Empatheia

Recently, emotion recognition has gained a lot of attention with the emergence of social media. Emotion recognition aims to discern and interpret human feelings towards a product, a marketing campaign, a political figure, a government decision, etc. The emotional states can be expressed through various modalities such as facial expressions, vocal intonations, and discourse (in this regard our dataset is made of oral speech analysed through written transcripts).  
However, there is a notable absence of annotated Arabic datasets for text and audio in political areas. Furthermore, advanced machine learning models and multimodal analysis integrating facial, vocal, and discursive aspects have not explored the correlation between these modalities in Arabic and the political domain. 
In this research, we present an Arabic dataset focused on political discourse designed for text and voice emotion recognition tasks. Additionally, machine learning and deep learning models are investigated to classify each modality (facial expressions, vocal intonations, and discourse) into distinct emotional categories, whereby we provide an analysis of the correlation between the three modalities. 

PARTENAIRES : Joëlle Chamieh, Rida El Chall, Vincent Legrand & Andrea Giammanco (UCLouvain & Lebanese University) - Interdisciplinary research project at the crossroads of Communication Sciences, Political Science, Computer Engineering & Physics, in co-operation with Louvain Transfer. 

DEBUT : 25 avril 2022