Objectiver l'intertexte ? Emmanuel Macron, deep learning et statistique textuelle

Abstract

International audience ; The present paper suggests that intertextuality can be brought out objectively by resorting to specific methodological tools. The case in point is political intertextuality in the speeches of the French president Emmanuel Macron. Deep learning (convolutional model) is first used to "learn" (satisfactory accuracy rate of 92.3%) the French presidential speeches since 1958: the speeches of De Gaulle, Pompidou, Giscard, Mitterrand, Chirac, Sarkozy and Hollande are then considered as the potential intertext of Macron's own speeches. Next, Macron's texts-hitherto unknown to the machine-are included in the model and the machine is instructed to assign Macron's quotations to one of his predecessors based on their linguistic content. Finally, the algorithm extracts and describes Macron's quotations and linguistic units (wTDS, lexical specificities, co-occurrences, morpho-syntactic labels) as they were interpreted by the machine in comparison to those of De Gaulle or Sarkozy, of Mitterrand or Holland. Macron's discourse is permeated with, sometimes explicitly but more often than not implicitly, by the discourse of former French presidents-a phenomenon that we shall refer to as "intertextuality"-and it turns out that Artificial Intelligence and textual statistics are able to identify such phenomena of borrowing, imitation and even plagiarism. ; Cette contribution propose un parcours méthodologique susceptible d'objectiver l'intertexte ; l'intertexte politique des discours du président français Emmanuel Macron en l'occurrence.Le deep learning (modèle convolutionnel) est d'abord utilisé pour « apprendre » (taux d'accuracy satisfaisant de 92,3%) le discours présidentiel français depuis 1958 : les discours de de Gaulle, Pompidou, Giscard, Mitterrand, Chirac, Sarkozy et Hollande sont alors considérés comme l'intertexte potentiel des discours de Macron.Ensuite, les textes de Macron – inconnus jusqu'ici du système – sont versés dans le modèle et nous forçons la machine à attribuer les passages de Macron à ...