mDeBERTaV3-subjectivity

This GitHub contains all the code and fine-tuned mDeBERTaV3 models that we used for our participation in CLEF2023's CheckThat! Lab (Task 2).

More details about this lab can be found here.

All our fine-tuned models are hosted on HuggingFace and can be found here.

Files and Directories

The following files and directories can be found in this repository:

├── src
│ ├── make_predictions.py
│ └── mdebertav3_grid.py
├── data
│ └── multilingual_adapted.tsv
└── requirements.txt

• src: Contains the source code files for the project.

  • make_predictions.py: Python script that can be used to make predictions using a fine-tuned model on a test set.
  • mdebertav3_grid.py: Python script used to run random grid searches using Weights and Biases.

• data: Directory containing the multilingual_adapted.tsv file. This file represents the adapted multilingual dataset created by curating and sampling from other datasets available for the task.

• requirements.txt: File listing all the required modules and their versions needed to run the scripts in this project.

Additional datasets related to this project can be found in the CLEF2023-checkthat-lab repository.

How to Run

To run the project, follow these steps:

1. Create a Python virtual environment and activate it:

python3 -m venv myenv

source venv/bin/activate

2. Install the required packages from requirementst.txt

pip install -r requirements.txt

3. Now you should be able to run the scripts in the src directory:

python3 scrc/make_predictions.py

Acknowledgments

We thank the Center for Information Technology of the University of Groningen for their support and for providing access to the Hábrók high-performance computing cluster.

Affiliation

This project is affiliated with the University of Groningen.

Authors

Folkert Atze Leistra

Tommaso Caselli

For questions, pleace reach out to [email protected]