🥇 Data Science Kitchen at GermEval 2021

This repository contains the codebase and resources for the paper Data Science Kitchen at GermEval 2021: A Fine Selection of Hand-Picked Features, Delivered Fresh from the Oven, which was submitted to the GermEval 2021 shared task. The task focuses on identifying comments that are toxic, engaging, and fact-claiming, with the goal of helping moderators and community managers prioritize content for fact-checking and engagement.

📁 Repository Structure

• config: Includes training configuration files in YAML format, which can be used to train the model on different feature sets.
• data: Contains the GermEval 2021 dataset for training and evaluation, as well as pretrained weights of the AdHominem model.
• src: The codebase, including functions for data preparation, model training and evaluation.

👷‍♀️ Installation

To run the code in this repository, follow these steps:

1. Clone this repository:

   git clone https://github.com/data-science-kitchen/germ-eval-2021.git
   cd germ-eval-2021

2. Install the required Python packages:

   pip install -r requirements.txt

🚀 Usage

We have provided a single script run_pipeline.py, which performs the entire process of data preparation, feature extraction, model training, tuning and evaluation via a single function call. To reproduce the results from the paper, simply run the following command:

python run_pipeline.py data/GermEval21_Toxic_Train.csv data/GermEval21_Toxic_TestData.csv config/all-features.yaml

Please refer to the corresponding function documentation for additional arguments you can provide to this function.

📑 Citation

If you use the code in this repository, please cite our paper:

@inproceedings{germeval2021-dsk,
    abbr = {KONVENS},
    author = {N. Hildebrandt and B. Bönninghoff and D. Orth and C. Schymura},
    booktitle = {Konferenz zur Verarbeitung natürlicher Sprache/Conference on Natural Language Processing (KONVENS)},
    title = {Data {S}cience {K}itchen at {G}erm{E}val 2021: {A} Fine Selection of Hand-Picked Features, Delivered Fresh from the Oven},
    year = {2021},
    pages = {88-94}
}

🧑‍⚖️ License

This project is licensed under the MIT License - see the LICENSE file for details.