Patterns of language dynamics in deep contextualized word representations (Kai Kugler)

Kai Kugler M.A., Prof. Dr. Achim Rettinger
research group: krAil – knowledge representation learning

In recent history no event has changed the world and standard knowledge of speakers worldwide as much as the corona pandemic. The associated language dynamics are not only noticeable by lexical diversity and conventions, but also in semantics and pragmatics.

The aim of the project as part of the joint project “Patterns. Linguistic Creativity and Variation in Synchrony and Diachrony” ( is the data-driven analysis of these semantic changes using linguistic entities in their respective context. Theoretically based on construction grammar and distributional semantics, models are to be extracted with the help of current deep learning methods (word embeddings and deep contextualized representations), in which the patterns to be examined are latent. These representations can be used to compare lexical units and their semantic changes in different contexts, at different times, in different languages or in different types of media.