Aprimorando a tarefa de detecção de ironia por meio de comitês de classificadores

Abstract

This paper presents an approach based on an ensemble to identify ironic tweets written in Portuguese. Irony is a linguistic phenomenon that can be perceived as a funny or peculiar aspect of a situation, using words that express the opposite of their literal meaning, often in a humorous context, and with intonation that reflects this. When dealing solely with text, detecting irony becomes quite challenging. This study evaluated several strategies for detecting ironic tweets using an ensemble in the tweet corpus of the shared task IDPT-21. The best result achieved 0.59 balanced accuracy using the bagging strategy, surpassing the state-of-the-art result of 0.55, indicating that the ensemble contributed to improving the results.