Optimizing Multilingual Sentiment Analysis in Low-Resource Languages with Adaptive Pretraining and Strategic Language Selection

In the realm of natural language processing (NLP), sentiment analysis has traditionally focused on resource-rich languages like English, often overlooking the linguistic diversity and potential of low-resource languages. This research aims to ad-dress this imbalance by concentrating on multilingual sentiment classification in twelve African languages: Hausa, Yoruba, Igbo, Nigerian Pidgin, Amharic, Algerian Arabic, Moroccan Arabic Darija, Swahili, Kinyarwanda, Twi, Mozambican Portuguese, and Xitsonga. Recognizing the critical lack of comprehensive datasets for these languages, our research not only highlights the need for such resources but also endeavors to fill this gap by assessing current NLP techniques and devising novel methodologies specifically tailored to these linguistically diverse yet under- resourced languages. Central to our study is the development and deployment of a transformer-based language model that is intricately adapted for African languages with limited resources. By refining the XLM-R model on the AfriSenti-SemEval 2023 Shared Task 12 multilingual sentiment classification datasets, we explore the model's efficacy in handling sentiment analysis tasks across these languages. The empirical results demonstrate the model's capability to perform sentiment analysis with considerable accuracy, even in languages that have traditionally been marginalized in the digital sphere due to minimal linguistic resources. This research not only contributes to the field of NLP by extending its reach to less-researched African languages but also underscores the importance of linguistic inclusivity in the development of AI technologies. By successfully navigating the challenges posed by multilingual sentiment classification in low-resource languages, our study takes a significant stride in closing the digital language divide. It emphasizes the potential of NLP to be a tool for cultural and linguistic preservation and highlights the necessity of developing technology that is accessible and relevant to a diverse global audience. The advancements made in this study pave the way for more inclusive and equitable AI applications, ensuring that the benefits of technology are shared across diverse linguistic landscapes.