Publikation

Titel:

Towards Augmenting Metadata Management by Machine Learning

AutorInnen:

Christopher Julian Kern
Thomas Schäffer
Dirk Stelzer

Kategorie:

Beiträge in referierten Konferenzbänden

erschienen in:

INFORMATIK 2021. Gesellschaft für Informatik, Bonn.

Abstract:

Managing metadata is an important section of master data management. It is a complex, comprehensive and labor-intensive task. This paper explores whether and how metadata management can be augmented by machine learning. We deduce requirements for managing metadata from the literature and from expert interviews. We also identify features of machine learning algorithms. We assess 15 machine learning algorithms to determine their contribution to meeting the requirements and the extent to which they can support metadata management. Supervised and unsupervised learning algorithms as well as neural networks have the greatest potential to support metadata management effectively. Reinforcement learning, however, does not seem to be well suited to augment metadata management. Using Support Vector Machines and identification of metadata as an example, we show how machine learning algorithms can support metadata management.

Download: