Detection of Interconnections, Factual Contradictions, Bias, and Coordinated campaigns in Web Media
Mass media shapes our opinions, affects our beliefs, and influences our social behaviour. Internet transformed the mass media by providing news readers and news producers with a real-time access to ever-growing amounts of multimedia information. However, it also poses significant challenges for the society. Popularity of social media (blogs and social networks) takes the monopoly on news distribution away from traditional media (newspapers and television). At the same time, it allows for quick distribution of false or unsubstantiated claims and biased news stories as well as for easier and imperceptible propagation of ideas by paid or ideologically driven groups of people.
Social behaviour is influenced by opinions. In order to develop well-founded opinions, we need to be exposed to balanced credible information, we need to be aware of information manipulation and media bias. In this project, we aim at developing computational tools for media analysis available for the community, which can help media scientists, journalists, and news readers to navigate in the ocean of news stories and opinions.
See project presentation.
Specifically, we work on developing the following tools.
Knowing the source of a media item, i.e. looking for information sources, citations, and modifications is crucial for evaluating its credibility. For each news item (image, text, link), we will automatically build a "life timeline" showing who was the first to publish this item, who cited and recited it, who modified it and how.
Prototype demo of News Life Cycle Detector is available.
News articles may contain incorrect or unmotivated claims. We will develop a service that will detect statements in a piece of text and find documents on the web that contain contradicting or supporting statements.
Media bias is inherent to the news production process. For news readers, it is difficult to analyse news articles critically and detect the bias. Most of the news readers live in isolated information bubbles. At the same time, being exposed to diverse information and being aware of the media bias is important for developing balanced views and correcting inaccurate beliefs. We will develop methods for automatically creating bias profiles of news articles. The bias profiles will be used to cluster together articles having a similar perspective, which will allow us to present diverse articles coming from different perspective clusters to news readers.
It is natural to assume that the most trusted claims are those that most sources agree with. However, in the current media situation, small groups of paid or ideologically driven people can significantly change the media landscape by pushing certain opinions. We will develop methods for automatically identifying cases of manufactured spread of opinions.