Text values are valuable information in relational database systems for analysis and machine learning (ML) tasks. Since ML techniques depend on numerical input representations, word embeddings are increasingly utilized to convert symbolic representations such as text into meaningful numbers. However, those models do not incorporate the context-specific semantics of text values in the database. To significantly improve the representation of text values occurring in DBMS, we propose a novel retrofitting approach called Retro which considers both, the semantics of the word embedding and the relational schema. Based on this, we developed RetroLive, an interactive system, that allows exploring how the retrofitted embeddings improve the performance for various ML and integration tasks. Moreover, the demo includes several interactive visualizations to explore the characteristics of the adapted vectors and their connection to the relational database.