direkt zum Inhalt springen

direkt zum Hauptnavigationsmenü

Sie sind hier

TU Berlin

Inhalt des Dokuments

Impact of Granularity on Readability of German Text: A Subjective Assessment

Location:  Zoom link (Please ask Saman Zadtootaghaj for access)  

Date/Time: 23.11.2020, 15:00-15:30 

SPEAKER:  Fynn Julius Heintz (TU Berlin)

Abstract: Automatic text assessment is a useful tool for applications such as text simplification. This thesis presents a data set of German texts that can be used to train a machine learning model to automatically evaluate a text’s complexity and understandability. It expands an existing data set of German sentences by adding texts at the paragraph level. The new data set can be used as a basis for future text complexity and understandability research at the paragraph level.

The paragraphs in the data set are annotated with subjective ratings of their readability, complexity, and understandability, combining the results of direct and indirect measurements. These ratings are collected on a web platform which is implemented for and also presented in this thesis. It can be reused for future text assessment studies. The ratings are then used to examine the relationship between the sentence and paragraph level and between the direct and indirect measurements. The results of this examination show a strong correlation between the different dimensions (readability, complexity, understandability), as well as between direct and indirect measurements. However, no apparent significant correlation between the ratings at the sentence and paragraph level can be found.

Zusatzinformationen / Extras

Direktzugang

Schnellnavigation zur Seite über Nummerneingabe