Specific "word" search not working

[rwm1962]

Brought to my attention by one of my members.

They are searching the word rapideal. It brings up a ton of results around the word rapid. The only post that includes the word rapideal is 4th in the results. Including the word in quote marks doesn't improve anything ie "rapideal".
Is there a fix or setting I've missed please?

 

[rwm1962]

Rebuilt the index - no better.

Tried another word press & "press" brought up loads of results but the first 2 results were for pressure. Press was the 3rd result.

If I use " around a word on google it brings back results for only that word.

I've also done a search on a random forum (a vb one) for press & it worked as I'd expect - results for press at the top of the list, not pressed or pressure.

 

[rwm1962]

Should have tried this first!

I have another XF forum. That search is working better. I did a test post with the word Rapideal - it came top of the list. Result. Rapideal is a unique one-off word on both sites now. Rapid is a common word across both sites.

That site is running the latest version of XES 1.1.6, the 'faulty' site was aversion or so behind. I've now upgraded it so both are running the same. Index rebuilt. No different.

Site 1:


Site 2



Please - what is wrong? How do I fix it?

 

[Mike]

Given that you posted this in Enhanced Search, this is probably a side effect of stemming. Stemming is programmatic and can get some cases wrong. The only alternative (within XF) would be to disable stemming.

 

[rwm1962]

Thanks - yes - it is ES. Would stemming get it wrong on one site but not the other?

I've turned it off & am in the process of rebuilding the index but seems wrong I can have stemming on one forum but not the other. Both on the same VPS.

 

[rwm1962]

Disabling stemming worked but getting less functionality for my buck on one XF install over another. Could XF version be a factor? The dysfunctional one is 1.4, the functional one 1.5.

 

[rwm1962]

This is working differently on my 2 sites - could it be the different versions of XF they're running on? Do I have to upgrade the older XF & see if that fixes it before opening a ticket?

 

[rwm1962]

made a new instance of the word 'rapideal' on the forum - only comes 3rd in the search list

 

[rwm1962]

Even on this forum it's not top:

 

[rwm1962]

But on my other forum where a similar word rapid is very common it's straight to the top

 

[Mike]

There shouldn't be any differences by version, but if you toggle stemming you need to reindex so that could account for differences.

In terms of relevancy, titles are higher relevance. Beyond that, relevance is a very complex algorithm with a lot of different factors. Note that with stemming, "rapid", "rapide" and "rapideal" are likely to all be the same word (as in, actually indexed as the same word). When you search for "rapideal", it will be run through the same stemmer and the result is what will actually be searched for.

 

[rwm1962]

There shouldn't be any differences by version, but if you toggle stemming you need to reindex so that could account for differences.

Index were rebuilt when that was toggled - search result was empty prior to doing that ( I checked out of curiosity)

In terms of relevancy, titles are higher relevance.

Ah - got it. One of my tests had the term in title, the other in the post body but not title, hence the different results.

Difficult one. I know for a lot of searches I'd want title to be the higher relevancy.

I'll have to give some thought to switching off stemming.

Thanks Mike - I've a better understanding how it works.

 

[AndyB]

I'll have to give some thought to switching off stemming.

Avoid doing that. Stemming is very useful and provides better results in most cases.

 

[rwm1962]

Cheers Andy - always good to have some input from experienced admins!

Are there any parameters that can be added to a search term to get around stemming. Quote marks don't work on ES as they do on google from what I've seen.

 

[AndyB]

Are there any parameters that can be added to a search term to get around stemming.

Unfortunately there isn't.