“Access to information is a fundamental right”
Artificial intelligence and archives
“I am a historian and used many audiovisual media from countless archives for my PhD,” says Cecilia Maas. “That’s when I noticed that tapping into and thus accessing them is very difficult. Listening to 4,000 original recordings to find just a few important parts was simply impossible.”
Together with the business information scientist Felix Mertineit and a handful of like-minded people, the young woman has now found a solution. Together, they will launch their joint start-up aureka already this year. “We use artificial intelligence (AI) to transcribe many different types of audio recordings and to support our customers in tapping into archives.” It works something like this: Let’s say an archive uploads recordings of radio broadcasts from the 1960s or 1970s to aureka. Artificial intelligence then turns the spoken word of these broadcasts into text. According to Cecilia Maas, this is called a “dirty transcript”. While voice recognition has made significant progress in recent years, it’s far from perfect. If required, users can contribute to tidying it up. However, for many it is usually enough to find a needle in the haystack that is actually valuable to their own research. AI can also help here by automatically creating an overview of the content, generating a list of people who were talked about or of places that were mentioned in the radio programme. “This is a great advantage to our users and helps them to decide where to look for the content they need,” says Maas.
With her idea, she took part in a competition at Free University of Berlin. This brought her into contact with Profund Innovation, the university’s start-up support service, who motivated her to apply for funding and transfer her project from the university into a business. “Many researchers develop their own software but typically only for a specific project. Once the funding has expired, the software is not maintained,” she says. “We wanted to create something that is useful to many and that develops over time. We want to transfer the progress of new technologies into the knowledge creation and cultural heritage sector.”
Potential customers for the founding team are wherever people sift through large troves of audiovisual data. Archives, libraries, and museums, for example, but also radio and TV stations. In her opinion, even people working in policymaking could benefit from this, for example, if speakers no longer have to listen through hours of recordings of parliamentary sessions. Of course, science and research are at the top of the founder’s wish list. “Anyone conducting a lot of interviews for their thesis, for example, would be in good hands with our service,” she says. “Of course, it would be great if universities would make licenses available to their students and staff.”
A social dimension, Cecilia stresses, has been an integral part of aureka from the very beginning. “We believe that access to information is a fundamental right,” she says. “This is primarily guaranteed by institutions, but technology can play an important role. With our company and our product, we want to support institutions with the challenges created when accessing information.” Made up of humanities scholars and computer scientists, the team is currently in the middle of beta testing. The official launch is scheduled for August. By then, the business founding process should also be complete, transforming a research transfer project into a genuine start-up.
Kai Dürfeld for POTENZIAL