MYANMAR-WA MACHINE TRANSLATION BASED ON TRANSFER LEARNING

Abstract

Artificial Intelligence (AI) is rapidly changing the world, affecting every aspect of human daily lives. Natural Language Processing (NLP) is itself a broad field that lies under AI. NLP depends upon linguistics and is responsible for making computers understand text and spoken words the same way humans do. NLP combines rule-based modeling concepts for human speech language, computational linguistics with some statistics, Machine Learning, and Deep Learning to enable computers to understand human speech language, which can be in the form of text or voice data. Machine Translation (MT), translates meaningful text from one specific language to another language without human involvement. The research area of NLP in machine translation has been a significant advancement in the research area of AI. NLP allows computers to comprehend, analyze, and generate human language in a way that’s more organic and contextual. It involves several subtasks such as part-of- speech tagging, sentiment analysis, named entity recognition, and more. These are applied in various stages of the translation process, augmenting the understanding of the specific source language and the generation of the target language. Incorporating NLP into machine translation has enhanced its capabilities and has led to the creation of more sophisticated translation models. However, it is important to note that NLP- based machine translation is still a developing field. Challenges such as handling low- resource languages, maintaining the source text’s style and tone in the translated version, and understanding cultural references and idioms still persist. Transfer learning, although its exact nature is unknown, enhances the quality of machine translation for low-resource systems. While there are many advantages to transfer learning, the three primary ones are reduced training time, improved neural network performance (for the most part), and reduced data requirements. The main problem of some Machine Translation (MT) systems is the need for a large range of parallel resource data for source-to-target language translation. To overcome this problem, previous research has shown that pivoting, if the pivot language is closely connected to the source and target language pair, produces translations of higher quality. In this exploration, pivot transfer learning-based MT is applied for the translation from Myanmar language to Wa language using English as the pivot language (Myanmar-English and English-Wa).