[SPTUsers] SPT Search results

[Michael McFadden]

Typically we have created our own synonyms file for software we have used in
the past.  It has been relevant to the information we added to our knowledge
management system.  Creating a synonym file large enough to fit every
solution would require more time than it is worth and wouldn't be very
useful to specific implementations.  If somehow SPT was installed with an
empty synonyms file that users can use how and if they wish would be a
perfect solution.  With the type of data we will be putting into SPT it
would be extremely useful to have.  The previous statements also apply for
stop words.  A lot of our users aren't technical in nature and don't
understand that potentially every word in a query will have an effect on the
results.  However if I were to choose one over the other, synonyms would be
the most useful in making searches more relevant.

On Wed, Apr 09, at 12:57:45PM, Michael McFadden wrote:
> Are there any stop words file to not allow words like "a" "to" "and" etc
to
> mess up the relevance of a search?  Also is there a way to setup synonyms
so
> that for example a search for the word "church" would also return results
> with "synagogue" or "cathedral"?

   Stop words are used by the recommender system, but the search engine does
   not support stop words or synonyms at this time.

   As to stop words, the assumption was that users would specify search
terms
   they thought were relevant, rather than type in a question or sentence.

   For synonyms, an issue was that they can be problematic because there is
   an implicit assumption as to the intended use of the word.  For example,
   if the user enters "tone", do we also return results for "pitch" and
   "timbre", or "attitude" and "manner", or "hue" and "shade", or "vigor"
and
   "elasticity", or all of the above?  If the assumptions are off by even a
   bit, it's easy to return a lot of search results that seem irrelevant to
   the user, and obscure meaningful results.

   In the case of both stop words and synonyms we chose the path that seemed
   to provide the most accurate and understandable results for the user.  I
   expect that we'll revisit both of these issues in the future, and will
   likely add support for both in some fashion in a future software release.

   Ed


--
   Edward Almasy                                     ealmasy at scout.wisc.edu
   Research Director                                   1308 W Dayton Street
   Internet Scout Project                                  Madison WI 53706
   Computer Sciences Department                        608-262-6606 (voice)
   University of Wisconsin - Madison                     608-265-9296 (fax)
_______________________________________________
SPTUsers mailing list
SPTUsers at scout.wisc.edu
http://scout.wisc.edu/mailman/listinfo/sptusers