12,976 student essays scored on a scale of 1 to 6 across multiple writing prompts. The data includes the full text of each essay alongside a single holistic score assigned by human graders.
Use Cases
- Train a transformer-based model to predict the score using the full_text as input
- Analyze the distribution of vocabulary and syntax across different score levels to identify key performance indicators in student writing
- Evaluate model generalization by training on specific prompt_id groups and testing on others
Strengths
- 12,976 unique essay entries with corresponding human-assigned scores
- score column provides a discrete target variable ranging from 1 to 6
- full_text column contains the raw string data of student writing
- essay_id and prompt_id columns allow for tracking specific responses and task contexts