Make AI Search Even Better with some Tuning
Tune the search experience and improve the results based on information specific to your organization. You can quickly test how the search behaves with the Search Preview, add vocabulary specific to your organization and create rules to influence the results.
Test with Search Preview
AI Search administrators (with the ais_admin role) can leverage the Search Preview UI to test AI Search queries using settings from a selected search profile.
The Search Preview UI is installed (along with other useful tools) with the Advanced AI Search Management Tools plugin.
To access the Search Preview UI, navigate to AI Search > Preview > Search Preview.
The Search Preview helps you assess search performance and debug and test queries for a range of profiles and user contexts.
Based on the test results, you can make changes to tune the search experience and improve the results. There are three areas: Synonyms, Stop Words and Result Improvement Rules.
Synonym Dictionaries
Although AI Search understands semantics, some unique words or acronyms are specific to your organization. These additional terms can be added to Synonym dictionaries to make the search more relevant. These dictionaries can also be used to add words with similar meanings and expand search queries. Synonyms allow the search to factor variations or ways a search term may be expressed.
You can create a new dictionary using the Synonym related list on the Search Profile or by navigating to Search Query Settings > Dictionaries. A synonym dictionary is language-specific.
By default, the synonym dictionary for English contains synonyms for IT words such as "computer", "laptop", "PC" and "Mac".
Stop Word Dictionaries
Stop words remove search query terms that don't produce meaningful results. Examples of such terms can include articles, conjunctions, personal pronouns, and prepositions. By defining these terms as stop words, you can improve search precision and result relevancy for users.
You can create a new dictionary using the Stop Words related list on the Search Profile. A stop word dictionary is language-specific.
By default, the list of stop words for English contains over 150 terms including "if", "its", "this" and "the".
Result Improvement Rules
Result Improvement Rules are rules to boost, block, or promote search results for specific searches or the user context, such as a user's location or device.
- Boost: increase the relevancy scores of search results that satisfy a custom condition
- Block: prevent specific records from appearing as Genius Results or search results
- Promote: make specific records appear first
To create a new Result Improvement Rule, navigate to the "Result Improvement Rules" related list on the Search Profile. Each rule is made of a date range, a language and an optional trigger. The rule is only active during the date range and for the user's language.
What's next?
At this stage, you have completed the configuration of AI Search! Learn how to deploy your configuration into production and monitor analytics to capture opportunities for improvement in this article.
Go back to the Quick Start Guide.
References
 |
Documentation |
 |
FAQ |
Don't forget to subscribe to our AI & Intelligence forum to get the latest updates!
Hi @cwillard in ESC for example, you'll want to make sure that you have the Autocomplete Suggestions within the AI Search application configuration is configured similar to the example below.
Product doc for reference:
Hi @Gerard Dwan,
Thanks for the reply. For me the search response is about 2 - 3 seconds I would say. I'm looking at the AI Search Analytics dashboard and the average is over the 4 second mark and has been like that since it's been implemented. I've re indexed all our tables to see if this would help at all but aside from that not sure what else we can really do to improve this.
Is there any way to see where the bottleneck lies? e.g. Whether is platform, network or application related?
Last 3 months
Last month (01-Feb - 28-Feb)
