English-to-Malayalam Machine Translation Framework using Transformers

Machine Translation (MT) is the task of automatically transforming one language into another while keeping the meaning of the input text and creating fluent text in the output language. Machine translation is in high demand in today’s world. Even while this field has risen in popularity, it still falls short of meeting the needs of people who speak various languages. This work investigates how multiple models can be utilised to translate English into Malayalam. Models such as Seq2Seq and pre-trained MarianMT model for encoder and decoder, as well as several attention models such as Bahdanau attention, multihead attention, and scaled dot product attention, have been examined, and both subjective and objective analysis have been performed to check the validity of translation. Fine-tuning of the MarianMT model has been done to improve the translation task. The findings show how these models can improve low-resource language translation.