WASHINGTON, United States – The Smithsonian Institution and Library of Congress are using artificial intelligence to search across collections from the American Revolutionary era, linking museum objects with newspapers, military records, letters and other historical material that have traditionally been held in separate collections.
The initiative, called Revolution Crossroads, focuses on the period from 1770 to 1810. Its purpose is to help researchers identify names, patterns and relationships that can be difficult to see when historical records are searched one collection at a time.
The project is not designed to let AI determine historical facts on its own. Smithsonian and Library of Congress researchers, curators and other experts review AI-assisted results and compare important findings with the original records.
AI connects records across institutions
Revolution Crossroads brings together several different types of historical evidence.
Smithsonian collections include military equipment, household objects, letters, maps, portraits, coins, garments and other materials from several museums. The Library of Congress contributes historic newspapers from its Chronicling America collection, while National Archives Revolutionary War pension files provide millions of pages containing information about veterans, their families and communities.
The Smithsonian says its project includes more than 10,000 records covering objects, artworks and documents from four Smithsonian museums. The Library of Congress has contributed 78,000 page-level historic newspaper records for the project, while the National Archives data includes more than 2 million pages of military and pension records.
Bringing those materials together gives researchers a way to examine relationships between physical objects and documentary evidence that may otherwise remain separated by institutional catalogs.
Named-entity recognition creates research leads
One of the project’s principal AI techniques is named-entity recognition, or NER.
The system identifies mentions of people, places, organizations, dates, events and ships in historical material. Those mentions can then be compared across different records to help researchers investigate whether apparently separate references may point to the same person, place or event.
The Smithsonian published an explanation of the process on October 2, showing how AI can move from scanned historical documents to extracted information and then to potential connections between records.
The technology is particularly useful for large collections in which manually searching every newspaper, pension file, catalog record or scanned document would require substantial time.
AI can expose connections that catalogs hide
The underlying problem is not a lack of historical material.
Instead, evidence about the same people and communities can exist in different forms and in different institutions. A newspaper might mention a person by name, while a museum collection contains an object associated with that individual and a pension file contains additional information about the person’s family or military service.
Revolution Crossroads is designed to make those relationships easier to investigate.
The Library of Congress says the project is exploring whether machine-based analysis and pattern recognition can provide new ways to examine human communities and everyday life during the Revolutionary era.
That makes the project broader than simply digitizing old records. The institutions are using AI to make already preserved historical material more searchable and to generate leads for further human research.
Human verification remains essential
The Smithsonian explicitly describes the AI-generated material as experimental.
AI-extracted text can misread handwriting, miss portions of a page or introduce errors. Named-entity systems can also confuse people who share the same name or fail to recognize that different spellings refer to the same person.
For that reason, the Smithsonian says important readings should be checked against the original images and that AI-generated identities and relationships should not automatically be treated as verified.
This creates an important boundary for the project: AI can help researchers find a possible connection, but the historical evidence must still establish whether that connection is real.
The project combines physical and documentary history
The partnership also brings together different kinds of institutional collections.
The Smithsonian holds extensive physical artifacts and museum records, while the Library of Congress provides major documentary collections such as historic newspapers. National Archives records add another layer through military and pension files.
The Smithsonian says the project is bringing these materials together so that connections become easier to see when collections are examined side by side.
That approach can be particularly useful for researching people who left relatively small traces in the historical record.
A person’s name appearing in a newspaper, for example, may become more meaningful when researchers can examine it alongside an object record or a pension document containing related information.
Technology is being tested, not treated as an authority
The project also provides a test of how AI should be used in historical research.
The Smithsonian describes Revolution Crossroads as an experiment in using emerging technology while remaining transparent about the sources, methods and limitations involved. The AI-enhanced datasets are being refined, and the institution warns that they can contain errors or omissions.
The Library of Congress similarly describes the initiative as an effort to explore how AI can augment research and discovery rather than replace human expertise.
That distinction is central to the project’s methodology. AI expands the scale at which researchers can search and compare material, while historians and curators remain responsible for evaluating what the evidence actually shows.
Revolution Crossroads extends beyond a single collection
The initiative is part of the institutions’ programming for the 250th anniversary of American independence.
The Smithsonian says Revolution Crossroads is intended to support historical research, public engagement and responsible access to Revolutionary-era materials. The project has also released AI-ready datasets through the Hugging Face platform, allowing researchers and developers to examine the data and build experimental tools around it.
The Library of Congress has continued to develop additional projects using the Revolution Crossroads data, including experiments designed to explore founding-era documents and historical newspapers through AI-powered tools.
The larger question is whether these methods can reveal connections that remain difficult to identify through conventional catalog searches.
For now, Revolution Crossroads offers a controlled example of what that approach can do: use AI to search a large and fragmented historical record, generate potential connections and then return those leads to human researchers for verification and interpretation.
Reporting Credit: Smithsonian Institution; Smithsonian Office of Digital and Innovation; Library of Congress; Revolution Crossroads project; National Museum of American History; National Postal Museum; National Museum of African American History and Culture; National Archives and Records Administration.
