Since September 1, 2021, PhD Jacob Orrje is pursuing a two-year, part-time project at the Center for the History of Science. The purpose is to inventory the digital needs of the Center and based on these to initiate development work.
This includes building databases and using text recognition for printed, typewritten and handwritten text to improve remote searching of the Academy of Sciences’ various archives. Other sub-projects may involve scaling up images from a lower resolution to a higher one using AI technology.
The past decade has seen rapid and exciting developments in machine learning and digital image analysis, and Orrje’s work at the Center for the History of Science is largely about investigating how these techniques can be used to make the Academy’s historical collections more accessible online. Many of our collections already exist as digital images, so as a first step he has inventoried what data is available and evaluated how we can improve access. In some cases digitizations were done a long time ago, and the quality is not sufficient for more advanced digital processing. Therefore, during the initial phase, Orrje has, experimented with how AI models, and more specifically so-called ”Enhanced Super-Resolution Generative Adversarial Networks” (ESRGAN), can be trained to increase the resolution of older digital reproductions of the Academy’s Proceedings without reducing the quality.
In parallel, he evaluates how a range of digital tools, which use machine learning to transcribe both printed and handwritten text, could be used to quickly and efficiently make important parts of the collections searchable in full text: such as the Nobel archives, the Academy’s minutes or The Academy’s Proceedings.
