Advances in Machine Learning for the Automated Analysis of Manuscripts

Abstract

Due to the increasing digitization initiatives of archives/museums/…, the number of digitally available documents, including historical manuscripts, is growing, facilitating access for experts and humanities scholars. Advances in machine learning, particularly in the area of deep learning, enable automated recognition of handwritten text, layout analysis, writer retrieval, and research-specific tasks. Applications such as handwritten text recognition allow on the one hand also ordinary persons to read historical scripts, and on the other hand a text search in a large text corpus for experts without the necessity to read everything. Additionally, writer retrieval enables to cluster documents based on the scribe. Manuscripts and their metadata can also be modeled in knowledge bases and knowledge representation learning can be used to produce interpretations automatically. Several state- of-the-art ML methods in the context of research projects will be presented to give an overview of the feasibility of automated methods.