The Personal Writes the Political: Rendering Black Lives Legible Through the Application of Machine Learning to Anti-Apartheid Solidarity Letters

Grants and Contracts Details

Description

The Personal Writes the Political: Rendering Black Lives Legible Through the Application of Machine Learning to Anti-Apartheid Solidarity Letters Project Narrative - This project uses machine learning (ML) models to extract data from an archive of letters exchanged between international supporters of the anti-apartheid movement and the families of political detainees in South Africa, known also as ‘struggle families’. These letters were part of a massive effort to support detainees and their families led by the International Defense and Aid Fund (IDAF), the largest anti-apartheid legal defense organization. Within IDAF, support for struggle families was codenamed ‘Programme II’ and it operated under the strictest secrecy between 1966 and 1990. During those years the South African government prohibited foreign entities like IDAF from providing material support to internal activists it labeled as subversive. Volunteer letter writers in Europe, North America and the Commonwealth helped circumvent this prohibition by volunteering to become trusted intermediaries, known as ‘cutouts,’ that sent funds originating with IDAF’s donors enclosed in seemingly anodyne correspondence. This ‘crowd-sourced subterfuge’ also shielded recipients from arrest by rendering letters and aid plausibly deniable. If challenged by the authorities, a family member need only explain that the letter was from a concerned citizen abroad who heard about the plight of their loved one in the news. This project intends to utilize newly developed optical character recognition (OCR) and handwritten text recognition (HTR) methods to render images of typescript or handwritten letters in the Programme II collection (hereafter ‘P2C’) into machine readable text. Once processed, we will then apply custom ML models to perform joint entity and relation extraction (JERE) to produce triplets, meaning two or more nouns related via a verb that indicate a qualitative relationship between two categories of data. Previous Digital Humanities (DH) projects based around letters used top-level metadata to create extensive network graphs that allow scholars to track the exchange of ideas contained in the body of the letter. While the extraction of letter metadata (sender, recipient, location, date) may be automated, the analysis of the contents of the letter is done either through close reading or older text- mining methods. This project builds upon that foundational work by creating an expansive knowledge base not just from top-level metadata of the letters as circulating objects but also from the relationships between entities embedded in contents (e.g. relationships between individuals, the action of organizations on individuals, and the journeys of individuals over space and time). A knowledge base derived from entity triplets will permit us to better understand the lives, struggles and contributions of letter writers in South Africa by collecting data on relations embedded in their own words.
StatusActive
Effective start/end date7/1/222/28/25

Funding

  • American Council of Learned Societies: $25,000.00

Fingerprint

Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.