Step 7: Applications and Further Information
These tools are interesting for various applications in Intelligent Tutoring
Systems. For example:
- (Unsupervised) clustering: The distance matrix of the dataset
enables cluster analysis, such that the number of possible solution strategies
for a given task can be estimated automatically.
- Classification: Based on a clustering, one can estimate the
underlying programming strategy of new, unseen solutions.
- Partner Selection: Similar solutions from a dataset of existing
solutions can be selected in order to create automated, example-based feedback
or pair students for peer learning setups.
- Dissimilarity Analysis: Alignment algorithms do not only provide
an estimate of similarity between solutions, but also give information on
similar and dissimilar parts of the solutions, because similar parts are
aligned. Such precise localizations might enhance automated
feedback. For example, dissimilarities between a new student solution and a
known, correct sample solution might indicate errors in the student solution.
This approach is not at all restricted to Java programming. Other learning
domains with solutions in sequential form include body movements in sports,
free text and formal arguments.
We do encourage you to employ this free software toolbox in your own research
or practical project. You can find more information at our
website.