On the first day of practical classes, we continued to perform the exercises that were available for us, using our dataset.
After some tests and conversations with the professor, we came to the conclusion that splitting the dataset into 50/50 (half toxic questions and the other half non-toxic questions) could be a problem, as it could give us more false positives than expected.
In addition to these tasks, we also decided to improve our article, since the first delivery will be next Sunday and we want to advance as much as possible in order to have the maximum and best feedback from the teachers.