PUB - Publikationen an der Universität Bielefeld

This is a dataset of time series or _traces_ of UML diagrams with the purpose to test the ability of artificial hint systems to provide next-step hints to students. The example learning task investigated in this dataset is to generate an [UML activity diagram](https://en.wikipedia.org/wiki/Activity_diagram) that models an algorithm to add two binary numbers. The dataset consists of twelve _training_ traces, which contain steps to a correct solution to the task, and twelve _test_ traces, which contain steps toward a possibly erroneous solution to the task. For each step of the test traces, hints of human tutors are available, accompanied by grades for these hints. To evaluate an artificial system, one can compare to the hints generated by human tutors.