Using fallback policy, I am able to handle the gibberish and completely irrelevant data. But what’s the solution to handle meaningful but not relevant to training data? Confidence value is high in these cases, which shouldn’t be actually.
Using fallback policy, I am able to handle the gibberish and completely irrelevant data. But what’s the solution to handle meaningful but not relevant to training data? Confidence value is high in these cases, which shouldn’t be actually.