Abstract

This paper examines tooled methods used to analyze digital corpora, linked with theoretical orientations of scientific research. Indeed, building corpora with various textual data is not neutral, and the process of constituting corpora gives specific meaning to the data. Currently, the use of Artificial Intelligence (AI), or Deep Learning, questions semioticians, and linguists, on the possible interpretative processes from these treatments which often come from "black boxes". From the concepts of instability, deformability, and chaotic units, this paper considers the hermeneutical dynamic which governs the use of digital corpora and their apprehension by digital tools. A double problem arises: hermeneutics of digital corpora, and hermeneutics of digital tools which allow the analysis of digital corpora. To approach this, the description of the different scientific models (Meunier 2019) makes it possible to propose a tooled analysis method that meets the principles of the Theory of discursive objects (Longhi 2015). This method is illustrated with the analysis of the term "enemy" in a corpus of political interviews, thus posing the different issues from a hermeneutical perspective for the digital analysis of discourse.