Is there a way to improve entities recognition training?

I have created an intent within which there are two entities to recognize, an object and a place.
For each of the two entities I have given numerous examples, with their synonyms.

Yet, in a way that seems arbitrary, it often does not recognize the object and/or place, especially if there are minimal changes, the relevance of which should not be so absolute if the correct structure (when and where to look for an object and a location) has been ‘learned’.

Is there any trick anyone can suggest for more effective training?
Many thanks in advance to anyone with suggestions!