Developer forum

The stemming handles different versions or soundings of words. I.e. plural version like "bike" and "bikes", but also sound alikes like "bikez" and "bikes" that sounds the same. So the stemming will ensure that "bike", "bikes", "bikez", "bikkes" etc are all indexed as "bike" and when searching you will find "bike" when searching for "bikes"

Synonyms can work regardles of the stemming. Synonyms would be other wors. I.e. "cycle" will become "bike".

But if you want to do stemmings as well as synonyms, you could benefit from a norwegian stemmer for your index. You can try "Norwegian" as the stemming. But that is not what is wrong with your setup.

See my setup attached.

Hi Nicolai,

Thank you for the details.

I have attached my setup. Unless I am missing some Type, the only difference is that your field is also Saved while mine is just Indexed and Analysed

Hi Nicolai,

I have checked my index using Luke and it seems that the value stored for productNameWithSynonyms is the same as the one for ProductName (which is supposed to be the source).

Which means that the configuration might be off.

Adrian

Hi Nicolai,

I have managed to figure out the difference.

I have initially considered group stem value as the keyword that has to be evaluated as a synonym. I have noticed in your example that you have used fast both as a group stem name and as a synonym.

I have done the same with my setup and I have included a custom string in the list of synonyms (adrian) in order to avoid any overlapping terms. 

And it WORKED!

However, the problem still needing a fix is the situation where the synonym contains a number.

If you look at my example, I was trying to match 6kt to sekskanskruer. Every other synonym works (including adrian) but not 6kt nor in6kt.

Maybe there are some issues with synonyms containing numbers?

Thank you,

Adrian

Cant tell. It is probably something related to how Lucene handles this. It could be stemming that does not want numbers in the text. So try using stem="sekskanskruer" and not "6kt" and see if it makes a difference.

Hi Nicolai,

I have tried it but the behavior is the same. It works for strings but not for combined numeric+strings no matter where the numbers are located inside the string.

Adrian

Hi Nicolai,

Any other suggestions? I really need to make this work and I have no idea if it's something I can control or it's something that you guys should fix.

Thank you,

Adrian

I do not think I can fix that.

The synonyms filter uses a tokenizer/analyzer inside that will split 6kt to 2 terms (6 and kt). So you would have to create another version of that analyzer that does not do this.

I guess you would have the same issue if you just index it with standard analyzer that it will become 2 terms and an fieldname<equals>6kt would probably also not work.

BR Nicolai

Hi Nicolai,

Thank you for your answer.

I don't think I will go over the trouble to create a custom analyzer because it is way out of our knowledge.

Is this something you would plan on including in future releases? I don't see it as an exception to our situation and it might be good to have built-in functionality that supports it rather than something custom.

Thank you,

Adrian