On Apr 17, Matthew Flatt wrote:
> At Fri, 17 Apr 2009 16:46:38 -0400, Eli Barzilay wrote:
> > On Apr 17, Neil Van Dyke wrote:
> > > Even a small application-specific synonym dictionary would be
> > > helpful.
> > >
> > > Or perhaps you already have synonym index entries in manuals, and
> > > can add more such entries to the manuals as you find key ones are
> > > missing.
> >
> > That's how it works now.
>
> You mean the latter, right?
Yes.
> When writing the old docs, I added explicit index entries for
> synonyms, and it was a pain. For the new docs, I threw out lots of
> synonym entries when porting, because a synonym dictionary seems
> like a better way to go, along with some automatic rules for plurals
> and things like that. Eventually.
(a) I think that the synonym information should come from the docs.
(That is, the semantic decision of what should be considered a
synonym belongs in the same world as writing the prose.) To be
used by the search, it should be some table that can be used by
the JS code.
(b) Actually using such a table is going to make things very slow at
this point, unless some normalization happens when the JS index is
generated so it doesn't have to do multiple comparisons when a
query is running. This might make the index bigger to, so it will
still have an effect on performance.
(c) I'm not sure how it will be done eventually, but there is also the
ordering issue to deal with -- using too many synonyms might make
the results appear in what looks like a random order.
--
((lambda (x) (x x)) (lambda (x) (x x))) Eli Barzilay:
http://www.barzilay.org/ Maze is Life!
_________________________________________________
For list-related administrative tasks:
http://list.cs.brown.edu/mailman/listinfo/plt-dev
