Page MenuHomePhabricator

Soundex search option
Closed, DeclinedPublic

Description

Author: kop

Description:
There should be a way to search for words that "sound like" (soundex) what
you've typed, especially in wiktionary.


Version: master
Severity: enhancement

Details

Reference
bz3140

Event Timeline

bzimport raised the priority of this task from to Lowest.Nov 21 2014, 8:45 PM
bzimport added projects: CirrusSearch, I18n.
bzimport set Reference to bz3140.
bzimport added a subscriber: Unknown Object (MLST).

Restored bug from flood attack.

(In reply to comment #0)

There should be a way to search for words that "sound like" (soundex) what
you've typed, especially in wiktionary.

What do you mean exactly? We now have good search suggestions for words that "look like" what typed, thanks to lucene.

kop wrote:

Soundex is this: http://en.wikipedia.org/wiki/Soundex

"Looks like" may make "sounds like" unnecessary. Dunno.

(In reply to comment #3)

Soundex is this: http://en.wikipedia.org/wiki/Soundex

English only? That would be bad.

"Looks like" may make "sounds like" unnecessary. Dunno.

Could you maybe do some test on Wiktionary? Lucene suggestions are sometimes surprisingly smart.

kop wrote:

Yes, English only would be bad.

While using Wikipedia I've noticed that it does come up with good suggestions.

Isn't going to happen for lsearchd as it's hit end of life.

This is a super cool idea though that we should look into at some point for Cirrus.

Relevant ES plugin possibly: https://github.com/elasticsearch/elasticsearch-analysis-phonetic

Yeah, I was looking at this to see if I could jam it onto the suggestions somehow or just use the plugin for search. The problem I saw (without digging too deap) is that Soundex and its ilk are pretty English centric.

Setting to NEW because REOPENED always looks to me like we tried to fix it and failed.

Until there's a multilingual way to do this we shouldn't even try.

"Looks like" bugs are things like bug 39501 and so forth, outside the scope of this.