We’re excited to announce that our team, in person of Prof. Dr. Bela Gipp, University of Göttingen, and Prof. Dr. Michael Granitzer, University of Passau, has been awarded a prestigious research grant of close to €600,000 by the German Research Foundation (DFG) to develop BARI—the Bias Analysis Research Infrastructure.
The project brings together researchers from the University of Göttingen, the University of Passau, the SUB Göttingen, and the GWDG. It aims to build a digital platform that enables systematic analysis of media bias in text content.
About the BARI Platform
BARI is designed to make media bias analysis more accessible across disciplines. It will offer an open, scalable infrastructure for identifying and classifying different types of bias in written content – initially focused on news articles. The system will integrate machine learning models, language data, and a community-driven taxonomy, allowing users to analyze and compare media coverage with minimal technical prerequisites.
The platform’s core components will include:
A published media bias taxonomy, ready for community feedback.
A set of automatic classifiers for distinct bias categories (e.g., selection bias, framing bias),
A user-facing web interface and a public API,
Annotation tools based on large language models (LLMs),
A feedback mechanism to involve users in improving the bias definitions and classification system.
Context and Motivation
Media bias is a widely discussed topic in research and public communication. While many existing tools either focus on specific forms of bias or require expert-level technical knowledge, BARI is intended to offer an integrated resource for broader use in fields such as journalism, media studies, linguistics, and social sciences.
The team emphasizes transparency, reproducibility, and adaptability. The system will be built following FAIR principles, and the software and data will be made available under an open license.
Collaboration and Support
The project has received endorsements from researchers in various disciplines. Letters of support reference the potential utility of BARI in DFG-funded projects, such as SFB 1472 (“Transformations of the Popular”) and SFB 1629 (“Negation as a Linguistic and Extra-Linguistic Phenomenon”). Additional use cases include work on disinformation detection, media training, and sentiment analysis.
The platform will be hosted and maintained in cooperation with GWDG, which will provide the necessary infrastructure, including GPU resources and long-term storage.
Next Steps
The BARI project will proceed through several stages:
Refinement of the media bias taxonomy,
Development and evaluation of classification models,
Implementation of the platform backend and frontend,
Collection and integration of community feedback.
The full platform is expected to be publicly accessible by the end of the funding period in late 2028.
