The ~ syntax can only be applied to a single phrase, i.e. "the greate one"~3. This sets the slop allowed for the phrase to be 3. The defintion of slop can be found by searching the archive, but it will not easily allow for what you are looking for. The Lucene query language has not yet recived Span systax support. You might want to check out Surround in contrib which is an alternate syntax with support for Spans. I also have a parser that adds proximity support...I am currently sitting on version 1.0 waiting for some free time to release. My syntax consists of binary operators including a proximity operator. Precedence works as expected (being a binary op sytax) and a search for (alien ufo crash site ) ~10 (Roswell NM New Mexico ) would make sure that each term on the left is within 10 of each term on the right (if you default space operator was AND).
-Mark On 5/9/07, Walt Stoneburner <[EMAIL PROTECTED]> wrote:
According to the documentation for Lucene's Query Parser Syntax, the tilde operator provides a proximity search. For instance, "Harry Hallows"~6 should match the text 'Harry Potter and the Deathly Hallows'. And while this is fine for two token phrases, I was wondering if it worked well for subexpressions: ( ( "alien" "ufo" "crash site" ) ( "Roswell" "NM" "New Mexico" ) ) ~10 Will this match "alien ship crashed in Roswell" in addition to "ufo spotted hovering over New Mexico"? I've been searching through the Lucene groups (http://www.gossamer-threads.com/lists/lucene/java-user/) for span and proximity, but have come up with nothing about such syntax or capability. Additionally, I tried entering expressions containing a tilde into Luke and then doing an explain structure. Surprisingly, I didn't see a difference between a simple case like "A B" and "A B"~10 in the explanation. Is this a problem with Luke, or did I possibly miss something trivial? Anyhow, the more important question is: Is it possible to do proximity searches on lists of terms? Thanks, -Walt Stoneburner, <[EMAIL PROTECTED]> http://www.wwco.com/~wls/blog/ --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
