Abstract-This article presents a survey of model-driven techniques for data model synthesis. During an extensive research, we identified more than 70 research papers in the field and more than 15 different graphical notations used for the source model representation. We have classified the proposed approaches into four distinct groups: function-oriented, process-oriented, communication-oriented and goal-oriented. Their contributions are presented in chronological order and evaluated based on several main criteria. Although the idea of model-driven design of the data model is more than 25 years old, the survey shows the richness and diversity of ideas, but only a small number of implemented automatic generators.