Just curious... do you know if Lucene can support "stemming"? For example, if a user enters "viewer" as the search term, a search engine that supports stemming will reduces the word to its root ("view") and returns all documents containing the root - like documents containing view, viewer, viewing, preview, review etc. Or conversely, if you enter view, it will expand the search with the alternates. (Giving weight to the preferred spelling).
Not a feature request, just wondering if it's a Lucene "freebie" that can be easily configured in. [originally posted on jforum.net by monroe]
I wonder if anyone has been successful in integrating stemming. I tried by replacing the Lucene StandardAnalyzer with the SnowballAnalyzer (using the Lovins filter). I modified the LuceneManager and the LuceneSearch modules to reflect this new analyzer (both the indexer and the query parsing need to reflect the choice of Analyzer). For some reason, I can't seem to get it to work. When I search, no stemming appears to being performed - it acts just like the StandardAnalyzer. I thought perhaps the re-indexing was not reflecting the new Analyzer, so I created new posts to test, but to no avail. I have the feeling I'm missing something in the configuration of JForum...
Anyone else try this?
Carlos [originally posted on jforum.net by cszamudio]
Update: I switched to the EnglishStemmer and that seemed to do the trick. Now wildcard and stemming work for searching the forums. Given that Lucene also support complex boolean expression, I also saw the need to add a new search filter that accepted a search expression (which I called filterByExpression), rather than use the filterByKeyword option. [originally posted on jforum.net by cszamudio]