Towards a Digital Archive of the Atlantic Slave Trades: Unlocking the Records of the South Sea Company

Lead Research Organisation: Lancaster University
Department Name: History

Abstract

Between 1500 and 1867, 12.5 million Africans were forced aboard slave ships and transported across the Atlantic Ocean; over half a million people who survived the Middle Passage undertook another seaborne 'Final Passage' within the Americas. These world-shaping forced migrations generated an immense quantity of archival material that is scattered across the globe and written in multiple languages. But accessing, understanding, and interpreting this material is presently difficult, especially for marginalized people who lack the resources to reach archives in Europe and North America.

This project aims to democratize access to the archives of the Atlantic slave trades by employing innovative digital methods. It will create the Digital Archive of the Atlantic Slave Trades (DAAST) an open-access digital resource that will enable users to access enormous quantities of manuscript material drawn from libraries around the world. Those records will be made comprehensible by using handwritten text recognition and machine translation technologies. Research description framework approaches will enable archival materials to be semantically linked, and tied to Slave Voyages (SV), a powerful dataset of the Atlantic slave trades. The DAAST will consequently enable users to comprehend how the trade operated; the experiences of its victims; and the motivations of its perpetrators, while simultaneously grasping the trade's size, complexity, and duration.

This project lays DAAST's foundations through a focus on Britain's South Sea Company (SSC)-one of the largest slaving entities in history. Founded in London in 1711, the SSC was soon after granted the asiento, the exclusive right to carry enslaved people to the Spanish Americas. Between 1713 and 1739, the SSC dragged forty-two thousand captives from Africa to the Americas; it shifted a further forty-eight thousand people from the British Caribbean to the Spanish Americas. Because of the SSC's key role in both trades, its voluminous records- which are distributed through archives in the United Kingdom, the US, and Spain-shed uniquely detailed light on the operation of the slave trade across the vastness of the Atlantic World.

In addition to enhancing efforts to research the slave trades' history, this project promises to have broader impacts on both slave trade studies and the Digital Humanities. By creating a new platform for making the archives of the slave trade accessible and comprehensible, this project will generate a resource that will grow exponentially and underpin generations of new scholarship. Researchers in other fields grappling with the challenges of making archival material available and comprehensible will also benefit by utilizing the methodologies pioneered through this project. Cultural institutions will be also be furnished with a powerful model that they can emulate as they seek to decolonize their collections by making them accessible to a global public via digital methods.

Publications

10 25 50
 
Title New Eighteenth Century English Handwritten Text Recognition Model 
Description We are utilizing Handwritten Text Recognition software (Transkribus) to machine transcribe approximately 20,000 pages of historical manuscripts. The speed and accuracy of HTR is dependent on the "models" that the software uses to recognize and transcribe text. These models improve through "training" using machine learning and AI technology. To facilitate the transcription of our materials, we have "trained" a new model that will recognize eighteenth century English handwriting. This model will soon be available for other Transkribus users and so it will considerably accelerate the machine transcription of historical sources. 
Type Of Material Computer model/algorithm 
Year Produced 2023 
Provided To Others? No  
Impact > Increase the speed and efficiency by which historical sources can be machine transcribed. 
 
Description British Library: Digitization of South Sea Company Manuscripts 
Organisation The British Library
Country United Kingdom 
Sector Public 
PI Contribution As PI, I have worked in collaboration with the BL to digitize their holdings of the South Sea Company's papers. On May 5 2002, I visited the BL to meet Dr Alexander Lock (the BL's CI on the project) and examine the SSC's records together. We systematically worked our way through the records and identified those that we wanted to have digitized for the project. Upon my return to Lancaster, I have maintained constant communication with Dr Lock, including frequent meetings via Zoom/ MS Teams.
Collaborator Contribution Since my visit to the BL, Dr Lock and his colleagues have digitized the materials that we identified during my visit, which totals 15,000 pages of material. The team are currently producing the meta data for the material and uploading it to the British Library's digital repository. They are also sending me the image files, which totals 1.8TB of material. I have dispatched an external hard drive to Dr Lock and he will, upon receipt, load the image files on it and send it to me. Once I receive that drive, which should be before March is out, I will move the image files to Transkribus so that the project team members in the US can access them.
Impact > 15,000 pages of manuscript material digitized > Delivery of image files of the above, totaling 1.8TB of data
Start Year 2022
 
Description Clements Library (University of Michigan): Digitization of Asiento Records 
Organisation University of Michigan
Country United States 
Sector Academic/University 
PI Contribution The research team have maintained constant communication with the Clements throughout the digitization process, and to arrange delivery of the digitized material. We have also consulted with the Clements' librarians on a common meta data schema for the digitized records. Since the delivery of the digitized material, we have uploaded it to Transkribus and are now commencing the process of undertaking Handwritten Text Recognition on each of the documents.
Collaborator Contribution In February 2022, the Clement Library's archivists began the process of digitizing the "Asiento" records within the papers of William Petty, the 1st Earl of Shelburne. This collection details the South Sea Company's slave trade and constitutes approximately 1,500 manuscript pages of material. The library completed the digitization of the material in January 2023, and have delivered the resulting image files to the project team via a file sharing portal.
Impact > 1,500 pages of manuscript pages digitized and delivered to the project team as image files. > Image files uploaded to Transkribus in preparation for handwritten text recognition
Start Year 2022
 
Description Huntington Library: Digitization of Chandos Papers 
Organisation Huntington Library
Country United States 
Sector Academic/University 
PI Contribution In February and March 2022, I consulted with the Huntington Librarians on which volumes of the Duke of Chandos' papers would be digitized in support of this project. I then maintained contact with the librarians throughout the digitization process, and to arrange delivery of the materials. Since the delivery of the digitized images, myself and the other team members have been entering the digitized material into Transkribus and have begun the process of undertaking Handwritten Text Recognition on the corpus. This has enabled us to produce a new handwritten text recognition model that we can now apply to the digitized records coming from the other partner libraries. The Huntington's Research division is also contributing $25,000 towards the hosting of a conference that will publicize the project's findings in its final year.
Collaborator Contribution The librarians at the Huntington Library have digitized six volumes of the papers of the Duke of Chandos (approx. 1800 manuscript pages) which detail the Duke's investments in both the South Sea and Royal African Companies. They sent us the image files via a file sharing service, and have also made the material available to the public via the Huntington's digital library (see below).
Impact > Six volumes/ 1800 pages of historical sources digitized > Handwritten text recognition model trained for eighteenth-century English
Start Year 2022