Rijksmuseum Amsterdam Use Case
Access and retrieval mechanisms for archives and museums typically rely on a rich description of the collection. Most cultural heritage institutions employ professional experts to describe their collections by manually compiling metadata for each item. For large and diverse collections, the knowledge of experts from other domains is indispensable. Cultural heritage institutions therefore seek to understand whether and how they can make use of external users to produce these annotations. In close collaboration with the Rijksmuseum Amsterdam, we aim at understanding which strategies and techniques lead to high-quality annotations by (crowds of) users that are external to the museum. The first challenge of the project is to identify the niche of relevant experts and to motivate them to contribute to the annotation of artworks. As a next step, the personalization mechanisms must make sure that the experts are shown items that correspond to their expertise. The quality of the annotations and annotators are to be evaluated using trust algorithms. Finally, all these aspects must be presented in an appropriate interface.
To evaluate our hypotheses, we develop a framework designed to support crowd annotation processes, called Accurator:
National Library of The Netherlands Use Case
The National Library of The Netherlands (KB) has a large newspaper archive that is available online for the general public. It contains around 1400 Dutch newspapers titles, with news coverage from almost four centuries, ranging from the year 1618 till 1995. For each newspaper article, a scan of the article itself, a scan of the containing page, the OCR text, a PDF version and metadata records are published using a persistent URI scheme. We focus on the subset of this collection that contains papers publishing about events related to World War II. This subset contains a diverse set of titles, including potentially controversial ones. Examples include titles of existing journals that continued publishing during the war by conforming to the restrictions of the occupying German authorities, titles of journals that were published illegally because they were banned by the Germans, and pro-Nazi newspapers of which the distribution is restricted in our days because its content is considered racist, containing hate speech, holocaust denials or other content that is restricted under current Dutch law. In the SEALINCMedia project we could use this content to create a semantic network of events that comprise related media assets from other archives. The users could not only benefit from a comprehensive collection of information, but could also be given the possibility to improve the links by assessing and commenting them.