Splendeurs et misères de la traduction automatique : la métaphore argumentative dans les discours institutionnels, quelle(s) issue(s)?

In recent years the evolution of neural machine translation (MT) has led the community of linguists to focus on its outcomes, and in particular on its results in terms of linguistic phenomena affecting lexicon and morphosyntax. We aim to explore the functioning of MT by observing the work of two open access tools: DeepL and eTranslation, general public version, meeting the challenge of the translation of institutional discourses. Placing ourselves in this field, we will test the effectiveness of these tools in rendering the figural and especially metaphorical significance that many leaders or institutional representatives adopt, namely the metaphor with an argumentative purpose. In order to avoid translation from the English text, we have selected speeches by the President of the European Council Charles Michel and the President of the French Republic Emmanuel Macron during the French Presidency of the EU. The language combination in question is FR-IT. The version accessible to professionals working at the DGT should not be neglected because the success of post-editing is unquestionable and the quality of MT in terms of interpretations and relevance of semantic variation is strictly correlated.