| # Evalution script (evalution.py) instructions | |
| * We provide an evaluation script that can easily calculate the scores achieved in each task, cluster and overall. | |
| - The predictions must be in .txt format. | |
| - All predictions files should be in the same directory and follow the naming conversions as in the "dummy_predictions" directory provided. | |
| - Each prediction file must contain the predictions made on the test set of the respective task with each line corresponding to one line. | |
| - The predictions should be written in full (e.g. 'strongly negative', instead of 0) and in cases of multiple outputs (e.g. TweetTopic) should be comma separated. | |
| * "dummy_predictions" directory provides the prediciton files produced by the roberta-base model tested and can be used as a template. | |
| * Steps to run the script: | |
| 1. Install requirements.txt | |
| 2. Run script: python evalution.py -p {directory with predictions}. e.g. python -p ./dummy_predictions |