HathiTrust wins $5m Mellon grant for AI-powered book discovery project
HathiTrust has received a $5 million grant from the Mellon Foundation to develop AI-powered services aimed at improving discovery and research across its collection of 19 million digitised volumes.
The five-year project, Transparent Books: AI Infrastructure for Discovery and the Public Interest, will see HathiTrust work with Northeastern University Library and nonprofit organisation Authors Alliance to explore responsible, equitable and transparent uses of artificial intelligence in large-scale digital collections.
The initiative is intended to create shared, academy-led infrastructure that can support more sophisticated discovery and research while providing lawful, noncommercial access to digitised material for academic and public-interest uses.
A key focus will be using AI to improve the descriptive metadata and text quality of material within the HathiTrust corpus. This could include better identification of subjects and languages and the creation of summaries, alongside interactive search tools designed to help users explore individual books and collections of works.
The partners will also look at ways of providing verified and transparent sources of book text and images at scale for academic and noncommercial purposes. Potential applications include text and data mining, computational analysis and the development of specialised language models.
Alongside the technical work, the project will develop shared principles covering areas including legal frameworks, research integrity, accountability and the environmental impact of AI. The partners say these principles will be designed to ensure that academic values and the public interest remain central to decisions about how the technology is deployed.
The project will also examine how services developed through the initiative can become financially sustainable while continuing to support academic and noncommercial research.
HathiTrust said the work would take account of the interests of publishers, authors, scholars and members of the public as it considers how AI can make digitised books easier to discover and use.
Mike Furlough, HathiTrust’s Executive Director, said, “For decades, libraries have sustained their values in the face of rapid and powerful advances in technology. Libraries created HathiTrust to ensure that the results of mass digitisation would address the unique needs of researchers and the academy. This project continues that longstanding effort to collaboratively make the scholarly record more accessible and usable for today and into the future.”