Multiple Stemmers seem to conflict

public
group: Search
ducdebreme - Wed, 2008-06-25 07:43

We have a multilingual website providing German, English, French and Italian content on a Drupal 5 instance.
We used to use Drupal's internal search and had many issues about terms that were not found. I found, that the problems might be solved using stemming.
So i installed stemmer for all the languages: porterstemmer, de_stemmer, ...

But we still had issues. I discovered that all the stemmers are called one after each other. They are all hooking into hook_search_preprocess(&$text). This way, all stemmers are called in a pipeline and thus destroying the content.
I thought, the problem might be solved by forcing the stemmers, only to act, if the content's language fits to them. But i found that the stemmer has no knowledge about the language of the node, that is about to be indexed.

  • Does anyone have ideas?
  • Are the issues targeted with new search approaches?

This is a great bug report/feature request

robertDouglass's picture
robertDouglass - Wed, 2008-07-02 17:17

Can you please file it in the issue queue for D7? The bug is that hook_search_preprocess is not language aware.


Thanks for the feedback...

ducdebreme - Fri, 2008-07-04 07:21

I added a feature request:
http://drupal.org/node/278443