AI Text-To-Image for the Representation of Treaties Texts. The Case Study of Le Vite by Vasari

A.I. Neural Networks (NN) constitute interesting support for the analysis of the text-image relationship, which, since ancient times, has stimulated essential reflections on the resulting ontological value. Among these, the Text-To-Image (TTI), designed to learn the transformation of a text into an image, is the most suitable to make an innovative contribution to this complex investigation. In the literature, the applications of the TTI are many and concern both the learning of the network and its training in constructing images starting from ad hoc descriptions. This paper instead investigates the training of the NN in transforming the descriptions of treatise texts into images. This operation is complex because it requires knowledge of the two different constructive systems of the "text architecture", one governed by the grammar of the sentences and the other by the grammar of the text, which introduces a multitude of variables that are not easily decodable by neural networks. The contribution presents the results of the first phase of the research with the development of a procedure based on the interception of sentences and paradigmatic relationships to be decoded for the formation of the network. The experimentation is conducted on the treatise Le Vite by Giorgio Vasari, which contains an accurate description of the grotesques, created by the author in the church of Sant'Anna dei Lombardi in Naples for which graphic and geometric analyses have already been carried out.