The article focuses on improving machine learning models for rhetorical analyses by using rhetorical feature sets for training artificial neural network models and it reports that the research reveals that a custom rhetorical analyses, outperforming more computationally-intensive architectures.