A harvest-fair volunteer helper was trained mostly on notes from one township. Advice for other townships is curt or wrong. What bias source is that?
Select an answer to reveal the explanation.
Short Explanation
Trained mostly on one township, curt or wrong for the others. That is representation bias: the corpus under-represents those townships. A class-balance plot can hint at skew, but this item is the trust principle. Tokenizer bugs and dual-use name different problems.
Full Explanation
Representation bias appears when some communities are missing or thin in the training corpus. The helper then works for the over-represented township and fails the others. A class-balance plot can reveal skew, but this item is the trust principle, not a prep-plot step. Tokenizer bugs and dual-use name different problems.