The research focus of Phillip Ströbel is on the Digital Humanities, where he is mainly concerned with text mining methods which help to make historical documents accessible to researchers from other disciplines. These methods include text classification, and topic modeling, especially.
Besides the enrollment as PhD student at the Institute of Computational Linguistics at the University of Zurich, Phillip is also engaged in the impresso project, which aims at developing new access points to historical newspapers for historians. The impresso project is a Sinergia project funded by the SNSF and involves researchers from the Digital Humanities Lab in Lausanne, the Centre for Contemporary and Digital History in Luxembourg, and the Institute of Computational Linguistics in Zurich.
Phillip’s project Transfer Learning and Text Data at the ETH Library Lab comprises the development of new access methods to the journals of e-periodica. More concretely, he examines to what extent transfer learning can be used to build semantic clusters of journal articles. To this end, he will use information from neural text classification systems to cluster the articles. A visualisation of these clusters then creates a novel method to access the texts which, combined with traditional keyword search, is supposed to assist users in finding relevant content.