ANLP-RG at NADI 2023 shared task: Machine Translation of Arabic Dialects: A Comparative Study of Transformer Models

In this paper, we present our findings within the context of Subtask 2 of the NADI-2023 Shared Task.This task requires the exclusive utilization of the DIALECT-MSA MADAR Bouamor et al. (2018) corpus to develop sentence-level machine translations from Palestinian, Jordanian, Emirati, and Egyptian dialects to Modern Standard Arabic (MSA).However, MADAR lacks a parallel Emirati-MSA corpus.To address this challenge, we pre-trained the AraT5 transformer model using different configurations of the MADAR corpus and compared their performance results with those of existing transformer models.The best model achieved a BLEU score of 11.14% on the dev set and 10.02% on the test set.