Re: Yes but

On 5/9/07, Kristin Antelman <kristin_antelman_at_ncsu.edu> wrote:
> Casey,
>
> You're right that exposing subject vocabulary in facets can lead users
> to the correct vocabulary, but in the current faceted catalogs it's hit
> or miss on recall using that link to the correct term, because your
> initial retrieval set was created from a keyword search.  At NCSU, we
> call this the "revolutionary war" problem (user searches "revolutionary
> war" while correct heading is "United States--History--Revolution,
> 1775-1783").

We've tried a couple of approaches in our prototype:

- searching for the search terms as keywords in the subject
authorities, including scope notes and showing the results as
"possibly related searches".  This helps utilise the scope note and
"see also" and "see instead" information in the authorities

- keyword stemming, so for example "sociology of education" will match
(but with lower relevance) "sociol" and "educat", (hence matching
stemmed indexed terms "sociology" and " educational")

- sorting the subjects found in the matched records and displaying the
LCSH hierarchy so that refinement can occur at any level in that
hierarchy.

For example: http://ll01.nla.gov.au/search.jsp?searchTerm=revolutionary+war
returns:

- Possibly related searches:
   For Revolutionary War, American See United States History
Revolution, 1775-1783
   (plus others)

-Related subjects:
 United States [1172]
   History -- Revolution, 1775-1783 [232]
     Biography [116] -- Juvenile literature [52]
     Registers [140]
     Campaigns [116]
     Personal narratives [108]
     Juvenile literature [84]
     Fiction [52]
     Sources [48]
    (plus others)

http://ll01.nla.gov.au/search.jsp?searchTerm=sociology+of+education returns

- Possibly related searches:
  For Education and sociology See Educational sociology
  (plus others)

-Related subjects:
 Educational sociology [2093]
   Australia [273]
   Periodicals [161]
   United States [119]
   Great Britain [91]
   Canada [49]
   New Zealand [56]
   Congresses [42]
   (plus others)

http://ll01.nla.gov.au/search.jsp?searchTerm=eskimos returns

- Possibly related searches:
   For Eskimos See also Inuit
    General note: Here are entered works limited to the indigenous
Arctic peoples
    of Greenland, Canada, and northern Alaska. Works discussing collectively the
    Inuit peoples and the related Eskimo peoples of southern and western Alaska
    and adjacent regions of Siberia, or works for which the individual
group cannot
    be identified, are entered under Eskimos.
    (plus others)

-Related subjects:
  Eskimos [308]
     Alaska [144] -- Nome -- 1890-1920 [48]
     Juvenile literature [128]
     Folklore [112]
     Canada [104]
     Greenland [104]
     Fiction [80]
     Social life and customs [76]
     Antiquities [52]
     Northwest Territories -- Pelly Bay Region -- Social life and customs [48]
    (plus others)

The UI is currently a hodge-podge of each-way bets with the kitchen
sink thrown in, just to demonstrate ideas.

> ...
>
> Of course, this problem is not a problem with faceted navigation, but of
> keyword searching.  The faceted navigation interface does, I think, lead
> to a false sense that the catalog is making the connection between
> keyword and controlled searching, when in many cases it's very much a
> partial connection.  (e.g., The lost recall for a user in our catalog
> searching "causes of the revolutionary war" is over 99%: 3 hits starting
> w/kw search and navigating to the correct heading vs. 388 books w/the
> correct heading).

Sure  - one of the things we are looking at is creating topic pages
which will be found by searches such as "causes of the revolutionary
war".  Initially they are likely to be generated from authorities, but
will be augmented with more content over time.  Also, we want to
provide the ability to "fan out" from a hit to find "more resources
like this" based on the most characteristic attributes of the found
resources (LCSH headings, tags, maybe one day citation links and
circulation data...).  Amazon's listmania is an inspirational model in
this regard.

Another important lever is full text searching of the IA Text Archive,
Google Books and Amazon to find resources held in your library(ies).
We've done a mockup of a search on "free trade globalization united
states" which is actually implementable now, except, sadly, for the
Google Scholar searching because I dont think it has an API yet:

http://ll01.nla.gov.au/mock3.html

The idea is to use the full text searching of these services, link to
their wonderful content which we dont have (full text, citation links,
google map mashups etc) but also annotate their results with library
availability.

Maybe this helps us to be perceived as a starting point for a search
which will combine results from multiple sources of information, which
will locate information immediately available electronically as well
as information locally available at the searcher's library, which will
assist the searcher with hints and suggestions and to do so in a
commercially disinterested way.

Kent Fitch