How to customize AI Search's search engine (assign search weight)

Junyu
Tera Contributor

Dear all,

We want to customize a set of search logic in AI Search for Virtual Agent.
Here are a few scenarios:
1. For different employees, the articles they find have different focuses: for example, IT personnel have a higher probability of finding IT-related articles (higher score)
2. Articles with higher ratings/helpful are easier to be searched: For example, articles with more user ratings of 5 stars/helpful are more likely to have high scores (similar to a recommendation system)

find_real_file.png

I don't know where they can be configured, thanks for the answers from the experts.

1 ACCEPTED SOLUTION

Richard Kiss1
Kilo Guru

Hi Junyu,

 

What you need is Result improvement rules. There is a ServiceNow Academy video about it, pretty much explains how it works and how to use it:

AI Academy: Personalize your AI Search experience with tuning

For example you can create a Result improvement rule like this:

find_real_file.png

and a boost action like this:

find_real_file.png

This way if the user's department containing IT ServiceDesk, they will be more likely to be presented with articles from Knowledge Bases containing IT.

 

Boosting KB articles by rating is a bit more tricky. You need a Result improvement rule again. Check the Activate on all queries checkbox if you do not want a condition.

find_real_file.png

Then create a Boost action for each rating tier you want to boost:

find_real_file.png

find_real_file.png

This way for all searches all the 5 star articles will be boosted, all 4 star articles will be boosted a bit less.

I hope it helps,

 

Richard

View solution in original post

5 REPLIES 5

Richard Kiss1
Kilo Guru

Hi Junyu,

 

What you need is Result improvement rules. There is a ServiceNow Academy video about it, pretty much explains how it works and how to use it:

AI Academy: Personalize your AI Search experience with tuning

For example you can create a Result improvement rule like this:

find_real_file.png

and a boost action like this:

find_real_file.png

This way if the user's department containing IT ServiceDesk, they will be more likely to be presented with articles from Knowledge Bases containing IT.

 

Boosting KB articles by rating is a bit more tricky. You need a Result improvement rule again. Check the Activate on all queries checkbox if you do not want a condition.

find_real_file.png

Then create a Boost action for each rating tier you want to boost:

find_real_file.png

find_real_file.png

This way for all searches all the 5 star articles will be boosted, all 4 star articles will be boosted a bit less.

I hope it helps,

 

Richard

Hi @Richard Kiss 

Thank you for your answer, 

Also I would like to ask if you know what Search Application do?

find_real_file.png

Thank you.

Hi Junyu,

 

You can think of a  Search application (Search Application configuration) as a container or parent record for a complete search experience. It contains/involves everything a searching capability needs in different parts of the ServiceNow platform. 

It involves a lot of things: how should a result look like on certain pages, what to search/not search, how to adjust results, what search engine to use etc.

When creating different search applications you can create searches behaving differently. If you want to change how a search behaves somewhere on the platform, the only thing you have to do is to switch the Search application. For example the Service portal using the Service Portal Default Search Application, but you can create a new one and set that application at the Service Portal record.

 

Richard

Hi @Richard Kiss1 The Result improvement rules seem really helpful. But I'm unsure how user criteria can be applied here. We have few user criteria for certain articles/catalog items so that they are visible only to few users. How do we show article or catalogs based on logged in user conditions? Currently, our Virtual agent AI search shows results just based on search text and ignores any existing user criteria & even shows expired articles. Any suggestions on how to handle this?