Computational Research

Exploratory Semantic Search for Humanists

As part of my research, I have developed a prototype search engine designed to address the limitations of traditional keyword search for humanistic inquiry. Unlike platforms that search for keywords within whole volumes, the platform I have developed enables exploratory semantic search at the passage level.

The system works by indexing the entire corpus—currently the 200,000 volumes of Eighteenth-Century Collections Online (ECCO)—into discrete 1,000-word passages. Users construct sophisticated, high-dimensional queries, often by providing multiple examples of passages that represent a concept from their own reading. The engine then retrieves the most semantically similar passages from across the archive.

This process is iterative. Scholars can provide relevance feedback, using the best results to refine the query and progressively sharpen their conceptual target. By transforming search from a simple act of retrieval into a dialogic process, it leverages the scholar's expert knowledge in a way that is grounded in the interpretive practices of the humanities. The project serves as a proof-of-concept for a new, more powerful mode of scholarly discovery.

An example of the search results, applying a filter for female authors:

An example of an expanded search result for *the picturesque:

The development of this search engine was supported by the Lasky-Barajas Dean’s Innovation Fund, the FAS Dean’s Competitive Fund for Promising Scholarship, and an S. T. Lee Library Innovation Grant. The initial prototype was developed at the Notre Dame Institute for Advanced Study in 2017.