A benchmark for toxic comment classification on Civil Comments dataset

A) Abstract

Toxic comment detection on social media has proven to be essential for content moderation. This paper compares a wide set of different models on a highly skewed multi-label hate speech dataset. We consider inference time and several metrics to measure performance and bias in our comparison. We show that all BERTs have similar performance regardless of the size, optimizations or language used to pre-train the models. RNNs are much faster at inference than any of the BERT. BiLSTM remains a good compromise between performance and inference time. RoBERTa with Focal Loss offers the best performance on biases and AUROC. However, DistilBERT combines both good AUROC and a low inference time. All models are affected by the bias of associating identities. BERT, RNN, and XLNet are less sensitive than the CNN and Compact Convolutional Transformers.

B) How to install the virtual environment

B.1) Recommended method

The recommended method is to use pyenv and poetry.

To do this, you must already have installed on your machine :

• Pyenv
• Poetry

Then, you just have to :

• Git clone the project
• Run pyenv install 3.8.9 to install python 3.8.9 (if not already installed)
• Run pyenv shell 3.8.9 to use python 3.8.9
• Run poetry install in the project folder
• Run poetry shell to enable the Python virtual environment.

B.2) Alternative method

In case you don't have pyenv and poetry, you must :

• have Python 3.8.9 installed on your machine
• virtualenv on your Python 3.8.9 (for example via pip)

Then, you just have to :

• Git clone of the project
• Run python -m venv .venv in the project folder. Be careful to choose the right version of python, for example python3.9 -m venv .venv
• Run source ./.venv/bin/activate to activate the virtual environment
• Run pip install -r requirements.txt

C) Team

Name	Email (@epita.fr)	Github account
Corentin Duchêne	corentin.duchene	Nigiva
Henri Jamet	henri.jamet	hjamet
Pierre Guillaume	pierre.guillaume	drguigui1
Réda Dehak	reda.dehak

D) Citation

In EGC 2023, vol. RNTI-E-39, pp.19-30.

@article{RNTI/papers/1002807,
  author    = {Corentin Duchêne and Henri Jamet and Pierre Guillaume and Réda Dehak},
  title     = {Benchmark pour la classification de commentaires toxiques sur le jeu de données Civil Comments},
  journal = {Revue des Nouvelles Technologies de l'Information},
  volume = {Extraction et Gestion des Connaissances, RNTI-E-39},
  year      = {2023},
  pages     = {19-30}
}

E) License

MIT license