When AI Meets the Archive: Transforming the Islam West Africa Collection with Large Language Models

The Islam West Africa Collection (IWAC) illustrates how large language models can automate labor-intensive archival workflows involving optical character recognition and named-entity recognition. Drawing on predominantly French-language sources, these processes complement experimental sentiment analysis of over 12,000 newspaper articles about Islam and Muslims. Automation improves searchability and supports exploration of media portrayals but can introduce plausible errors; classification can obscure distinctions among different readings. These computational interventions constitute interpretive layers requiring historical source criticism. These workflows depend on preservation and digitization and redistribute labor toward verification. Unequal access to computing infrastructure constrains their adoption.