Re: Apropos commands and regexps

[Kai Großjohann]

Miles Bader <address@hidden> writes:

> `or' is clearly wrong; even in emacs' `limited' universe, it generates
> way too many hits.
>
> E.g., (apropos "\\(find.*file\\|file.*find\\)") gets about 50 hits,
> whereas (apropos "\\(find\\|file\\)") gets over 700!
>
> Maybe your idea of `at least N matches' is a good compromise.

Information Retrieval research has shown that weighting and ranking is
what's needed.  Just list the "good" matches first.  With Boolean
searches, people need to issue a lot of queries to select an
appropriate answer set.  If you have ranking, fewer queries will be
sufficient.

But it might be useful to somehow indicate to the user the nature of
each match so that the user can decide what they want.  For example:
this group of matches contains all words, the following group of
matches misses the word foo, ...

kai
-- 
Silence is foo!