The scatter plot depicts a sample of sentences containing the term “liber” in various forms. Each sentence is depicted by one dot. The positions of the dots on the y and z axes are based on similarities between vectors of the token “liber” within these sentences. In principle, sentences relatively close to each other in the y and z dimensions should employ the term in similar sense. The x axis depicts temporal distributions of the sentences. However, the results are so far unsatisfactory. For more details, consult https://github.com/CCS-ZCU/latin-contextual-embeddings.