Details
| Original language | English |
|---|---|
| Article number | 112 |
| Journal | Discover Artificial Intelligence |
| Volume | 5 |
| Issue number | 1 |
| Publication status | Published - 18 Jun 2025 |
Abstract
This work investigates the application of machine learning for the analysis of video journalism to get insights into media bias in the German video journalism landscape. For this purpose, a custom dataset made up of subtitles from video data of major German news outlets ranging across the political spectrum was created. Media bias was assessed utilizing mention and sentiment analysis with respect to the major political parties in Germany. Sentiment analysis, performed using german-news-sentiment-bert, revealed significant differences in the reporting sentiment between media outlets. The German public broadcast outlet ARD was found to report with neutral sentiment less frequently than the mean, instead using negative sentiment significantly more often, especially while mentioning parties on the political edges. Mention analysis revealed that politicians get mentioned more often when in governing coalitions and, furthermore, it revealed a slight association between the assumed political ideology of media outlets and how frequently they report on political parties with a similar ideology, i.e., right-leaning outlets mention left-leaning parties and politicians less frequently and vice versa.
Keywords
- Bias, Journalism, Machine learning, Media, Mention analysis, Sentiment analysis
ASJC Scopus subject areas
- Computer Science(all)
- Artificial Intelligence
- Computer Science(all)
- Information Systems
- Computer Science(all)
- Computer Vision and Pattern Recognition
- Computer Science(all)
- Human-Computer Interaction
Cite this
- Standard
- Harvard
- Apa
- Vancouver
- BibTeX
- RIS
In: Discover Artificial Intelligence, Vol. 5, No. 1, 112, 18.06.2025.
Research output: Contribution to journal › Article › Research › peer review
}
TY - JOUR
T1 - Towards Automatic Bias Analysis in Multimedia Journalism
AU - Hinrichs, Reemt
AU - Steffen, Hauke
AU - Avetisyan, Hayastan
AU - Broneske, David
AU - Ostermann, Jörn
N1 - Publisher Copyright: © The Author(s) 2025.
PY - 2025/6/18
Y1 - 2025/6/18
N2 - This work investigates the application of machine learning for the analysis of video journalism to get insights into media bias in the German video journalism landscape. For this purpose, a custom dataset made up of subtitles from video data of major German news outlets ranging across the political spectrum was created. Media bias was assessed utilizing mention and sentiment analysis with respect to the major political parties in Germany. Sentiment analysis, performed using german-news-sentiment-bert, revealed significant differences in the reporting sentiment between media outlets. The German public broadcast outlet ARD was found to report with neutral sentiment less frequently than the mean, instead using negative sentiment significantly more often, especially while mentioning parties on the political edges. Mention analysis revealed that politicians get mentioned more often when in governing coalitions and, furthermore, it revealed a slight association between the assumed political ideology of media outlets and how frequently they report on political parties with a similar ideology, i.e., right-leaning outlets mention left-leaning parties and politicians less frequently and vice versa.
AB - This work investigates the application of machine learning for the analysis of video journalism to get insights into media bias in the German video journalism landscape. For this purpose, a custom dataset made up of subtitles from video data of major German news outlets ranging across the political spectrum was created. Media bias was assessed utilizing mention and sentiment analysis with respect to the major political parties in Germany. Sentiment analysis, performed using german-news-sentiment-bert, revealed significant differences in the reporting sentiment between media outlets. The German public broadcast outlet ARD was found to report with neutral sentiment less frequently than the mean, instead using negative sentiment significantly more often, especially while mentioning parties on the political edges. Mention analysis revealed that politicians get mentioned more often when in governing coalitions and, furthermore, it revealed a slight association between the assumed political ideology of media outlets and how frequently they report on political parties with a similar ideology, i.e., right-leaning outlets mention left-leaning parties and politicians less frequently and vice versa.
KW - Bias
KW - Journalism
KW - Machine learning
KW - Media
KW - Mention analysis
KW - Sentiment analysis
UR - http://www.scopus.com/inward/record.url?scp=105008549245&partnerID=8YFLogxK
U2 - 10.1007/s44163-025-00362-1
DO - 10.1007/s44163-025-00362-1
M3 - Article
AN - SCOPUS:105008549245
VL - 5
JO - Discover Artificial Intelligence
JF - Discover Artificial Intelligence
IS - 1
M1 - 112
ER -