This talk offers an overview of the current state of Arabic machine translation using deep learning techniques, focusing on the challenges and potential advancements in this area. While deep learning models are highly capable, they often struggle to fully capture the complex morphological and syntactic characteristics of Arabic. Furthermore, the scarcity of large-scale parallel corpora in Arabic limits the development of robust models. To tackle the issue of out-of-vocabulary words, this study introduces innovative solutions that make use of Arabic’s inherent morphosyntactic features. The challenges posed by long sentences, which can be problematic for neural architectures, are also examined.
Speaker: Azzeddine Mazroui
Azzeddine Mazroui is a Full Professor at Computer Science Research Laboratory, Faculty of Sciences, Mohammed First University (Oujda, Morocco). He received the “Doctorat d’Etat” in numerical analysis at the Mohammed First University, 2000, and PhD in probability and statistics at the Pierre & Marie Curie University France, 1993. He is the director of the Natural Language Processing team of the Laboratory Researches in Computer Science. His main research interests include the areas of NLP and image processing.