Abstract: Techniques for managing information describing a hierarchy of relationships between educational institutions are described. According to various embodiments, first feature data describing a first school and second feature data describing a second school is accessed via one or more databases. A confidence score is then generated based on a machine learned model, the first feature data and the second feature data, the confidence score indicating a probability that the second school is a sub-school of the first school. Thereafter, based on a comparison of the confidence score to a threshold, is it is determined that the second school is a sub-school of the first school. Hierarchy information identifying a hierarchy of relationships between a plurality of schools is then generated or modified, the hierarchy information indicating that the second school is a sub-school of the first school.