On 1/16/07, Mike Klaas <[EMAIL PROTECTED]> wrote:
> It appears to be matching author:"Shelley Ohara" but when I do this > search no results are returned, searches like author:"Shelley O hara", > author:"Shelley O'hara" work as expected. Any ideas?
This problem is why some sloppiness is recommended when dealing with WordDelimiterFilter. "Shelley Ohara"~1 should work.
Hmm, shouldn't "ohara" be generated at the same position as "o", not "hara"? It looks like it is failing to do exact phrase matching because the index contains "shelley o (ohara|hara)"
The problem is, if you do it one way, the other way breaks. If you index "ohara" with "o", then a field like "O'hara Shelley" wouldn't match a query like "oraha shelly". There are a few possible options: - optionally index ohara at *both* "o" and "hara" - pick the "alignment" based on the token position in the stream... right-justify the catenations if it's the first token, otherwise left-justify. One could try to identify proper names and do the justification correctly too (blech). -Yonik
