You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@lucene.apache.org by "Michael McCandless (JIRA)" <ji...@apache.org> on 2015/06/29 23:21:06 UTC
[jira] [Updated] (LUCENE-6638) Factor graph flattening out of
SynonymFilter
[ https://issues.apache.org/jira/browse/LUCENE-6638?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Michael McCandless updated LUCENE-6638:
---------------------------------------
Attachment: LUCENE-6638.patch
Starting patch, full of nocommits, probably violating TokenFilter requirements, but the few tests (including the wtf/wow that's funny/what the fudge example from LUCENE-6582) seem to pass :)
I think the hardest thing still needing fixing is properly dealing with holes which are quite difficult for graph processing because they essentially produce a fully disconnected graph...
> Factor graph flattening out of SynonymFilter
> --------------------------------------------
>
> Key: LUCENE-6638
> URL: https://issues.apache.org/jira/browse/LUCENE-6638
> Project: Lucene - Core
> Issue Type: New Feature
> Reporter: Michael McCandless
> Assignee: Michael McCandless
> Fix For: 5.3, Trunk
>
> Attachments: LUCENE-6638.patch
>
>
> Spinoff from LUCENE-6582.
> SynonymFilter is very hairy, and has known nearly-impossible-to-fix bugs: it produces the wrong graph, both accepting too many phrases and not enough phrases, because it never creates new positions.
> This makes improvements like LUCENE-6582, to fix some of its bugs, unnecessarily hard.
> I'd like to pull out the graph flattening into its own token filter, and I think I have a starting patch that works. I started with the "sausagizer" stage on the branch from LUCENE-5012, but changed the approach so that it should not have so many adversarial cases.
> I think this should make SynonymFilter quite a bit simpler, hopefully to the point where we can just fix its bugs already.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org