Ethical AI and Librarianship
A Resource Guide
Making and Using AI in the Library: Creating a BERT Model at the National Library of Sweden
Field
|
Description | |
---|---|---|
Title | Making and Using AI in the Library: Creating a BERT Model at the National Library of Sweden | |
Type | Articles | |
Creator | Chris Haffenden, Elena Fano, Martin Malmsten, Love Börjeson | |
Link | https://crl.acrl.org/index.php/crl/article/view/25748/33669 | |
Creation Date | 2022 | |
Last Updated Date | -- | |
Summary | This article presents a case study of how the National Library of Sweden (KB) applies AI techniques to its extensive text collections for developing a language model BERT in Swedish, referred to as KB-BERT. It outlines use cases for the KB-BERT in the context of academic libraries, such as assisting in classifying incoming digital materials, enriching metadata for the library’s digital collections, improving OCR cohesion of digitalized collections to make them more accessible to users. The conclusion section argues that national libraries are ideal sites for ethical AI development for three reasons: 1) National libraries hold publicly funded collections that can be used to train language models serving the public good. 2) National libraries can leverage their high-quality, language-specific archives to support the development of models for under-resourced languages, filling a gap left by private tech companies. 3) By developing models in-house, national libraries can examine issues of bias and data representativity and ensure responsible data documentation, and adopt ethical AI practices, providing a more accountable alternative to models e.g., BERT that are trained on un-curated internet data and risk encoding harmful biases. | |
Topic | AI and Librarianship. Ethical AI. Digital Collection. | |
Source and Link | The KBLab Blog. https://kb-labb.github.io/posts/2022-01-24-kblab-publishes-an-article-in-crl/ | |
Access | Open | |
Accessibility | Open | |
Audience | Libraries – Academic. Information Professionals. Scholars and Students. | |
Platform or Format | Document (.pdf) | |
Length | 19 pages | |
Geography | Europe/USA | |
Language | eng | |
Description Date | 06/09/2025 |