[LUCENE-760] Spellchecker could/should use n-gram tokenizers instead of rolling its own n-gramming

• Type: Improvement
• Status: Closed
• Priority: Minor
• Resolution: Won't Fix
• Component/s: modules/analysis
• Lucene Fields: New
• Assignee: Otis Gospodnetic
• Reporter: Otis Gospodnetic
• Created: 22/Dec/06 23:49
• Updated: 28/Aug/22 11:33
• Resolved: 14/May/08 06:13

Description

The SpellChecker.java under contrib/spellchecker currently does its own creation of n-grams while it creates the index to search for alternative spelling suggestions, and then it again creates appropriate n-grams when it receives a query string/word to lookup alternative spelling suggestions for. Very clear sentence, I know.

I think it might be better if n-gram chomping could be outsourced to n-gram tokenizers that just made their way into contrib/analyzers via LUCENE-759.

If I see nods or if I don't get any nays I'll go and refactor SpellChecker.java a little bit to allow this.
SpellChecker has a page on the Wiki: http://wiki.apache.org/jakarta-lucene/SpellChecker

Thoughts?