Instructions for Participants

 📕 | How to Participate

To express your interest in participating, please fill this form. This will help us organizing the shared task, and update you with relevant information.

To enrol, go to the "My Submission" tab on Codabench here.

📨 | How to Submit

TL;DR

The GitHub repository reports the scripts to assess your submission and prepare the files for the submission to the CodaBench leaderboard.

Required submission format

Detailed formatting guidelines and automatic assessment can be found in the GitHub repository. 

The output must be provided in .jsonl format, where each line represents a single JSON object containing the following fields:

• id: the identifier of the clinical note as reported in the development data

• predictions: a list of dictionaries, where each dictionary has 2 items, item and prediction. 

Here is an example of the submission file content: 

{"id":"0001","predictions":[{"item":"Exam:haemoglobin","prediction":"10g/dl."},...]} {"id":"0002","predictions":[{"item":"Exam:haemoglobin","prediction":"unknown"},...]} 

... 

{"id":"9999","predictions":[{"item":"Exam:haemoglobin","prediction":"11g/dl "},...]}

Submission link

Submissions have to be uploaded into the CodaBench page.

Submissions in Italian

The English development and test clinical notes are translated from Italian. In the submission file, participants only report the id of the clinical note, and the predictions. Therefore, Submission files are language-agnostic.