Semantic enrichment of historical archives: semi-automated topical annotation of Heritage of Nazi Persecution collections

Natural Language Processing (NLP) has recently gained interest among Holocaust researchers as NLP tools can assist with the exploration and curation of vast document collections. We present a unified NLP framework for the processing of documents related to the Heritage of Nazi Persecution, such as testimonies or diaries. The framework includes newly trained tools achieving state-of-the-art performance in the domain for Topical Text Segmentation, Semantic Tagging of text segments, Named Entity Recognition, Emotive Phrase Extraction, as well as a Machine Translation system and a Data Curation Tool, which provides a Graphical User Interface to manually pipeline and post-process the outputs of the NLP models. Furthermore, we describe the underlying architecture and data ingestion pipelines orchestrating all components on a platform that can serve as a central hub for researchers to assist them with the processing of their documents.