The goal of ERPAD („Extraktion räumlicher Daten aus historischen Dokumenten“) is an intelligent handling of historical documents and written sources of information. In German and European archives large amounts of historical documents are present containing information with extremely high value. Connecting these historic documents is particularly difficult without having an understanding of the historic context and only possible with expert knowledge. Due to a deficit of experts in different historic fields much information cannot be collected and accordingly not effectively analyzed.

As a result, within the project an automated method for the determination and connection of historic documents based on geographical, temporal and thematic relationships will be developed. Therefore, recent developments from the field of machine learning (deep learning) and object character recognition using neural networks and GPU-based calculations (CUDA/OpenCL) will be applied. The resulting datasets can afterwards be manipulated, combined and analyzed by means of current technologies.