[ http://issues.apache.org/jira/browse/LUCENE-517?page=comments#action_12369928 ]
Yonik Seeley commented on LUCENE-517: ------------------------------------- Yes, the error bars seem kind of large for the normal usage of norms, which is just length normalization if you don't include boosts. You could still use a single byte, but increase the number of bits dedicated to the mantissa to get better resolution (but with less range). http://lucene.apache.org/java/docs/api/org/apache/lucene/util/SmallFloat.html You could easily make the change for your index, but if would break existing indexes if we changed the default in Lucene. > norm compression breaks ranking for small fields > ------------------------------------------------ > > Key: LUCENE-517 > URL: http://issues.apache.org/jira/browse/LUCENE-517 > Project: Lucene - Java > Type: Bug > Components: Index, Search > Versions: 1.9 > Environment: N/A > Reporter: Randy Puttick > > The scheme of compressing document norms to one byte loses a lot of > information. This completely breaks search ranking on small fields because > there is no way to see the difference between documents with shorter and > longer fields that contain the same number of matching query terms. > Unfortunately the export of norms as a byte array seems to be pretty well > embedded in the code base so a fix would seem to require a major rev. -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: http://issues.apache.org/jira/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
