Clemens Neudecker

Vast amounts of historical documents are being digitized with subsequent optical character recognition (OCR), but the quality assessment of the results is challenging for larger quantities. Ground Truth-based evaluation requires sufficient and representative data that are expensive to create. Following recent work, we investigate whether confidence scores automatically provided by text recognition systems can serve as a proxy. Based on an analysis of the relationship between word error rates and word confidence scores for several OCR engines, we find that the latter can serve as a useful indicator of OCR quality. In a second step, we explore the scalability and reliability of combining Ground Truth and confidence scores for quality assessment of text recognition in several experiments on a heterogeneous dataset comprising almost 5 million pages of historical documents from 1456–2000. The deeper analysis of the evaluation results provides insights into typical issues for OCR of historical documents, suggesting potential directions for future work.

In this article, Clemens Neudecker discusses various use cases, from text recognition to image analysis or classification to contextualization, that have already demonstrated how digitization and curation can benefit from the use of AI. But applying black-box technologies from the private sector to cultural data without fully understanding the implications also involves risks. This paper presents two examples of the adoption of AI in cultural heritage from the Staatsbibliothek zu Berlin (Berlin State Library), highlights some of the recent criticisms of data practices in the domain of AI, and offers ideas and suggestions regarding the potential role and contribution of digital curators and cultural heritage institutions for the benefit of AI.

Identification of images that are close to each other in terms of their iconographical meaning requires an applicable distance measure for text-image or image-image pairs. To obtain such a measure of distance, we finetune a group of contrastive loss based text-to-image similarity models (MS-CLIP) with respect to a large number of Iconclass visual concepts by means of natural language supervised learning. We show that there are certain Iconclass concepts that actually can be learned by the models whereas other visual concepts cannot be learned. We hypothesize that the visual concepts that can be learned more easily are intrinsically different from those that are more difficult to learn and that these qualitative differences can provide a valuable orientation for future research directions in text-to-image similarity learning.

In diesem Artikel werden digitale Kuratierungstechnologien vorgestellt, wie sie aktuell im Rahmen eines vom Bundesministerium für Bildung und Forschung (BMBF) geförderten Verbundprojekts unter der Leitung des DFKI für die Anwendung in verschiedenen Branchen und Domänen entwickelt werden. Der besondere Augenmerk liegt dabei auf deren Anwendungsszenarien in Bibliotheken – nach einer Einführung in die Herausforderungen, die sich durch die Massendigitalisierung in Bibliotheken stellen, werden exemplarisch Einsatzmöglichkeiten von Kuratierungstechnologien im Kontext von Bibliotheken vorgestellt. Anhand von Beispielen aus der bibliothekarischen Arbeit im 21. Jahrhundert, sowie der Nachnutzung digitaler Bibliotheksdaten in Wissenschaft und Kreativwirtschaft, werden die vielfältigen Möglichkeiten und Potenziale der Nutzung von Kuratierungstechnologien zur Datenaufbereitung, -anreicherung und Bereitstellung von attraktiven Diensten in Bibliotheken sowie für die Entwicklung neuartiger Forschungs- und Geschäftsfelder in deren Umfeld aufgezeigt.This article presents smart technologies for digital content curation that are currently under development in a collaborative project, coordinated by DFKI and funded by the German Federal Ministry of Education and Research. After a brief introduction into the concept of digital curation, the article goes on to explain the various technologies and their use cases across a number of sectors. In the following, we go into greater detail regarding the particular needs and requirements for digital curation technologies in the domain of libraries. Due to their large-scale digitisation activities, libraries are creating vast amounts of digital data. Technologies for digital curation can help leverage the full potential of these data for applications in scholarship, digital humanities, and the creative industries and produce new opportunities for research and industry, as illustrated by a number of examples from the Berlin State Library.

This paper provides a rare glimpse into the overall approach for the refinement, i.e. the enrichment of scanned historical newspapers with text and layout recognition, in the Europeana Newspapers project. Within three years, the project processed more than 10 million pages of historical newspapers from 12 national and major libraries to produce the largest open access and fully searchable text collection of digital historical newspapers in Europe. In this, a wide variety of legal, logistical, technical and other challenges were encountered. After introducing the background issues in newspaper digitization in Europe, the paper discusses the technical aspects of refinement in greater detail. It explains what decisions were taken in the design of the large-scale processing workflow to address these challenges, what were the results produced and what were identified as best practices.

The National Library of the Netherlands, (Koninklijke Bibliotheek, hereafter KB), has been innovating its services and organization for the past 20 years and expects to continue to do so in the future. The central question in this article is: what makes innovation work in the organization of the KB? We will focus on two use cases: the development of the recently opened Delpher portal, giving access to 30 million pages of digitized Dutch heritage, and the current development of the KB ResearchLab that gives internal and external researchers a platform for experiments. A review of innovation theory and practice (Balk 2013) provides us with a checklist of factors that determine the innovation capacity of a library, grouped in four themes: Leadership and culture, Knowledge and organizational learning, Collaboration capacity and Organizational design. By applying this innovation checklist to the use cases discussed, we hope to contribute to the body of best practice in innovation in national libraries. Finally, we will look ahead at the development of the National Digital Library of the Netherlands, integrating services for the public library community into the KB in the near future and share some potential scenarios for the future of the library landscape in the Netherlands with the audience.