--- language: - en - cs license: mit task_categories: - text-classification pretty_name: Edustories dataset_info: features: - name: id dtype: int64 - name: description dtype: string - name: anamnesis dtype: string - name: problem_categories dtype: string - name: problem_possible_categories dtype: string - name: intervention dtype: string - name: intervention_categories dtype: string - name: intervention_possible_categories dtype: string - name: outcome dtype: string - name: implication_possible_categories dtype: string - name: age, school year dtype: string - name: hobbies dtype: string - name: diagnoses dtype: string - name: disorders dtype: string - name: anamnesis_cs dtype: string - name: intervention_cs dtype: string - name: outcome_cs dtype: string - name: annotator_id dtype: int64 - name: description_cs dtype: string - name: outcome_categories dtype: string - name: outcome_possible_categories dtype: string splits: - name: train num_bytes: 9225947 num_examples: 1492 download_size: 4975724 dataset_size: 9225947 configs: - config_name: default data_files: - split: train path: data/train-* --- # Dataset Card for Edustories dataset This repository contains the Edustories dataset (under review at ARR). The data contains structured descriptions of situations from classes documented by candidate teachers. Each entry, also called a casuistic, is structured into a `description` of the background, `anamnesis` describing the situation, an `intervention` describing the intervention of the teacher in the situation, and an `outcome` describing the final state of the intervention. Each entry was semi-automatically parsed from the original free-text journal and associated with additional information from our database. All entries were anonymised. In addition, our annotators manually associated each entry with a set of multiple categories that best fit the described situation, intervention, and outcome. ## About the dataset The dataset comes from student teachers, who collect case studies from their supervising teachers during their teaching practicum. These data are collected through standardized forms that the student teachers complete with their accompanying teachers. The collection of the dataset runs between 2023–2026. All students involved in the collection are informed of the use of the data and have given written consent. Additional case studies will be collected on an ongoing basis from practising teachers who choose to publish their anonymous case studies. All data is subject to multiple stages of anonymisation, so it does not contain any real names of schools, school staff, or students. ## Dataset format The dataset contains the following attributes: * **Identifier**: `id`. Selected entries have duplicate annotations, allowing evaluation of annotation uncertainty. * **Structured story**: `description`, `anamnesis`, `intervention`, and `outcome`, describing the situation, intervention, and its outcome in free text. * **Annotated category labels**: * `problem_categories` — category assignments for the described problem(s) * `intervention_categories` — category assignments for the described intervention(s) * `outcome_categories` — category assignments for the described outcome(s) * **Uncertain category labels**: * `problem_possible_categories` — possible problem category assignments marked as uncertain by annotators * `intervention_possible_categories` — possible intervention category assignments marked as uncertain by annotators * `outcome_possible_categories` — possible outcome category assignments marked as uncertain by annotators * **Student attributes**: `age`, `school year`, `hobbies`, `diagnoses`, `disorders`, detailing the profile of the student(s) acting in the entry. * **Teacher attributes**: `approbation` and `practice_years` of the teacher acting in the entry. * **Original free-text features in Czech**: `description_cs`, `anamnesis_cs`, `intervention_cs`, and `outcome_cs`, containing structured parts of the story in the original Czech language. ## Notes Paper under review.