Domain adaptation for Hindi to English neural machine translation
Pragya Tewari, Anurag Singh Baghel · 2022
Machine translation is an area to increase the knowledgeable society of Indians without any language barrier.Machine translation (MT) is the procedure by which a fully automated software is applied to translate a source content from one original language (such as English) into target content of another language (such as Spanish) preserving the sense of the input text, and delivering fluent output text in the target language.Neural machine translation (NMT) is a deep learning-based technique for machine translation that provides state-of-the-art translation performance in scenarios where large-scale parallel corpora are available. A model that has already seen comparable sentences is more likely to produce a better translation when automatically translating a sentence. Neural Machine Translation (NMT) scheme that is tailored to that domain will also be the best at translating a domain of sentences. The emphasis of this paper is on domain adaption methodologies for NMT schemes. In this study, we define domain adaptation as any strategy for improving translations for a specific topic or language genre. Adapting model parameters with phrase translations in the area of interest is one example, as is restricting the model output to only produce vocabulary in the domain of interest.