A Deep Learning Approach to Classify and Quantify the Multiple Emotions of Arabic Tweets

Abstract

In this paper, we introduce both a Multi-Label Classification (MLC) method to determine all the emotions expressed in an Arabic tweet and a Multi-Target Regression (MTR) method to determine the emotions’ intensities. MLC involves the prediction of zero or more classes per sample. It is one of the interesting research topics in Natural Language Processing (NLP), especially for the Arabic language due to scarcity of works related to it. MTR is a harder task compared to MLC, but it lends itself as a suitable representation for Emotion Analysis (EA), which is gaining more interest due to the increasing use of social media and the wide range of applications related to it. This work introduces a new study on the use of Deep Learning (DL) techniques for emotions classification and quantification in Arabic tweets.