[SAC-FAST] broader questions

[Stephen Hearn]

The broader questions I have about FAST at this point are:  What are we 
evaluating? A subject descriptor system? A conversion algorithm? The 
quality of a set of catalog records? Hopefully not the last one, but we do 
seem to be vacillating between the first two.

My understanding is that the conversion is meant to build the FAST 
vocabulary with minimal human effort, and then the vocabulary would be 
presented to untrained metadata builders for use as a set of controlled 
descriptors (for want of a better term). In that case, it's really not that 
important whether the converted records are well described by their FAST 
headings. One would not apply FAST headings in a real context by first 
applying LCSH strings and then converting them to FAST. The question is, 
how effectively can untrained users of FAST select appropriate FAST terms 
to describe an object? Are all the terms they need available, and can they 
readily be found? It is in this context that I worry about inconsistencies 
in the way FAST is constructed (many inherited from LCSH, or resulting from 
practice and coding changes), and about how it would be presented for 
searching.

I looked at the experimental interface OCLC has put up for searching FAST 
(at http://fast.oclc.org ). When I tried a keyword search on "rome" I got 
581 results. Scanning the list was baffling until I realized that they're 
sorted first by tag value--emperors and other persons first, then museums 
and other bodies in a new alphabet ... The Roman Empire finally turns up in 
the 151s under "E" for "Eastern Hemisphere..." This is not a friendly 
interface for the naive user.

I also tried a browse search on "education" as a "Full Fast heading." The 
result was a display promising 329 hits for "education" and individual hits 
for headings beginning "education ..." (e.g., "education and crime"). 
Browsing the 329 hits proved a daunting prospect, given uncertainty as to 
whether some one of them might prove to be a "best fit." I tried an 
alternative--an advanced search on "education" and "aids" as "topical 
keywords" and got 91 hits with more required browsing (I want audio-visual 
aids for veterinary education; I find "Education--Audio-visual aids," but 
maybe one of these more specifically formulated headings is better. "Adult 
education--Audio-visual aids"; "Vocational education--..."; will I find one 
for veterinary education? Again, not an easy task.

So, looking at FAST headings as a potential user, my main complaint would 
be that they're too articulated, to complicated, and there are too many of 
them. However well the converted headings work as alternate descriptors on 
the catalog records we looked at, I still have serious doubts about the 
usability of FAST as a file of authorized headings for the untrained 
metadata creator, and likewise, for the untutored searcher. I'm sure that 
further improvement of the searching interface will help; but can it 
overcome the basic problems of over-articulation and complexity? At this 
point, FAST is more like an alternate system of pre-coordinated subject 
strings (but lacking LCSH's voluminous documentation) than it is like a set 
of simple descriptors.

Stephen


****************************************************
Stephen Hearn
Authority Control Coord./Database Mgmt. Section Head
Technical Services Dept.
University of Minnesota
160 Wilson Library               Voice: 612-625-2328
309 19th Avenue South              Fax: 612-625-3428
Minneapolis, MN 55455      E-mail: s-hear at tc.umn.edu