NLP-FBK/SGB-crf-italian

This dataset containes the data presented in Toward Automatic Filling of Case Report Forms: A Case Study on Data from an Italian Emergency Department, G. Kaczzmarek et al., 11th Language & Technology Conference: Human Language Technologies as a Challenge for Computer Science, Linguistics and Low Resourced Languages, 2025, Poznań, Poland

Case Report Forms are a set, standardised list of medical items which doctors need to fill with the correct values. Each patient accepted to the Emergency Department with Dyspnea is asked a list 134 questions to fill these items. The task consists in automatically filling the items from the clinical history of the patient. Each example in the dataset refers to one CRF item for one patient, i.e., for each patient there are as many examples as items in the CRF (134). The items are divided into 7 groups, each of them represented by a split. The dataset comes in 2 subsets (train and development).

• document_id: unique patient identifier,
• item: crf item that needs to be filled,
• ground_truth: ground truth filling value for the item,
• group_id: the type of document the clinical_case belongs to,
• clinical_case: clinical history of the patient.

CRF:filling SHared Task at CL4Health@LREC2026

The dataset has been used for the CRF:filling Shared Task at CL4Health@LREC2026, with a split modification (the train set has ben split into train and development, while the deelopment set became the test set). Please consider refering to NLP-FBK/dyspnea-crf-train, NLP-FBK/dyspnea-crf-development, and NLP-FBK/dyspnea-crf-test for the Shared Task data.