A multivariate model for classifying texts' readability