The Berlin-based start-up Nuzzera was selected from 685 total applicants in Europe for the prototype projects funding from the Google Digital News Initiative (DNI) for the project Spectral, as Google DNI itself announced today. Spectral is a new approach to a news recommendation algorithm that uses machine learning to combat filter bubbles.

Nuzzera’s Spectral will automatically classify articles and arrange them in spectra of text complexity and political orientation from political left to political right. In doing so, the algorithm should approach the limits of the personal reading spectrum and at the same time recommend high-quality and sophisticated articles beyond the reader’s own spectrum in order to offer new perspectives. The aim is that the readers be challenged but not overwhelmed by the difficulty presented. In contrast to conventional recommendation systems, Spectral does not try to match articles as precisely as possible, but rather as diversely as possible to the habits of the reader. Spectral is planned to help readers overcome their political bias while improving their ability to read more complex texts. The main challenge in developing the application is to find the sweet spot between reducing the effect of filter bubbles, broadening readers’ scope while at the same time helping them enjoy reading.

Schematic illustration of one of the spectral hypotheses. Here: Broadening the reading spectrum regarding the political orientation of articles.

Janine Perkuhn, co-founder of Nuzzera: “Our user research has shown that even readers who are aware of their filter bubble cannot consistently escape it in everyday life. That’s why we are working with Spectral on an algorithm that aims to broaden the perspective while providing a positive user experience - after all, reading should be fun.”

Schematic illustration of one of the spectral hypotheses. Here: Shifting the reading spectrum regarding the complexity of articles.

The aim of Spectral is to unleash the full potential of journalism by classifying, identifying and distributing articles at a level of detail that was previously not possible. Machine Learning is designed to help journalism meet every single reader at an individual level of interests and skills. Through this kind of recommendation, Spectral can help journalism to challenge readers and positively influence their reading habits. The aim is to increase the demand for high-quality content and the diversity of perspectives in order to promote a sustainable ecosystem with high-quality journalism. The complete automation of these processes makes it possible to provide such a service in a way which is economically viable.

Marco Herzog, co-founder of Nuzzera: “We have created a flood of information that the individual cannot cope with. With technologies such as machine learning, we can address each reader individually and expand their capabilities. I’m proud that Google DNI is supporting us in this and promoting Spectral.”

The Digital News Initiative (DNI) is a partnership between Google and news publishers in Europe. It aims to support high-quality journalism through technology and innovation. Google has set up a 150 million euro innovation fund, which is to be distributed over a period of three years to projects of various sizes. Companies, publishers, and individuals can apply for funding every six months, this being the fourth round of applications.

Janine Perkuhn