NewsMediaBias-Plus: Multimodal News Analysis for Media Bias Detection

A multimodal dataset for analyzing media bias and disinformation, combining textual and visual data from news articles. It is hosted by the Vector Institute on Hugging Face and was last updated in February 2025. The dataset is categorized for text-classification and multimodal tasks and is sized between 10,000 and 100,000 entries.

Use Cases

• Training multimodal classifiers for media bias detection based on combined text and image content.
• Analyzing visual framing and textual slant in news articles for disinformation research.
• Benchmarking models on the specific task of text-classification with visual context.

Strengths

• Dataset is explicitly designed for the analysis of media bias and disinformation, providing a clear research focus.
• Combines both textual and visual data from news articles, enabling multimodal analysis.
• Hosted by the Vector Institute, a recognized AI research organization.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The license is listed as 'cc-by-nc-2.0', which restricts commercial use.

Provenance

Source

Vector Institute via Hugging Face.

Collection Method

Likely collected from online news articles, but the specific gathering method is not detailed.

Freshness

Last updated 2025-02-18 22:22:45.