On Thu, 16 Jul 2009, Pavel Sanda wrote:

just an idea - couldn't be this a bit automatized? something like export
to plain text, sort & uniq all words by alphabet, manually delete the
boring words (shouldn't be too much in tech documentation) and then run
some script mapping it back into the text.


  I don't think this will work. There are often variants of a word (e.g.,
'environment' and 'environmental') and we need to decide which, or both, to
index. Most importantly is trying to think of terms other than those we use
by which someone would try to find a reference in the text.

  This last point was driven home a couple of years ago when I wanted to
find the formula used to calculate Net Present Value (NPV) in my spreadsheet
application (XessSE). I looked in the built-in help and in the manual. No
luck. I called the vendor and the tech support guy also tried to find it
without success. Then it occurred to him to look under "Formulas" in the
index. Sure enough, that's where they were all listed, but they were not
listed individually. It did not occur to me to look for "Formulas" when I
wanted a specific one. Whomever wrote the index assumed everyone would look
in for the generic name rather than for the specific name. A perfect reason
for plenty of cross references ("See ..."). It is situations like this that
prevent automation of indexing.


Richard B. Shepard, Ph.D.               |  Integrity            Credibility
Applied Ecosystem Services, Inc.        |            Innovation
<http://www.appl-ecosys.com>     Voice: 503-667-4517      Fax: 503-667-8863

Reply via email to