Bridge is a dataset of student and human-tutor interactions annotated by expert teachers. The dataset consists of 700 online math tutoring conversations, subsetted to excerpts where a student makes a mistake. Each excerpt has been annotated for the type of mistake, the strategy for correction, and the intention. The teacher-annotators also provide an expert response to the student's mistake.

This dataset is useful for training models to identify and respond to student mistakes, both to improve AI-based tutors and to provide useful feedback to human tutors.

Permissions and licensing

License

Creative Commons Attribution-NonCommercial 4.0

Key restrictions

This dataset is published under a non-commercial license and is available for research purposes only.

How to access the data

This dataset is available to anyone. To access the data, head to the Data download tab and follow the instructions provided.