Campus & Community
Differentiating apples from Apples

Students utilizing card catalogs at Widener Library, 1945.
Courtesy of Harvard University Archives
Harvard Library search tool will comprehend intent behind the terms
Over the past half-century since card catalogs transitioned online, the method of searching for library materials has remained largely unchanged. Users input keywords into a search field, the system searches for those keywords, and retrieves results.
As collections and data have expanded significantly, fine-tuning for accurate results has grown more complex. For instance, if you search a library catalog using “the history of Apple,” you will predominantly receive results related to the fruit rather than the corporation. The system grasps only the vocabulary, not the context.
A team from Harvard Library is developing a novel search tool to alter this dynamic.
By leveraging generative artificial intelligence and semantic search technologies, the upcoming Collections Explorer will transcend keyword search limitations to interpret the intent behind your queries. It will enable you to pose questions and conduct your search using natural language.
Which poems of Emily Dickinson feature handwritten notes? What materials does Harvard possess regarding the evolution of germ theory? Share insights on the Black empowerment movement in America.
Visualize asking any of these inquiries, word-for-word, on the library’s website and receiving the exact results you seek. This will soon be a reality.
Innovating a new approach to search and discovery
With over 20 million physical and digital items in various formats — from ancient manuscripts to scholarly articles, from unique maps to original poetry recordings — identifying the appropriate item for your research within Harvard’s extensive collections is a challenging task.
For the librarians and technologists at the library, the emergence of generative AI provides a chance to address this challenge while rethinking traditional library search methodologies.
Martha Whitehead, University librarian and vice president for Harvard Library, acknowledged the need for evolution in library searches, and she tasked her team to find a means to integrate AI into the search process.
“How can Harvard Library exemplify what is achievable in this new era of library discovery supported and enhanced by AI?” she inquired.

Collections Explorer is scheduled to be publicly launched in the fall.
Photo by Scott Murry
Working in collaboration with Mozilla.ai, the nonprofit’s segment focused on open-source and reliable AI, a Harvard Library team led by Stu Snydman, associate University librarian and managing director of Library Technology Services, commenced their efforts.
Harvard Library’s three-month collaboration with Mozilla.ai resulted in a prototype for a new AI-powered search tool, Collections Explorer. Developed by Library Technology Services, this tool utilizes generative AI to search across different repositories and formats. Its alpha phase, which recently completed user testing, is anticipated to launch publicly in the fall.
Engaging with Collections Explorer
The Collections Explorer is user-friendly and clear. If you’re interested in Chinese artwork at Harvard, you can phrase your query — “Does Harvard possess any artworks from China?” — as if you are conversing with a librarian.
Accompanying results from the archives of Harvard Art Museums and collections of Chinese calligraphy and painting, you will also discover illustrations of Chinese flora from a global botanical illustration collection. The results include clarifications as to why they match your inquiry.
The Explorer additionally proposes supplementary prompts, such as “Prominent Chinese paintings and sculpture at Harvard University” or “Examining the Chinese art treasures at Harvard.” Elements of serendipity and creativity are integrated into the system.
When you ask the tool “What resources does Harvard have regarding the history of germ theory development?” alongside results from the library’s collections, the system may recommend trying: “In what ways did public education campaigns of the 19th century intersect with germ theory?” or “What are the methods for antiseptic surgery?” Each new prompt presents a fresh opportunity for exploration.
“With Collections Explorer, our new discovery system crafted for the AI era, Harvard Library caters to the needs of its community and the public in inventive and groundbreaking ways,” Whitehead remarked. “We eagerly anticipate the next 50 years.”