Le deep learning auxiliaire de l’ADT dans le choix de textes à étiqueter en vue d’un corpus de comparaison : à propos de l’étude stylistique des lettres de Pierre Damien

To carry out a complete and reliable morphosyntactic labeling of Latin texts is a particularly time-consuming task. It is therefore necessary to choose wisely the texts to be included in a labelled comparison corpus when one wishes to study the intertextual distances between a given author, in particular a medieval one, and his predecessors. A stylistic research on the letters of Peter Damian (11th century) was the occasion to question the methods to be implemented to operate this selection. The intertextual distances were first computed on the forms using additive tree analysis. The results were then compared to the predictions of the deep learning, attributing with variable recognition rates passages of Damian to various authors of the comparison corpus. Where ADT relies primarily on the lexicon, the Convolutional Neural Network takes into account morphosyntactic parameters, with strong areas of activation suggesting a recognition of linguistic patterns that Damian shares with some of his predecessors.