Estimating the transferability of state-of-the-art models in predicting moral values