Abstract: Semantic heterogeneity often arises in long-term datasets due to changes in how data is categorized, named, or measured. This complicates querying, as users must manually account for inconsistencies over time—a time-consuming and error-prone task. The seminar introduces theoretical foundations and tools to tackle this issue through two strategies: query rewriting and data preprocessing. Both rely on tailored storage models and algorithms to ensure correct and semantically consistent retrieval. A prototype, MellowDB, enables querying heterogeneous data as if it were homogeneous, given a complete history of semantic changes. The approach models changes as discrete, time-stamped operations—translation, grouping, and ungrouping—with the possibility of extension. The system was tested on a real dataset: causes of death in Brazil (1979–2021).