Abstract:In the current era of rapidly growing digital data, evaluating the political bias and factuality of news outlets has become more important for seeking reliable information online. In this work, we study the classification problem of profiling news media from the lens of political bias and factuality. Traditional profiling methods, such as Pre-trained Language Models (PLMs) and Graph Neural Networks (GNNs) have shown promising results, but they face notable challenges. PLMs focus solely on textual features, causing them to overlook the complex relationships between entities, while GNNs often struggle with media graphs containing disconnected components and insufficient labels. To address these limitations, we propose MediaGraphMind (MGM), an effective solution within a variational Expectation-Maximization (EM) framework. Instead of relying on limited neighboring nodes, MGM leverages features, structural patterns, and label information from globally similar nodes. Such a framework not only enables GNNs to capture long-range dependencies for learning expressive node representations but also enhances PLMs by integrating structural information and therefore improving the performance of both models. The extensive experiments demonstrate the effectiveness of the proposed framework and achieve new state-of-the-art results. Further, we share our repository1 which contains the dataset, code, and documentation

What problem does this paper attempt to address?

The problem that this paper attempts to solve is that in the current era of rapid growth of digital data, evaluating the political leanings and factuality of news media has become crucial for finding reliable information online. Although traditional news media profiling methods (such as pre - trained language models (PLMs) and graph neural networks (GNNs)) have shown certain effects, they also face significant challenges: 1. **Pre - trained Language Models (PLMs)**: They mainly rely on text features and ignore the complex relationships between entities. 2. **Graph Neural Networks (GNNs)**: They perform poorly when dealing with media graphs containing unconnected components and insufficient labels. To solve these problems, the authors propose MediaGraphMind (MGM), an effective solution based on the variational expectation - maximization (EM) framework. By leveraging the features, structural patterns, and label information of globally similar nodes, MGM not only enables GNNs to capture long - distance dependencies to learn more expressive node representations, but also enhances the performance of PLMs by integrating structural information. ### Specific Problem Description 1. **Evaluation of Political Leanings and Factuality**: With the development of the Internet, the spread of false information and misleading content has accelerated, and it has become particularly important to quickly and accurately evaluate the political leanings and factuality of news media. 2. **Limitations of Traditional Methods**: - **PLMs**: They only focus on text features and cannot capture the complex relationships between entities. - **GNNs**: They have difficulty dealing with unconnected components and sparse labels, which limits their performance in classification tasks. ### Solution MGM solves the above problems in the following ways: - **Utilization of Global Information**: By using the information of globally similar nodes, it enhances the ability of GNNs to capture long - distance dependencies. - **Combination of PLMs and GNNs**: It combines structural information with text features to improve overall performance. - **Variational EM Framework**: It stores pre - computed node representations through an external memory module to reduce computational costs and improve efficiency. ### Experimental Results The experimental results show that MGM significantly improves the performance of the baseline GNN model, achieving a 10% improvement on multiple evaluation metrics, and reaching a new state - of - the - art level on the ACL - 2020 and EMNLP - 2018 datasets. In conclusion, this paper aims to solve the key challenges in evaluating the political leanings and factuality of news media by proposing the MGM model, thereby providing a more reliable method for identifying news sources.

MGM: Global Understanding of Audience Overlap Graphs for Predicting the Factuality and the Bias of News Media

GREENER: Graph Neural Networks for News Media Profiling

ChatGPT v.s. Media Bias: A Comparative Study of GPT-3.5 and Fine-tuned Language Models

Mapping the Media Landscape: Predicting Factual Reporting and Political Bias Through Web Interactions

MGMFN: Multi-graph and MLP-mixer fusion network for Chinese social network sentiment classification

Understanding Political Polarization via Jointly Modeling Users, Connections and Multimodal Contents on Heterogeneous Graphs

Quantifying Generative Media Bias with a Corpus of Real-world and Generated News Articles

KGAP: Knowledge Graph Augmented Political Perspective Detection in News Media

People on Media: Jointly Identifying Credible News and Trustworthy Citizen Journalists in Online Communities

A Survey on Predicting the Factuality and the Bias of News Media

Machine-Learning media bias

MMGA: Multimodal Learning with Graph Alignment

GPT-4V(ision) as A Social Media Analysis Engine

Sentence-level Media Bias Analysis with Event Relation Graph

Multilingual Coarse Political Stance Classification of Media. The Editorial Line of a ChatGPT and Bard Newspaper

Predicting Factuality of Reporting and Bias of News Media Sources

MGMNet: Mutual-Guidance Mechanism for Joint Classification of Multisource Remote Sensing Data

MediaRank: Computational Ranking of Online News Sources

GMRD

SENTINET: A DEEP SENTIMENT ANALYSIS NETWORK FOR POLITICAL MEDIA BIAS DETECTION

Multilevel Graph Matching Networks for Deep Graph Similarity Learning