529347 - Add custom wordforms for common multi-word search terms

Status: VERIFIED FIXED

Description

[Chris Ilias [:cilias]]

I've come across a case where someone looking for issues with ZoneAlarm used the search term 'zone alarm'. The instances of "ZoneAlarm" did not appear in search results. This would also apply to "book marks", "fire wall", "home page", "web site", etc.

SUMO 1.5 worthy?

Comment 1

[James Socol [:jsocol, :james]]

Adding infix indexing (aka: substring searches) is extremely expensive for the indexer. However, an alternative method is to add a wordforms list with specific overrides, like

Zone Alarm > ZoneAlarm
book marks > bookmark (singular for stem matching)
fire wall > firewall

etc.

Comment 2

[[:Cww]]

plug-in -> plugin
addons -> add-ons

Comment 3

[James Socol [:jsocol, :james]]

Created attachment 414112 [details] [diff] [review]
add wordforms.txt

NB: This patch needs to be applied directly in the scripts/sphinx directory, not from /trunk/.

This adds a wordforms.txt and references it in sphinx.conf (also correcting references to stopwords.txt). These .txt files are found in ROOT_PATH from localsettings.py--you may need to move things around.

Comment 4

[James Socol [:jsocol, :james]]

Oh, and a good search for testing is "face book". Before adding wordforms, I found 8 results in my database from Nov 18. After adding wordforms it was over 250.

Comment 5

[Michael Morgan [:morgamic]]

Comment on attachment 414112 [details] [diff] [review]
add wordforms.txt

Looks reasonable to me.  Could you fix the justification for wordforms?  It's not aligned with the other stuff.

Comment 6

[James Socol [:jsocol, :james]]

r56764.

Justified both the wordforms.txt and the instructions in sphinx.conf (will attach).

Comment 7

[James Socol [:jsocol, :james]]

Created attachment 414294 [details] [diff] [review]
justification

Comment 8

[Vishal Kamdar [:vish_moz]]

Verified, FIXED.