Smart search: Issues with "-" in some queries

I believe you need to change the analyzer type to "Simple / Stop words / White space with stemming" because subset won't work properly.

https://docs.kentico.com/k11/configuring-kentico/setting-up-search-on-your-website/using-locally-stored-search-indexes/creating-local-search-indexes#Creatinglocalsearchindexes-Reference-Searchindexproperties

Hi Brenden, thanks for your reply. I just tried those analyzer types and none of them work. I am now going through and trying all of the other types because I need a solution. I am doubtful any of them will work any better.

Another weird thing I noticed if I search for ZH-2 I get results but if I search for ZH-32 I get nothing.

Update: None of the analyzer types worked. I am very puzzled why some search queries are returned without issue (ZH-2) and others don't match anything (ZH-32).

Hi Brenden, thanks for your reply (again!). To answer your questions: yes, I rebuilt the indexes after each time. Yes, the fields are searchable and work perfectly 99% of the time, except for the cases I mentioned.

I appreciate your telling me about Luke, I will definitely check it out. In the meantime, I just learned something after some more experimentation. The reason ZH-32 was not coming back is that it was not ever mentioned in isolation on the page (it was always referred to as ZH-32-2-2). Adding a space so it became ZH-32 -2-2 fixed the issue, and now it is returning the proper result.

The analyzer type I am using, Subset, does support partial matching, and I am not sure why the hyphen is breaking the partial match. A simple experiment is a manufacturer name, Sub-Zero. If you search sub you get a partial match. But if you search sub- or sub-z you get nothing. Search the full string sub-zero and it works fine again.

Again, it's all based on your analyzer. Even though you're seeing the results you expect quite a bit of the time, the fact that you can't see "sub-" or "sub-z" tells me the analyzer type is failing due to not matching the WHOLE string AND there is an issue with the special character.

Play around with Luke and test out what content is in your actual index. You'll only waste more time fooling around especially if you can't see what data is in your index.

What was the issue you found Deb?

The issue is that Kentico treats a hyphen the same as a space. There is nothing I can do, short of a custom analyzer, to correct this. Thanks for following up!

I'd challenge answer a bit more, for two reasons:

• Kentico's search is built off of Lucene, so it's a "limitation" of Lucene, not Kentico (just to be clear).
• I don't beleive this is a limitation at all. Mainly because you can search any kind of text, how it's analyzed is based on the analyzer type. I have a site which is using "$" as a filtering mechanizim on page data using a Smart Search index. I'm using the Standard analyzer and the field is setup as "searchable", "content" and "tokenized". The filter is wrapping the value in double quotes. So when it performs a filter (which is very similar to what you're doing), it's wrapping the value $ like so "$".

I'm not saying you're wrong, but I'm stating I would NOT create a custom analyzer because of this. If you haven't, do check out what the value is stored in your search index and try to do different queries using the Luke tool. That exposes all your indexed data and will give you a raw look at what is stored as well as create queries to get the right data.

Hi Brenden,

The "filtering mechanism" (as you call it) definitely sounds like a good solution. I am curious if there a way to set it from within the Smart Search node in Kentico or if it has to be built inside the project. I don't have access to the .NET project but I do have a dev colleague that I could recruit. I'm just not sure the extra effort is worthwhile when I could just edit the text on a page to match the search query on an as-needed basis.

I found this article on custom indexes ... is this what you are talking about? If not, can you point me to a link on Kentico's website that outlines what you are suggesting?

Thanks again for your continued help!

-Deb