LIBRA: Measuring Bias of Large Language Model from a Local Context