Arabic Multiclass Emotional Analysis Using a Bidirectional GRU-LSTM Deep Model on Twitter

Abstract

Recognizing emotional states within Arabic tweets is a challenging task due to the inherent complexity of the Arabic language. This complexity encompasses aspects such as ambiguity, agglutination, dialectal variations, and morphological richness. Additionally, user-generated data frequently include misspelled words, abbreviations, acronyms, slang, informal language, and non-standard punctuation, further complicating the task. In our research, we have implemented an efficient multiclass emotional recognition classifier. To achieve this, we proposed a Bidirectional GRU-LSTM deep learning model. This model is designed to capture high-quality global features within the input tweet. Our experimental evaluation, conducted using a benchmark dataset, has produced highly promising results and significant performance improvements.