The goal of SETS is to improve the performance of AI in the domain of humanities research on archival sources.

Our primary focus is on handwritten archival documents that contain multiple layers of meaning: their meaning is encoded as text, but also in distinctive graphical marks, physical attributes and visual layout patterns. Meaning further accumulates over time as the documents are modified, change ownership, and are reused and reinterpreted beyond the intention of their initial creation.

Handwritten archival documents are as important as they are difficult to access and read. Archival collections measure in the millions of shelving kilometers worldwide and hold unique documents of high information density per square centimeter. These materials are crucial to the understanding of our shared past, which in turn informs choices about the future. Analyzing these materials is not simply a matter of information retrieval. It requires careful attention, mediation and interpretation.

The name SETS is rooted in our belief that leveraging the mathematical flexibility of set theory will be fundamental to achieving our goal. Set theory will allow us to represent knowledge as a dynamic graph structure where entities, relationships, and even concepts are defined within the graph itself as nested sets. In a SETS graph, meaning resides in the topological structure, which is inherently mutable and extensible, allowing the framework to evolve alongside scholarly inquiry and discovery. This architecture complements the complex and layered history of handwritten archival documents and the scholarship that makes them accessible, thus enabling the creation of new knowledge.