Structured data in the form of Web tables and open data repositories lays the foundation for the Semantic Web.However, quality concerns are often raised, mostly with respect to accuracy and completeness. While missing value imputation is the go-to solution to fill in blanks, (1) it can only approximate known unknowns and (2) struggles if values are missing not at random. As a remedy, this paper combines semantically rich narratives with proven-quality knowledge graphs to dynamically assess the completeness of individual data sets while accounting for the user’s intent, too. Having determined which values are actually missing, we leverage state-of-the-art NLP techniques to identify functionally dependent attributes as proxies for later value imputation. Being functionally dependent (at least to some degree), these attributes provide the necessary context in the sense of relatedness allowing for more sophisticated imputation techniques. As a proof of concept we demonstrate our approach’s benefits in a real world setting using real-life narratives on the open data repository of the World Health Organization.