Best way to limit the number of matches + negative examples

[jonathan-dufault-kr]

I have a document that has 6 entities that I want to extract, and I've provided 16 examples of things to match, but the tool is identifying over 40 entities.

Is there an optimal way to instruct the tool to be parsimonious in what it returns? I could put it in the prompt, but I was wondering if this is the ideal, or if there's some parameter to adjust.

I was also wondering if there's a mechanism to instruct the tool of things I don't want to match. Like if I wanted to extract vehicle information about public transit vehicles (bus, train, airplane), but not personal vehicles. Currently I'm providing 8 examples of the public transit vehicles, and 8 examples of personal vehicles, but I feel like that's making the task too complex.