Zum Inhalt springen

Research on text simplification is commonly motivated by the assumption that available text is too complex. So far, an empirical basis for this claim is missing. This paper provides a starting point for substantiating the claim that complex text is ubiquitous. With a focus on German public administration, this paper reports the results of a data-mining audit study investigating more than 49,900 websites of 26 large counties in Germany. We show that difficult text is a universal problem. The vast majority of text on official German county websites is either difficult or very difficult. Despite slight differences between counties, text across counties was difficult to understand. We also found no strong correlation between the number of inhabitants, a county’s area, or population density. We make actionable recommendations on how to continue studying this in the future.

About the authors

Hendrik Heuer
Prof. Dr. Hendrik Heuer is a research professor at the Center for Advanced Internet Studies (CAIS) & the University of Wuppertal. He heads the research program Design of Trustworthy Artificial Intelligence. His research focuses on Human-Computer Interaction and Machine Learning. He is trying to increase Trust in Artificial Intelligence through Understanding, Control, and Co-design. He is also interested in Participatory Software Development for Machine Learning and Combating Misinformation and Disinformation. In the past, he was an Interim Professor of Applied Computer Science at the University of Bremen and a Postdoctoral Research Fellow at The Variation Lab, Harvard University.

David Fröhlich, capito
Verena Riegler, capito
Michael Radeka, capito
Julia Gspandl, capito