Machine Learning sheds light on historical initials, tracing the evolution of printing styles
At the national Deutscher Historikertag in Leipzig, September 21, 2023, bachelor student Anika Merklein (Technische Universität Berlin) was awarded the Peter-Haber-Preis for her poster "Analyse gedruckter Initialen aus der Frühen Neuzeit mit neuronalen Netzen”. Anika Merklein is a member of the Max Planck Institute for the History of Science (MPIWG) research project “The Sphere. Knowledge System Evolution and the Shared Scientific Identity of Europe” which is led by BIFOLD Fellow Prof. Dr. Matteo Valleriani and she is supervised by the BIFOLD postdoctoral researcher Dr. Oliver Eberle. Her poster summarizes a machine learning research project in historical sciences, which focuses on finding initials of similar styles across different books and printers from the late Middle Ages to the early modern period. By using artificial neural networks to extract and compare different representations of style, this research helps to better understand the dissemination, transmission, and imitation of woodcuts for printing during that time period. Additionally, it can identify connections between printers and provide missing information about them. Despite not coming from a historical science discipline and competing against Ph.D. students and postdocs, Anika Merklein placed third. Congratulations!
I am Bachelor student of Computer Science at the Technische Universität Berlin. The idea of the project came up during my work as a student assistant at the Max Planck Institute for the History of Science. I had already worked with the initial images and experimented with Machine Learning procedures with them. My work colleague Dr. Hassan El-Hajj pointed me to the Call for Posters of the Digital History conference at Humboldt-Universität zu Berlin at the beginning of the year. Dr. Oliver Eberle and I then worked out the concrete procedure as a topic for my bachelor thesis. Anika Merklein
Poster: Anika Merklein, Hassan El-Hajj, Oliver Eberle und Matteo Valleriani: Analyse gedruckter Initialen aus der Frühen Neuzeit mit neuronalen Netzen. Digital History Tagung 2023 - Digitale Methoden in der geschichtswissenschaftlichen Praxis, Transformationen und ihre Konsequenzen, Berlin, 23-26 May 2023. DOI: https://doi.org/10.5281/zenodo.7950688
