Does ES interpret search terms differently to standard XF search?

[CTXMedia]

For example, if I search here (on the XF community site) for "CycleChat" it doesn't return any of the posts that have "CycleChat.net" in them. To find "CycleChat.net" I have to search for it exclusively.

Can you use quotes and asterisk and plus or minus symbols to denote phrase, wildcard, and include/exclude?

Are there any "special" ways you can demark your search words or phrases to get ES to behave in a particular fashion when returning results?

Just curious ...

Cheers,
Shaun

 

[Anthony Parsons]

I believe the instructions for ES said that XF was using the standard analyzer, not word stemming, which would return such results as you are citing above.

If you use word stemming option, then cyclechat would return wildcard results basically.

 

[Mike]

Stemming != wildcard.

ES may do tokenizing a bit differently than MySQL. It's debatable which is more correct really.

 

[Anthony Parsons]

Sorry... whatever the word is for searching for "test" will return "tested" "testing" etc... that is what the stemming option does, yes?

 

[Mike]

Stemming effectively indexes the root of a word. So in that case, test = tests = tested = testing, however != tester != testtube (yes I know, it's 2 words ). Where as "test*" would match all of those.

And the same query syntax is used for ES as MySQL (well, that's partially because we make it use the same).

 

[Anthony Parsons]

Ah... ok, I get it now. Wildcard would go after anything after test*, stemming will only go after adjectives, plurals, etc?

 

[CTXMedia]

Okay, thanks Mike, so it doesn't matter about the backend - XF will "handle" the query in a similar way; but if you *do* enable stemming, you get a slightly broader relevance?

 

[Mike]

Stemming just tends to be more user friendly. You can clearly see Google doing it at times.

(Google actually takes it a whole step further with aliases. For example, search for "IE" and it will match "Internet Explorer".)

 

[CTXMedia]

I must say I'm impressed with how fast it is and how little impact it has had on server load (haven't noticed any increase in average load values - even with 200+ members logged in and beavering away).