Abstract: This paper outlines strategies for the automatic detection of irony in literary-critical texts, which leverages a manually annotated corpus. We propose a typology of ironic statements established through observations in the corpus, which includes articles from the Mercure de France and Barbey d'Aurevilly's Les Bas-bleus. This study highlights an operational relationship to irony that varies according to author, sometimes revealing different literary-critical strategies. The annotated corpus is constructed in order to enable the future training of machine learning models.