Registration

If you would like to participate in this shared task, please fill out the registration form. Once we receive your registration information, we will send the training data to your email address. Please check your email regularly. If you do not receive a reply for two days, please re-send the form.

No need to register for each task seperately, you can participate in both tasks as long as the group stays the same. Different groups require seperate registrations.

Submitting Your Results

Once the deadline for registration is due we will publish the link to send the results. The submissions should be formatted as follows:

File Naming

The name of the should end with “_lemmatization.csv” or “_tokenprediction.csv” according to the task. If you participate in both tasks The results should be in a CSV file that is identical in structure to the train and the test files but has a ‘prediction’ column where you put the lemma or the predicted token.

Writing the Technical Report

Papers should not be longer than 4 pages of content (not including references). Each task gets its own technical report. To allow blind reviewing, please do not include author names and affiliations within the paper and avoid obvious self-references. The technical report should include specific examples for the chosen task and the model’s predictions as well as an error analysis section.

Papers must be submitted according to the submission deadline.

Prizes

To win a task, a team must provide the results file along with their evaluation score and the technical report. We will re-run the results file against our evaluation script for authentication and incorporate the score of the technical report according to the reviewers. The combined result of the report score and test score will be the final score for the team.

The winning team for each task will win 500 Euros

The first runner-up for each task will win 250 Euros

To receive the prize groups need to make the code available on Github and provide details of a single bank account.

Winning models will be used by the eBL and ARCHIBAB projects.