Models of word sense clustering have mainly been explored on synchronous, modern data. In contrast to these synchronous data sets, various historical word sense clustering data sets have been developed. This enables the evaluation of word sense disambiguation models on historical corpora and the exploration of their potential to detect changes in clusters over time (lexical semantic change). The aim of this thesis is to assess multiple context-based approaches to word sense disambiguation and lexical semantic change detection by relying on deep contextualized word embeddings and powerful token-based vector space models.