Service Desk portal search does not find obvious request types

[sdwire]

The "What do you need help with?" search on our Help Center portal is behaving in an unhelpful way.

I can type in two simple words. Of the three request types it suggests, two of them have neither of the two words in the title, help text, or any fields. However, one of the other request types that has both of those words in its title does not show up.

I have already completed a full re-index of JIRA, and the behavior has not changed.

As it currently behaves, the customer portal is nearly useless. How can I get the portal suggestion to return relevant results - request types that actually have the words I typed?

[Damian Rosochacki]

Hi Steve,

I'm running into a similar problem. The portal can only scan the request type name and description, which limits its usefulness. I wish we could add tagged words or phrases to the request type to make them searchable in a more natural way.

Example:

I have a request type called "Printing problems" with a description that says "Printer out of ink? Can't find a printer?"

If I search for "print" it finds it. If I search for "can't," it finds it. If I search for "can't print," it fails.

If I could tag the request with words phrases like print, printer, printing, 'can't print', 'printer doesn't', 'printer is', it would be much easier to use the portal for the end users. As is, I've instructed my users to only use keywords when using the search bar. 

[sdwire]

Oh - and the request types I expect to find are indeed searchable.  If I remove one of the two words from the search, I can see the request type I expect to find. But when I add another word that's actually in the title, it goes away and is replaced with others that have neither of the words that I typed.

[Miroslav Suchopar]

Hi,

I have been dealing with similar issue recently and I have found an article describing the HelpCenter search algorithm - so called smart graph. See https://www.atlassian.com/blog/jira-service-desk/introducing-smart-graph-machine-learning-in-action

Maybe it could help you. 

It's a hybrid algorithm that combines keyword search and machine learning. Thanks to machine learning, once you create a request with summary "email new request" and manually choose appropriate request type, next time the search result for the same string "email new request" will return desired request type. 

So, it means that by creating couple of requests, you can teach the algorithm to return correct results for any request type. 

--ms

[Sue Lund]

Is there a way to unteach it so it stops pulling up the wrong request types.  It seems it must be doing that because people have submitted the wrong request types.

[Miroslav Suchopar]

Hi Sue, 

I believe, when you or your users create several new requests with correct request type, the smart graph updates it's associations and starts returning correct request type at the top of the list. 

The associations are also updated when agent updates the request type in existing ticket, which was created with the wrong request type. 

You can also try deleting the wrong request type (if possible) and creating new one. However, I am not sure if all the associations associated with given request type will be removed, too. 

[Sue Lund]

There is no way to manually override it, is there?
We don't want to update request types or delete and recreate them.

[asolou]

Having the same issue, we have a request lets say, Request new email. if we rephrase it to Email new request or something. It doesnt find it. Even tho we use the same words in a different order. 
Are you aware of any open suggestion we can vote on for this?

[Varsha Gupta]

Anyone have solution for these.?

[Jousef]

Hey all,

Don't have an answer to this, only wanted to add to this that not only can you only search for keywords as listed as labels/request type name/description but also will the search not work for incomplete words.

Searching for an article called Pandora's Box that I want to open for example. :)
If I type "Pando" nothing will be displayed in the search results.

Mistakes while searching something quickly are happening all the time and having a smarter search would be much appreciated.

-Jousef