Artificial intelligence in the KB
The KB uses artificial intelligence, also known as AI. These are systems that can improve their own performance by learning from experience. This allows the technology to perform tasks faster or to recognise patterns, such as a system that suggests keywords when you add a new book to a catalogue. While customers do not always see this technology, they sometimes do notice the results. In this example, these results are the keywords they see in the catalogue.
At the KB, we also use AI for recognising texts and images. This means that texts in Delpher and DBNL are becoming increasingly readable and searchable.
Who is AI of interest to?
How does AI change the services provided by libraries? How does AI change the way we work? What can the library contribute to AI development? And how can the library help deploy AI responsibly? The KB always keeps those questions in mind.
The KB uses AI for its customers, researchers, network and its own staff. We use it to optimise our search technology, which is important because the KB manages millions of digital books, newspapers, magazines and more. Searching through all that information used to be complicated but, thanks to AI, customers can now quickly get the answers they are looking for.
We also check whether we can recommend book titles that you as a customer might like to read. We do this with recommenders, whereby a system analyses the content of books so that it can make comparisons with what you have read before.
We also use AI in our collaborations with researchers and research institutions. Twice a year, a researcher comes to the KB to conduct research. We try to answer their research question by using the latest technology and put the experience we gain back into making our datasets and services even more accessible. A good example of this is the use of Handwritten Text Recognition (HTR) to convert handwritten and old printed works into digital text.
The KB does not do all this alone. We collaborate with, among others, the Netherlands Institute for Sound and Vision, National Archives, Huygens ING, NBD Biblion, universities and libraries. We are also represented in network associations such as NDE, CLARIAH, Cultural AI Lab and NL AIC.
Why the KB uses artificial intelligence
AI is not an end in itself, but an important tool. It helps the KB develop and offer services that match the needs and expectations of our users.
We can also support researchers in answering research questions. We do this, for example, in the researcher-in-residence programme or in other collaborations with research and heritage institutions.
To give the populace greater insight into the opportunities and risks posed by AI, the KB joined forces with libraries and other parties to launch the AI Parade. This is a travelling event that lets people have their say on AI and its impact on society.
Finally, there’s the question of how AI affects the work of librarians. How can it help improve work processes in the KB? Thanks to AI, tasks can be done faster, better and more efficiently. As a result, less input from people is needed. In all this, we have to take into account the ethical aspects and privacy. That is why we have drawn up a guide of seven AI principles. For example, we do not want to use applications that actively seek to manipulate people's behaviour or thinking. We also make sure that people stay in charge and are the ones making the decisions.
Choices and applications made by the KB
The KB has already achieved great results by working with AI. We have also written down where we want to go with this. More is possible and achievable every day, which is why we continue to work with different target groups and partners.
Find more information on the KB's results and the policy on AI here:
- De digitale spoelkeuken: a type of digital dishwashing method used to check whether a website meets 5 values, compiled by Public Spaces.
- The Netherlands AI Coalition (Culture and Media Workgroup). The art of AI for all. The connective power of culture and media (2022). A plea for more research on AI in the Culture and Media sector.
- National Digital Heritage Strategy 2021-2024 (2021). Dutch Digital Heritage Network & the Ministry of Education, Culture and Science About the best possible access to our digital heritage.
- Michel de Gruijter et al., Krant & foto’s verbonden (Newspaper & photos linked). An exploration into the use of artificial intelligence to link heritage collections (2022). Regarding automatically linking newspaper photos and press photos.
- Dorien Haagsma, Onderzoekassistentie van de Demosaurus (2021) (Research assistance of the Demosaurus). Suggestions for keywords when entering book titles.
- Seyran Khademi, DePTH (DeeP Training on History) (2021). On recognising objects in illustrations in children's books.
- Jan Willem van Wessel, AI in libraries: seven principles (2020) Seven values with which the KB aims to test AI applications.
CuratorBot
In collaboration with TU Delft, the KB developed the 'CuratorBot' - an AI chatbot that will talk to you about the Visboeck (Fish Book) by Adriaen Coenen, a fish auctioneer from Scheveningen. This late 16th-century manuscript is particularly popular with visitors. It contains all kinds of illustrations of marine creatures from fish to sea monsters but, to an untrained eye, the text is not readable and many illustrations raise questions. What is this? What does this have to do with the sea? Why did Adriaen Coenen add this to his book?
The CuratorBot is 'fed' all kinds of information sources about the Visboeck, from the web to publications, and knowledge from our own collections specialists. It can therefore tell you all about unicorns, manatees and barnacle geese. The CuratorBot gives us insight into the usability of AI chatbots for heritage institutions. Which questions does the CuratorBot struggle with? What do visitors to the KB want to know? The CuratorBot was available for live testing in the KB reception hall until June 2023.
