SemEval is a dataset of science questions with graded short-answer responses. Each answer is categorized as falling into one of five correctness categories: Correct, Partially_correct_incomplete, Contradictory, Irrelevant, or Non_domain. The corpus contains 253 questions with over 13,000 student responses. Topics cover 15 science domains ranging from 6th grade to college material.

This dataset is useful for training models to grade student short-answer responses to science questions.

Permissions and licensing

License

Creative Commons Attribution-ShareAlike 3.0 Unported

Key restrictions

You may use this data for any purpose, even commercially. Any derivative work must give credit to the original source and be shared under the same license as the original.

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.