In order to ensure adequate education and training in a statistics-driven field, large sets of content-compliant training data (CCTD) are required. Within the context of practical orientation, such data sets should be as realistic as possible concerning the content in order to improve the learning experience. While there are different data generators for special use cases, the approaches mostly aim at evaluating the performance of database systems. Therefore, they focus on the structure but not on the content. Based on formulated requirements, this paper designs a possible approach for generating CCTD in the context of Big Data education. For this purpose, different Machine Learning algorithms could be utilized. In future work, specific models will be designed, implemented and evaluated.