Unraveling Media Perspectives: A Comprehensive Methodology Combining Large Language Models, Topic Modeling, Sentiment Analysis, and Ontology Learning to Analyse Media Bias

TL;DR

The paper tackles scalable analysis of media bias in political news by introducing a methodological framework that fuses hierarchical topic modeling, sentiment analysis, entity analysis, and ontology learning to detect event selection and labeling biases across sources. It leverages BERTopic with SBERT embeddings for topic structure, RoBERTa and spaCy-based sentiment methods, bert-base-NER for entities, and GPT-4-driven ontology learning to create core reference ontologies, all evaluated through three case studies. The authors define a Media Bias Spectrum to visualize deviations in topic publishing rates and sentiment across newspapers, and demonstrate how ontology-derived comparisons illuminate omissions and commitments. While showing promise for tool-assisted media bias analysis, the work acknowledges limitations related to data scope (RSS-based datasets), topic modeling noise, and ontology duplication, proposing directions for time-aware analyses, explainable-AI enhancements, and automated ontology consolidation to improve robustness and usability.

Abstract

Biased news reporting poses a significant threat to informed decision-making and the functioning of democracies. This study introduces a novel methodology for scalable, minimally biased analysis of media bias in political news. The proposed approach examines event selection, labeling, word choice, and commission and omission biases across news sources by leveraging natural language processing techniques, including hierarchical topic modeling, sentiment analysis, and ontology learning with large language models. Through three case studies related to current political events, we demonstrate the methodology's effectiveness in identifying biases across news sources at various levels of granularity. This work represents a significant step towards scalable, minimally biased media bias analysis, laying the groundwork for tools to help news consumers navigate an increasingly complex media landscape.

Figures (22)

• Figure 1: Forms of media bias in and out of scope of this work. Based on Park2009 and Hamborg2019
• Figure 2: Overview of our proposed methodology and mapping to analyzed forms of media bias
• Figure 3: Classification of ontologies. Based on Roussey2011.
• Figure 4: Prompt template used for extracting ontology elements
• Figure 5: Example for visualization of deviation of percentages from mean