When using Natural Language Search (NLS), keep these limitations in mind.
- Natural language search works with these standard objects: Account, Case, Contact, Knowledge, Lead, Opportunity, and Task.
- NLS supports non-English language queries for renamed standard objects, renamed fields, and picklist values. Non-English languages aren’t supported for queries that contain boolean field values, dates, operators such as "top" and "my", and stop-words such as “how” and “to”.
- In Government Cloud, NLS doesn’t support queries that use location and date filters. For example, you can’t perform searches like my contacts in Boston and cases closed in the last 2 weeks.
- NLS doesn’t return search results for exact dates, such as closed cases January 23 2020.
- NLS doesn’t return search results for records with fields that an admin or end user can’t access due to field level security.
- NLS only returns one record if you have multiple articles with the same name. To receive multiple articles with the same name, enable Natural Language Search Filters (Pilot) and filter them accordingly.
Example A user searches for My accounts in Denver. Einstein Search doesn’t
compare the term “Denver” against the fields Billing Address and Shipping Address because the
user is restricted from accessing these fields. Instead, a keyword search for the entire query
is done, returning results for matches against any accessible fields. For example, if the name
of a record matches any of the terms in My accounts in Denver, that
record is returned.
- Limitations for Non-English Terms
Natural language search offers full support for English language queries. It can also understand queries using non-English languages, but there are some limitations. If a non-English query isn’t supported, Einstein Search does a keyword search instead.