Service Incident: New DOI registrations are working again. Re-registration of failed DOI registrations (~500) are still affected by the service incident at DataCite (our DOI registration agency).
Published August 23, 2019 | Version v1
Report Open

Exploration possibilities Automated Generation of Metadata

Description

How can we use smart technologies to simplify the description of publications? This is one of the research questions on the KB Research Agenda that we intend to answer in the coming years.

At present, the Koninklijke Bibliotheek (KB), National Library of the Netherlands assigns the resource description (also referred to as “generation of metadata” or “creating bibliographic records”) by hand and partially by adopting the data we acquire through other channels. In part due to the growth in electronically generated material (“born digital”) and the growth in website storage, we expect a growing need for the retention of increasing numbers of publications in the coming years. For this reason we explore to optimise the options for manual description of

publications. Two current developments offer opportunities: the growing volume of publications available in entirely electronic format, and the fact that smart technologies, for example artificial intelligence (AI) applications such as machine learning, are expanding the possibilities for having electronic texts be interpreted automatically (by computer).

In this white paper we describe the state of our initial explorations of the options for automated generation of metadata of publications. We first present an overview of the ways in which organizations and enterprises outside the KB are using smart technologies to analyse and describe sources such as news articles, books, television broadcasts and photographs. We then discuss how we within the National Library are currently describing titles in order to indicate where in the process we see opportunities for automated attachment of metadata. In the third chapter we discuss the result of our own experiments with the automated assignment of keywords to publications. We conclude with the lessons we have learned so far and discuss our next steps.

A Dutch version of this whitepaper is available at https://zenodo.org/record/3373316

Files

KB_Whitepaper_Verkenning automatisch metadateren_ENG_HR-PRINT.pdf