Hunt Institute for Botanical Documentation

The Hunt Institute is excited to announce one of the most significant updates yet to our public prototype of the Archives Text Discovery Platform: semantic search. In addition to searching for exact words and phrases, researchers can now search the transcribed contents of our digitized Archives collections by meaning.

This opens an entirely new way to discover material across the collections. Keyword searching is extremely useful when a researcher already knows the names, terms or phrases that might appear in a document. Semantic search makes it possible to begin more broadly, with an idea or research topic, and find relevant material even when the documents use very different language. In a large body of archival material, where the most important connections are often hidden in individual letters, notebooks and other documents, that can make an enormous difference.

For example, a search for "medicinal, culinary, or other traditional uses of plants" brings together material from across our collections relating to herbs and spices, plants used as food and medicine, medicinal preparations, ethnobotanical field notes and other records of human uses of plants. These connections would be difficult to uncover with a single keyword search because the relevant documents do not describe these subjects in the same terms.

Rather than generating an answer to a question, semantic search points researchers directly to the archival pages themselves. Results can be opened to view the AI-generated transcription alongside the original digitized page and to follow links back to the corresponding archival description.

Keyword and Semantic search are both available through the same interface, and each offers a different way to explore the collections. Keyword search remains ideal for finding known words and phrases, while semantic search can reveal material connected by subject or meaning. Together, they greatly expand the ways researchers can explore the contents of our digitized Archives and uncover connections that might otherwise remain buried in thousands of pages of material.

Semantic search is still experimental, and some results will inevitably be more relevant than others. We will continue refining the system as we learn from its use. For now, we are especially excited to make this new form of discovery publicly available and to see what researchers uncover with it.

We invite you to try the Archives Text Discovery Platform and let us know what you find.