH and faceted browsing for the Code4Lib Journal in 2007:
"Facet-Based Search and Navigation With LCSH: Problems and Opportunities"
http://journal.code4lib.org/articles/23

The "lattice" works nicely in Simon's example below, but that's not always how LCSH is written--sometimes it's not possible to decompose just at the dashes:
"these demands for step-by-step and clearly defined division are often in conflict with the desire to have phrases that make sense to users (e.g., “African American women poets” versus “African Americans” + “Women” + “Poets”). "

Lots of other gems in what Kelley wrote, along with brief descriptions of alternatives such as FAST [1] and RSWK. FAST might be particularly interesting for Freebase, Thad.

-Jodi
 [1] http://www.oclc.org/research/activities/fast/default.htm

On 4 Nov 2010, at 00:55, Simon Spero wrote:

> On Tue, Nov 2, 2010 at 2:46 PM, Ray Denenberg, Library of Congress <rden@loc.gov> wrote:
> Quote:  'wouldn't the result be loads of syntactically "valid" but nonsensical combinations?'
> 
> The more difficult question is; what meaning, if any,  do pre-combined* LCSH heading strings have.   If Drabenstott (1998) is correct, then the interpretation of the order of subdivisions, as applied by catalogers, and as understood by patrons and reference librarians, does not match the official interpretation.
> 
> Under the official interpretation, a sentential paraphrase is generated by treating the compound as being right headed, and reading off the modifiers from  right to left - for example "Weasels -- United States -- Nineteenth Century -- History" roughly corresponds to  "History of Nineteenth Century US Weasels". 
> 
> Even without the problems of user and cataloger understanding, any account of the semantics of subdivided headings must explain the entailed relationships between the pre-combined heading and its components.   The heading used as an example in the previous post bears closer examination.
> 
> [Note that subject headings refer to what documents are about - see e.g. Svenonius (2000, p.130) ]
> 
>  (1) Sailboats -- Design and construction -- New Zealand
> 
> has the canonical form.
> 
> (1a) Sailboats -- New Zealand -- Design and construction
>  
> Since the subdivisions are orthogonal, both headings denote the set of documents about "The design and construction of Sailboats in New Zealand".
> 
> Since these documents are about the design and construction of (Sailboats) in  (New Zealand), they must necessarily also be about  the the design and construction of something, somewhere. 
> 
> (2) Design and construction. 
> 
> Since everything about (1) must  also about (2), we can infer (from the definition of BT )
> 
> (3) Sailboats -- New Zealand -- Design and construction  BT  Design and construction
> 
> Since these documents are about (Sailboats) located in New Zealand  (in particular, their design and construction), they must necessarily in some way also be about 
> 
> (4) New Zealand
> 
> Which entails
> 
> (5) Sailboats -- New Zealand -- Design and construction  BT  New Zealand
> 
> Being about the activity design and construction of (Sailboats) in New Zealand, the documents must be about the activity of  design and construction of (something) in New Zealand:
> 
> (6) New Zealand -- Design and construction
> 
> Entailing
> 
> (7) Sailboats -- New Zealand -- Design and construction  BT New Zealand -- Design and construction  
> 
> Note also that
> 
> (8) New Zealand -- Design and construction BT New Zealand
> and 
> (9) New Zealand -- Design and construction BT Design and construction
> 
> We also have 
> 
> (10 ) Sailboats -- New Zealand -- Design and construction   BT Sailboats -- Design and construction
> 
> (11) Sailboats -- Design and construction BT Sailboats
> (12) Sailboats -- Design and construction BT Design and construction
> etc.
> 
> These BT relationships form a semi-lattice whose GLB is the most specific subject.
> 
> Note that  the ontological entities  described by documents about "Sailboats" and "Sailboats -- New Zealand"  are [Sailboats] but documents about "Sailboats -- New Zealand -- Design and construction"  describe a [Process].
> 
> The geographic subdivision does not change the class of entity, but the topical subdivision does. 
> 
> Form/Genre subdivisions confuse matters further, since they refer to the ontological class of the document itself. Of course, without subdivison markers ($v or $x), it's impossible to tell from the string form whether a subdivision that might be a Form/Genre is serving that role, rather than being a topical modifier. 
> 
> Simon 
> 
> 
> Drabenstott, Karen Markey (1998). Understanding subject headings in library catalogs. 
> Tech. rep. University of Michigan. url: http://deepblue.lib.umich.edu/handle/2027.42/57992.
> 
> Svenonius, Elaine (2000). The Intellectual Foundation of Information Organization. Cambridge, Mass.: MIT Press. url: http://www.netlibrary.com/AccessProduct.aspx?ProductId=39954.
>