One of the tools that we use in our AI-lab, is to try out AI in real contexts. These may be user experiences/services, components in internal library workflows, or services for services. Some examples are described below.


Recommendations system are the day-to-day of our digital lives. We are used to use them without realizing it, being in retail, news, or social media. Given the immense collection we handle, it is sometimes hard for our stakeholders to find relevant related content when they navigate through the main site. In order to alleviate this issue, we have built a content-based recommender system for both images and books.

Read more…


To assist with the development of the Sami bibliography, the AI-lab has developed a helper application that reads text from books to identify Sami-relevance and feed it to the bibliographers for potential addition to the bibliography.

Read more…

Named Entity Recogniton

To make using our models easier, we are working on APIs that perform common and useful tasks using our models.

One such task is Named Entity Recognition (NER), which is the task of picking out named entities – people, places, organizations etc. – from a text. For instance, “Ola Nordmann” is a named entity, and so is “Nasjonalbiblioteket”.

Read more…

SPIFF – the photo assistant

The AI-lab has made several implementations to support the workflows for photograph digitization at the NLN. These are made to study how AI can be integrated in such workflows, and to understand the potential effect.

Read more…

Front page detector

As part of the digitization process, at NB we digitized collections of newspapers. This process involves the psychical binding of several issues together to optimize scanning capacity. However, after the scanning is done, a manual process must be carried out to mark where a newspaper ends and another starts, so the digital records is properly, cataloged, stored and served. We at the Nb AI-lab have been trying to come up with solutions to better assist our colleagues in this labor intensive task. As such, we have developed a system to detect, given an single image of page, whether the image corresponds to a front-, middle-, or back-page in a given newspaper.

Read more…