Towards a Multilingual System for Automatic Irony Detection
Jihen Karoui, Farah Benamara, Véronique Moriceau · 2019
This chapter assesses the portability of both the annotation scheme and the automatic detection system for other languages. It presents two experiments. The first experiment consists of testing the portability of the multilevel annotation scheme for irony for other languages in the same family as French: English and Italian. The aim of this first experiment is not only to test the performance of the annotation scheme on other Indo-European languages that are culturally similar to French, but also to measure the impact of a set of pragmatic phenomena on the interpretation of irony. The second experiment consists of testing the performance of the feature-based automatic irony detection system on tweets written in Arabic. For the purposes of this experiment, the authors constructed the first corpus of ironic and non-ironic tweets in Arabic, studying the performance of features and assessing the algorithms used to classify tweets as ironic/non-ironic.