azziimm7/Low-Resource-Language-Translation-Arabic-and-Nigerian-Pidgin-in-the-Spotlight

Low-Resource Language Translation: Arabic and Nigerian Pidgin in the Spotlight Project title :Low-Resource Language Translation: Arabic and Nigerian Pidgin in the Spotlight Description: *Abstract:* Large Language Models (LLMs) have become essential tools for various applications, including machine translation, customer service, and other automated systems. These models, however, predominantly support high-resource languages such as English, Arabic, and German, while low-resource languages remain largely underrepresented. Globally, around 1.2 billion people speak low-resource languages, yet the development of AI technologies for these languages lags significantly behind. This research seeks to address this gap by focusing on Nigerian Pidgin, a low-resource language spoken by approximately 100 million people across West Africa, alongside Arabic, a high-resource language. In this study, we will curate the first large-scale parallel dataset for Arabic-Nigerian Pidgin translation by extracting parallel sentences from the Holy Bible, a widely available resource in both languages. We will then apply existing literature on low-resource translation techniques using this dataset to fine-tune state-of-the-art language models, improving translation performance for Nigerian Pidgin. Additionally, we will explore few-shot learning with larger language models such as GPT-4, Gemini, and Claude to evaluate their effectiveness in handling low-resource translation tasks. This research aims to provide critical insights into improving machine translation for low-resource languages, contributing to a more inclusive development of language technologies.