[WEB4LIB] Search engines and positronic brains

Bob Duncan duncanr at lafvax.lafayette.edu
Thu Jun 3 08:52:24 EDT 1999


At 12:45 PM 6/2/99, Joyce Ward, Director, Content Classification, Northern
Light Technology LLC wrote:
>I took the Ask Jeeves challenge, then I did the same query:
>
>What are the Riverbank Publications?
>
>in Northern Light.
> . . .
>Note that the top 10 documents or so are relevant, and that we generate a
>a folder for 'Cryptography'.

I hate to be picky, and I'm sure Joyce is a very nice person who works very
hard to make our lives easier, but...

Out of the first 10 documents, two are 404s and the other 8 are various
forms of bibliographies.  I suppose a reasonably intelligent person could
figure out based on the bibs that the Riverbank Publications have
*something to do with cryptography*, but to consider these results
"relevant" to the query "what are the Riverbank Publications?" is
stretching the definition of relevant close to the breaking point.

Also, the "or so" documents (namely results 11-15) refer to a painting
("The Riverbank");  a grassroots coalition ("save Riverbank Park");  a
piece of software ("RiverBank");  an industrial complex;  and a publication
("Riverbank Review of Books for Young Readers").

NL did a good job of floating the most-likely-to-be-relevant docs to the
top of the 2100+ results returned, but I doubt this had little to do with
so-called "content classification" -- the search "riverbank publications"
(with quotes) turns up 11 records in NL.  Looks like simple phrase
weighting to me.

Bob Duncan

~`~`~`~`~`~`~`~`~`~`~`~`~`~`~`~`~`~`
  Robert E. Duncan
  Systems Librarian
  David Bishop Skillman Library
  Lafayette College
  Easton, PA 18042
  610-330-5156
  duncanr at lafayette.edu
  http://www.library.lafayette.edu/


More information about the Web4lib mailing list