[Snowball-discuss] Porter2: Accident and Accidentally

Martin Porter martin_porter@softhome.net
Fri Jan 24 17:23:02 2003


Steve, hello.

The short answer, I think, is that it can be fixed, but should not be. The
long answer is wrapped up in
http://snowball.tartarus.org/texts/introduction.html, which you might like
to look through.

On balance, incident/incidental etc have rather different meanings and it
would not help to stem them. On the other hand incidental/incidentally etc
have rather similar meanings and it is probably useful to stem them.

If you want to adjust the stemming rules, you have to work with a large
representative vocabulary and look at the overall change that tweaking one
rule makes. It is then surprising how often a very plausible extension to a
stemmer turns out on inspection to degrade rather than improve performance.

Martin