Lifting the treasure of an archive — the Furche Navigator as a digital time machine.

The task was to develop a new concept for digital reading at Die Furche. In the conception work we looked for ways to achieve a real digital transformation of the medium. For us, digital transformation means something other than just doing things on a computer — it means discovering possibilities that only digital ways of working make possible.

It was clear that one of the Furche's most important assets is its archive. In many conversations with the editorial team and management we explored what could be done with it. To feel the value first-hand, our management actually sat down in the Austrian National Library and read old issues. Wonderful articles published as early as the 1940s and 1950s — gathering dust in printed form, in two places only: in Die Furche itself and in the National Library. There are certainly private collections, but we had no access to those.

Technically the project was built on the research outcomes of our STRG.behave project — and is still operated with it today. STRG.behave measures and vectorises every new article and assembles a dynamic timeline of archive articles that match. The editorial team has an interface to add or exclude articles manually. We advise against interfering with the semantic logic — manual edits are weighted very heavily in the self-learning system and significantly alter results. The timeline is built by a neural network. The whole system runs on Google Cloud.

STRG.behave grew out of a research project asking how media can measure their users' real interests without depending on third-party providers. The idea is that most of that third-party data is of extremely low quality and rarely makes sense. No neural network, however good, can fix that. To make this work we had to implement semantic analyses that go far beyond keyword finding — analysing sentiment and computing the actual meaning of long-form stories.