Unlocking the Archives. A pipeline for scanning, transcribing, and modelling entities of archival documents into Linked Open Data

Activity: Talk or presentationAcademic

Description

In the project Golden Agents: Creative Industries and the Making of the Dutch Golden Age, heterogeneous resources on the production of the creative industries in the Dutch Golden Age from heritage institutions (e.g. Rijksmuseum, KB, RKD) are brought together as linked data. Added to this, the digitisation of the enormously rich collection of the notarial acts in the Amsterdam City Archives, will provide data on the consumption of cultural goods by the inhabitants of all layers of society in Amsterdam during the Dutch Golden Age. This archive currently plays a pioneering role in the massive digitization process of archival inventories. In the project Alle Amsterdams Akten [All Amsterdam Acts] handwritten notarial deeds are indexed on the level of inventories, documents, person namens and geolocations outside Amsterdam. At the same time, the full text of these documents is being made searchable by using the advanced Handwritten Text Recognition [HTR] tool Transkribus in the project Crowd Leert Computer Lezen [Crowd learns the computer to read] in combination with corrections of the transcriptions by volunteers. In the Golden Agents project, novel ways are explored to extract all entities of objects
that are mentioned in such notary deeds between 1578 and 1750 that are relevant to get insight in the cultural goods of Amsterdamers in the Dutch Golden Age. Once extracted and identified, almost all types of these objects can be linked to thesauri such as the Getty’s Art & Architecture Thesaurus [AAT] and reconciled with textual/linguistic references to an item in an external (authored) dataset, such as the STCN, ICONCLASS and those of the RKD. The model that is used to express the combined, enriched and created data, is compliant with major and widely used data models in the GLAM world, such as the CIDOC-CRM.
Here the full pipeline from archives to annotations is represented that comprehends the successive stages of scanning, transcribing, correcting and modelling the entities of archival documents into Linked Open Data. It provides the creation of transparent datasets that can be replicated, evaluated and used for quantitative analyses in digital humanities research.
Period03 Jun 202005 Jun 2020
Degree of RecognitionInternational

Keywords

  • entity extraction from archival documents
  • GLAM
  • Linked Open Data
  • Dutch Golden Age