Managing semantic evolution in databases: from theory to implementation
Pedro Ivo Siqueira Nepomuceno, Kelly Rosa Braghetto · 2025
Semantic heterogeneity in long-term datasets emerges as categories, groupings, and units change over time, requiring users to manually adapt queries and results. The master’s thesis proposes a formal framework to address this issue through two strategies: query rewriting and data preprocessing. It introduces storage models and algorithms that manage semantic evolution via discrete, time-stamped operations (translation, merging, and splitting), enabling queries to be written as if the data were homogeneous. A prototype, MellowDB, was evaluated using Brazilian mortality data (1979–2021). Results show both approaches are production-ready, with data preprocessing generally outperforming query rewriting except in highly write-heavy scenarios.