You may wish to include data from several different entities in your search. Unlike database searching, searching with indexes is not conducted using joins. Remember, the main benefit of using an index is to allow the work of searching to essentially happen up-front, when the index is created rather than when the search is invoked. Accordingly, all database tables should be denormalized for indexing. The alternative, which is to create separate indexes, search them separately, then attempt to merge results is much more complex and inefficient.
Example say you have the following entities: Entity Person with attributes name, date of birth, and a foreign key pointing to an Address entity Entity Address with attributes street address, city, and country. You wish to create a search that allows you to search for persons by name, DOB, street address, city and country. You would create a searchable index that contains all the data from both tables.
When you have multiple entities contributing to a single search index, bear in mind that updates to any of the tables concerned can lead to the search index requiring an update.