Vector Institute Releases NMB+ Multimodal Dataset for Ethical AI Development
The Vector Institute’s AI Engineering team has launched Newsmediabias-plus (NMB+), a new multimodal dataset designed to support the development of ethical artificial intelligence systems. Comprising approximately 90,000 news articles collected from reliable global sources between May 2023 and September 2024, the dataset integrates full-text articles with comprehensive publication details and images. It features extensive bias categorization, addressing critical issues such as gender and racial biases, ideological leanings, framing, and environmental concerns. Building on previous UnBIAS work, NMB+ incorporates both text and image analysis to help train AI models in detecting disinformation and ensuring adherence to trust and safety principles. The dataset is intended for academic researchers, NGOs, and socially focused groups aiming to analyze media trends and combat fake content. Developed under strict ethical standards with independent audits and human reviews, NMB+ ensures data accuracy and reliability. It is available on Vector’s Hugging Face page under a non-commercial license, providing a robust tool for creating transparent and accountable AI systems capable of understanding and mitigating media bias across multiple modalities.
Wire timeline
Vector Institute Releases NMB+ Multimodal Dataset for Ethical AI Development
The Vector Institute’s AI Engineering team has launched Newsmediabias-plus (NMB+), a new multimodal dataset designed to support the development of ethical artificial intelligence systems. Comprising approximately 90,000 news articles collected from reliable global sources between May 2023 and September 2024, the dataset integrates full-text articles with comprehensive publication details and images. It features extensive bias categorization, addressing critical issues such as gender and racial biases, ideological leanings, framing, and environmental concerns. Building on previous UnBIAS work, NMB+ incorporates both text and image analysis to help train AI models in detecting disinformation and ensuring adherence to trust and safety principles. The dataset is intended for academic researchers, NGOs, and socially focused groups aiming to analyze media trends and combat fake content. Developed under strict ethical standards with independent audits and human reviews, NMB+ ensures data accuracy and reliability. It is available on Vector’s Hugging Face page under a non-commercial license, providing a robust tool for creating transparent and accountable AI systems capable of understanding and mitigating media bias across multiple modalities.
Vector Institute for Artificial Intelligence