Multitask learning for Arabic Dialects Identification and Machine Translation

Multitask learning (MTL) is a popular approach in natural language processing (NLP) to improve the performance of related functions. This paper presents a comprehensive approach to address dialect identification and translation challenges in Arabic NLP. The proposed model utilizes state-of-the-art transformer architectures for both tasks, ensuring robust performance in capturing contextual dependencies and linguistic nuances inherent in Arabic dialects. The primary objectives are to accurately identify Arabic dialects and facilitate their translation to Modern Standard Arabic (MSA) using a shared representation learning approach. The model architecture consists of a shared embedding layer followed by a transformer encoder for each task, employing multi-head self-attention mechanisms for hierarchical contextual information extraction. The model’s performance is evaluated using the F1-score for dialect classification, and the BLEU-score for translation performance.