Skip to end of metadata
Go to start of metadata

Finnish Academy organized a seminar with the topic "Deep learning and the humanities" where the idea was to bring together researchers from the Digihum-program, as they progress along with their research targets.

As a keynote there was Professor Roger K Moore, whose slides can be found from the academy's event page .  He had a very lively presentation about the fast technology evolution with regard the speech recognition and he also talked about the difficulties, that can occur when a machine tries to understand a human, how many points of misunderstanding there can be. Some of his articles listed.  One way forward, which he mentioned would be iterative translation, where machine does translation as the talking progresses, instead of waiting to the end. This enables clarifying as e.g. translation progresses.


Also Comhis-program was presented by the work package leaders. Dr. Kimmo Kettunen told about the article extraction with the digitized newspapers, Mikko Tolonen about how new information can be revealed from the metrics, and Hannu Tolonen, how similar paragraphs or news items have, back in the day, flown freely from one newspaper to another for which there were potential tools to experiment on and to check how they work with this quite complex corpus of the newspapers and journals.


Also other Digihum-projects were presented and finally also the international 'Digging into data' challenge winners from Finland got also to present their topic. It seemed that there is lots more research questions to tackle, which the existing projects start, but which can be made even further within that umbrella.  The joint projects enable universities with different focus areas to collaborate and find new solutions, which creates new benefits on both sides. For example, quite concretely, project called Oceanic Exchanges (OcEx), where University of Turku is a member, illustrates the global connectedness by utilizing the digitized newspapers from different collections.



P.S. Do remember that the open data distribution packages from the digitized newspapers  are available via .



  • No labels