We have created an Expression Rule named abc, where data is fetched from a Record Type with Smart Search enabled. The results are sorted in descending order of similarity score.
Smart Search is enabled on the Full Name and Address fields. For example, the address field contains values like "St.", and we expect that when a user searches for "Street", addresses containing "St." should receive the highest similarity scores.
However, we are observing unexpected behavior. For instance, the address "1817 W Broadway Columbia MO 65218" sometimes receives a higher similarity score than addresses that explicitly contain "St.".
Additionally, we would like to understand how Smart Search handles nickname matching—for example, whether a search for "Leo" would return results like "Leonardo" with a high similarity score.
Request:
Could you please help us understand how similarity scores are calculated in Smart Search?
Is there documentation or guidance on how abbreviations (e.g., "St." vs "Street") and nicknames are handled?
Are there any best practices or configuration tips to improve the accuracy of semantic matching in such cases?
Discussion posts and replies are publicly visible