Improved Models for Media Bias Detection and Subcategorization

TL;DR

This paper introduces enhanced neural transformer models for detailed detection and classification of media bias in news articles, evaluating zero-shot and fine-tuned approaches, and leveraging synthetic data to boost performance.

Contribution

It presents a novel taxonomy of 27 bias types and demonstrates how synthetic data can improve bias detection models.

Findings

Fine-tuned models outperform zero-shot in bias classification.

Synthetic data enhances model accuracy and robustness.

Granular bias detection benefits from detailed class taxonomy.

Abstract

We present improved models for the granular detection and sub-classification news media bias in English news articles. We compare the performance of zero-shot versus fine-tuned large pre-trained neural transformer language models, explore how the level of detail of the classes affects performance on a novel taxonomy of 27 news bias-types, and demonstrate how using synthetically generated example data can be used to improve quality