Enhancing Sentiment Analysis in Amharic: Leveraging Transformer-Based Language Model for Low-Resource African Languages

One of the most extensively researched applications in natural language processing (NLP) is sentiment analysis. While the majority of the study focuses on high-resource languages (e.g., English), this research will focus on low-resource African languages, namely Amharic. The annotated tweets in Amharic have a significant number of code-mixed tweets. The curated datasets necessary to build complex AI applications are not available for the majority of African languages. To optimize the use of such datasets, research is needed to determine the viability of present NLP procedures as well as the development of novel techniuqes. This paper details our initiative to establish a sentiment analysis which is a computational technique used to detect positive, negative, and neutral sentiments in Amharic language tweets. Our research utilized the AfriSenti-SemEval 2023 Shared Task 12 Twitter datasets, which include human-annotated data in Amharic. For this task, we employed the XLM-R, a transformer based pre-trained language model. This training was aimed at enhancing the model's ability to accurately classify sentiments in tweets written in Amharic. The results demonstrate that our model which is XLM-R, trained on AfriSenti-SemEval Shared Task 12 datasets, produced an Fl score of 79.02% for the Amharic language on the sentiment analysis test. In AfriSenti-SemEval 2023 shared task 12 (Task A) and this research achieved the top 1 ranking, based on the Fl scores officially released by the SemEval organizers.