An attempt at the SemEval 2021 Task 5: Toxic Spans Detection.

The Toxic Spans Detection task of SemEval2021 required participants to predict the spans of toxic posts that were responsible for the toxic label of the posts. You can read the task overview paper for more information about the task, data, evaluation, and performance of the participating systems.

python3.8 -m pip install poetry

and then:

python3.8 -m poetry install

To serve the Jupyter notebooks, available under notebooks/, run:

python3.8 -m poetry run jupyter notebook

The team that has ranked first in this competition (HITSZ-HLT) has achieved an F1 score of 70.83% , with the median score across all 36 participating teams being 67.58%.

Our approach, as presented here, yields: 64.46%.