Page MenuHomePhabricator

Semantic Drilldown gives spurious results
Closed, DeclinedPublic

Description

Author: vipul.wikis

Description:
In this wiki:

http://groupprops.subwiki.org

When I try the semantic search [[Weaker than::Characteristic subgroup]] I get a bunch of results:

http://groupprops.subwiki.org/w/index.php?title=Special:Ask&q=[[Weaker+than%3A%3ACharacteristic+subgroup]]&p=format%3Dbroadtable&eq=yes&limit=50

When I try the same semantic search using a filter created for Semantic DrillDown, I get a number of additional results which are spurious:

http://groupprops.subwiki.org/wiki/Special:BrowseData/Terminology?_search_Weaker_than_in_terminology=Characteristic_subgroup

It seems to me (although I am not sure) that the Semantic DrillDown is using some approximate matching or substring matching instead of matching the exact term being searched so anything satisfying [[Weaker than::Characteristic subgroup of finite group]] also comes up in the answers.

There does not seem to be (to my knowledge) any way of disabling this to only restrict to exact matches.


Version: unspecified
Severity: normal

Details

Reference
bz28505

Event Timeline

bzimport raised the priority of this task from to Low.Nov 21 2014, 11:25 PM
bzimport set Reference to bz28505.

I just saw this bug now - I think I didn't notice it before because I didn't think it was related to one of my extensions. I changed "Semantic searches" in the title to "Semantic Drilldown" so it'll be clearer.

Anyway, I see the issue - values like "Characteristic subgroup of abelian group" are showing up when searching for "Characteristic subgroup". Yes, if you use the "combo box" input type, it does a substring search, instead of a search on the exact text. I think having it as a substring search is the right approach, since users can type in anything into that field. So, in this case, I'd say you have two basic options: don't use the "combo box" input type, or rename some of the values so that they're not substrings of other values.

By the way, in your explanatory text, "minimized" might be a better word choice than "circumcised".

I'm changing the status of this to "WONTFIX" - I can't think of a way around the problem, other than with the options I listed above.