< Back to previous page

Project

Towards a KBR Data Science Lab using Knowledge and Data-Driven Artificial Intelligence for Digitization, Informatics, Iconography and Communication in the Cultural Heritage Sector (FEDTWIN4)

The objective of the KBR Data Science Lab (DSL) is to establish a long-term collaboration between the Royal Library of Belgium (KBR) and the Digital Mathematics (DIMA) research group of the Department of Mathematics and Data Science (WIDS) of the Vrije Universiteit Brussel (VUB) on research and development in data science, in particular Artificial Intelligence (AI), applied toRT-1: Data Science for library informatics and iconography;RT-2: Data Science for automation/optimization of digitization workflows in the cultural heritage sector;to leverage the value of the vast amount of digitized or born-digital collections in the cultural heritage sector (CHS).Dedicated government investments in the digitization of the collections of the Federal Scientific Institutions in Belgium provides researchers and the public valuable resources of heritage (historical but also artistic) collections. As a flagship project, the KBR’s digitization of the Belgian Press 1830-1950 has produced rich resources in terms of document images of historical newspapers. This has already led to the successful joint KBR and DIMA Belspo-project ADOCHS (see Section 3.8).In the near future, the KBR’s systematic digitization of its collection of about 4.500 medieval codices and its collection of about 1 million prints and drawings is producing a huge range of resources in terms of iconography which constitutes a strong potential for further research investigations within the KBR DSL and disclosure of the material.Systematic research efforts for exploiting the potential of these resources are however still missing. On one hand, information locked in the digital collections is yet to be extracted, organized and manipulated in meaningful ways. For example, it is difficult to query the digital collections semantically (e.g. automatically retrieve all issues of ‘Le Patriote’ which contain stories on criminal activities). On the other hand, new methods have to be developed to automate/improve existing digitization workflows. One such example involves image quality assessment (IQA), where existing IQA models are content-specific, therefore fail to provide reliable supervision to digitization processes containing different types of images or images with mixed content (e.g. text and natural scene photos).The objective of the KBR DSL is to address both of these challenges in a systematic way, employing knowledge and data-driven AI. This links both KBR and DIMA in a mutually beneficial way. On one hand, the two research tracks of the KBR DSL are aligned with the development and progress of the library. DIMA on the other hand, will also benefit greatly from KBR DSL to further develop its expertise in data science, and to make contributions not only to developing new tools for CHS, but also to the academic field.The KBR DSL will be a continuation of the strong collaboration between KBR and DIMA (cfr. DSPH algorithm, Section 3.5) and will be developed over a period of 10 years. The progress of the lab will be put in dialogue with stakeholders (e.g. the KBR Senior Management Team and the digitization department), domain experts (e.g. other CHS & data scientists) and the general public.
Date:1 Nov 2021 →  Today
Keywords:Data science, data-driven & knowlegde-driven AI, information harvesting, iconography in libraries, digitization workflows in libraries
Disciplines:Computer science