Transformers vs. Traditional Models: Analyzing Media Bias in Bangladeshi News Through Sentiment Analysis

Media bias shapes public perception, making its detection a pivotal area of research, particularly in regions with diverse and influential media landscapes such as Bangladesh. This study conducts a sentiment analysis of Bangladeshi news articles, leveraging advanced Natural Language Processing (NLP) models, including Bidirectional Encoder Representations from Transformers (BERT), alongside traditional methods like Support Vector Machines (SVM) and Long Short-Term Memory networks (LSTM). Through a comparative analysis, the research underscores the superior capability of transformer-based models in identifying nuanced contextual patterns and implicit biases in news content. The findings provide critical insights into the dynamics of media coverage and highlight the potential of state-of-the-art NLP techniques to enhance objectivity and transparency in journalism, contributing significantly to the field of computational linguistics and media studies.