Our paper, Disentangling Structure and Style: Political Bias Detection in News by Inducing Document Hierarchy is accepted to the Findings of EMNLP 2023!

 

Abstract

We address an important gap in detecting political bias in news articles. Previous works that perform document classification can be influenced by the writing style of each news outlet, leading to overfitting and limited generalizability. Our approach overcomes this limitation by considering both the sentence-level semantics and the document-level rhetorical structure, resulting in a more robust and style-agnostic approach to detecting political bias in news articles. We introduce a novel multi-head hierarchical attention model that effectively encodes the structure of long documents through a diverse ensemble of attention heads. While journalism follows a formalized rhetorical structure, the writing style may vary by news outlet. We demonstrate that our method overcomes this domain dependency and outperforms previous approaches for robustness and accuracy. Further analysis and human evaluation demonstrate the ability of our model to capture common discourse structures in journalism.

 

Main Method

The proposed architecture primarily utilizes the frozen SBERT (Reimers and Gurevych, 2019) to encode the news article by sentences. Then, it calculates semantic relationships between the sentences through BiLSTM and shallow multi-head attention module, which allows us to understand the hierarchical discourse structure of the document. With this explainable property, our model precisely classify the political bias of the given news article in the low resource setting.