Uncovering the essence of diverse media biases from the semantic embedding space

Media bias significantly influences public perception and can lead to societal injustices. Existing analysis methods often rely on human effort or focus on specific bias types, failing to capture the complexities of multiple biases. This study proposes a general media bias analysis framework using embedding techniques to objectively quantify bias across various topics. Analyzing over 8 million event records and 1.2 million news articles, the findings reveal that media bias is regionally diverse and sensitive to current events. The framework reveals notable biases among US news outlets, highlighting both diverse biases on various topics and common stereotypes, such as gender bias. This framework aims to improve understanding and combat media bias for a more objective news environment.

Hong Huang, Hua Zhu, Wenshi Liu, Hua Gao, Hai Jin, Bang Liu