A surge in public availability of emerging GenAI-AD has brought back the promises of automated accessibility for people who cannot see or see well. This article tests those promises through a double-rendering method that asks GenAI-AD engines to describe a simple portrait of a person and then returns these generated texts into GenAI-AD engines for visualizations of what they earlier had described, revealing insights about GenAI efficacies, ethics, and biases.

In this paper, we investigate two approaches to building artificial neural network models to compare their effectiveness for accurately classifying rhetorical structures across multiple (non-binary) classes in small textual datasets. We find that the most accurate type of model can be designed by using a custom rhetorical feature list coupled with general-language word vector representations, which outperforms models with more computing-intensive architectures.