Detecting Linguistic Bias in Government Documents Using Large language Models