Training data
학습 데이터
Also known as: dataset
The data a model learns from. Its quality and bias decide the outcome.
A model imitates what it has seen. It struggles with situations absent from the data and reproduces whatever bias is present.
Consistency often beats volume. A small set with a clear labelling standard can outperform a large noisy one.
For sensitive data such as children's photos or voices, collection and retention rules come first: what consent covers, how long it is kept, when it is deleted.
- ReviewProvenance, labelling standard, bias, personal data.
- In educationDecide consent and deletion first.