Libraries at University of Nebraska-Lincoln
Digital Libraries, Intelligent Data Analytics, and Augmented Description: A Demonstration Project
Date of this Version
Lorang, Elizabeth, Leen-Kiat Soh, Yi Liu, and Chulwoo Pack, "Digital Libraries, Intelligent Data Analytics, and Augmented Description: A Demonstration Project," Submitted to the Library of Congress, 10 January 2020. Rev. 15 June 2020.
From July 16-to November 8, 2019, the Aida digital libraries research team at the University of Nebraska-Lincoln collaborated with the Library of Congress on “Digital Libraries, Intelligent Data Analytics, and Augmented Description: A Demonstration Project.“ This demonstration project sought to (1) develop and investigate the viability and feasibility of textual and image-based data analytics approaches to support and facilitate discovery; (2) understand technical tools and requirements for the Library of Congress to improve access and discovery of its digital collections; and (3) enable the Library of Congress to plan for future possibilities. In pursuit of these goals, we focused our work around two areas: extracting and foregrounding visual content from Chronicling America (chroniclingamerica.loc.gov) and applying a series of image processing and machine learning methods to minimally processed manuscript collections featured in By the People (crowd.loc.gov). We undertook a series of explorations and investigated a range of issues and challenges related to machine learning and the Library’s collections.
This final report details the explorations, addresses social and technical challenges with regard to the explorations and that are critical context for the development of machine learning in the cultural heritage sector, and makes several recommendations to the Library of Congress as it plans for future possibilities. We propose two top-level recommendations. First, the Library should focus the weight of its machine learning efforts and energies on social and technical infrastructures for the development of machine learning in cultural heritage organizations, research libraries, and digital libraries. Second, we recommend that the Library invest in continued, ongoing, intentional explorations and investigations of particular machine learning applications to its collections. Both of these top-level recommendations map to the three goals of the Library’s 2019 digital strategy.
Within each top-level recommendation, we offer three more concrete, short- and medium-term recommendations. They include, under social and technical infrastructures: (1) Develop a statement of values or principles that will guide how the Library of Congress pursues the use, application, and development of machine learning for cultural heritage. (2) Create and scope a machine learning roadmap for the Library that looks both internally to the Library of Congress and its needs and goals and externally to the larger cultural heritage and other research communities. (3) Focus efforts on developing ground truth sets and benchmarking data and making these easily available. Nested under the recommendation to support ongoing explorations and investigations, we recommend that the Library: (4) Join the Library of Congress’s emergent efforts in machine learning with its existing expertise and leadership in crowdsourcing. Combine these areas as “informed crowdsourcing” as appropriate. (5) Sponsor challenges for teams to create additional metadata for digital collections in the Library of Congress. As part of these challenges, require teams to engage across a range of social and technical questions and problem areas. (6) Continue to create and support opportunities for researchers to partner in substantive ways with the Library of Congress on machine learning explorations. Each of these recommendations speak to the investigation and challenge areas identified by Thomas Padilla in Responsible Operations: Data Science, Machine Learning, and AI in Libraries.
This demonstration project—via its explorations, discussion, and recommendations—shows the potential of machine learning toward a variety of goals and use cases, and it argues that the technology itself will not be the hardest part of this work. The hardest part will be the myriad challenges to undertaking this work in ways that are socially and culturally responsible, while also upholding responsibility to make the Library of Congress’s materials available in timely and accessible ways. Fortunately, the Library of Congress is in a remarkable position to advance machine learning for cultural heritage organizations, through its size, the diversity of its collections, and its commitment to digital strategy.
Artificial Intelligence and Robotics Commons, Digital Humanities Commons, Library and Information Science Commons, Museum Studies Commons
Presentations delivered to the Library of Congress as part of this project and work-in-progress reports are available separately via the UNL Digital Commons.