Stance Classification in Afan Oromo Using Deep Learning Approaches

Stance detection is an important task in natural language processing (NLP) that seeks to determine a speaker’s or writer’s position toward a given topic. While substantial progress has been achieved for major languages, low-resource languages such as Afan Oromo remain largely underexplored. This study introduces a deep learning–based approach for stance detection in Afan Oromo, leveraging a newly collected and annotated dataset of over one million sentences from social media platforms, particularly Facebook. Three deep learning models—Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and Bidirectional LSTM (Bi-LSTM)—were implemented and evaluated. Among these, CNN achieved the highest accuracy of 85.9%, outperforming LSTM (81.4%) and Bi-LSTM (79.8%). The superior performance of CNN is attributed to its ability to capture local spatial features in text, which is particularly beneficial for short, informal social media posts. These results demonstrate the feasibility and effectiveness of deep learning techniques for stance detection in low-resource languages. Furthermore, the findings contribute to advancing language technologies for Afan Oromo and open pathways for future research in social media analysis, sentiment monitoring, and political discourse understanding in local contexts.