Resolving Elliptical Compounds in German Medical Text

Preparation

1. Get access to GGPONC following the instructions on the project homepage and place the contents of ellipses_2023_01_30 in the data/ellipses folder (or adapt the path in experiment.yaml to point to another directory)
2. Install Python dependencies pip install -r requirements.txt
3. Fill and adjust the config file experiment.yaml. If you want to use any of the approaches that require OpenAI services you need to specify your API key in the config

Notebooks

In notebooks, we provide the following Jupyter Notebooks to reproduce the results from the paper:

• 01_Dataset.ipynb
  • Corpus Statistics
• 02_Baseline_Aepli.ipynb
  • Updated implementation of a rule-based baseline by Aepli & Volk, 2012
  • In order to run the baseline, you also need to download the full GGPONC 2.0 corpus and place it under data/ggponc_v2 (or adapt the path in experiment.yaml to point to your data folder)
• 03_Generative.ipynb
  • Generative approach using encoder-decoder Transformer-based model (default: mT5)
• 04_Zero_Shot.ipynb
  • Zero-shot approach using ChatGPT (API key needed)
• 05_TopK.ipynb
  • Multiple-choice approaches using generative model to generate the k most likely sentences and ChatGPT to choose the best one (API Key needed)

Running Generative Transformer Experiments with HuggingFace and Hydra

In scripts, we provide a Hydra script that runs training with the optimal hyperparameters set in experiment.yaml.

To run such an experiment, do: python scripts/run_experiment.py

If you have installed and configured Weights & Biases, it will automatically sync your runs.

To run a hyperparameter sweep, specify your desired paramters in experiment.yaml under params and pass the optiom -m to Hydra, e.g.: python scripts/run_experiment.py -m

Citation

If you find our data or code useful for your work, please cite the following paper:

@inproceedings{kammer-etal-2023-resolving,
    title = "Resolving Elliptical Compounds in {G}erman Medical Text",
    author = "Kammer, Niklas  and
      Borchert, Florian  and
      Winkler, Silvia  and
      de Melo, Gerard  and
      Schapranow, Matthieu-P.",
    booktitle = "The 22nd Workshop on Biomedical Natural Language Processing and BioNLP Shared Tasks",
    month = jul,
    year = "2023",
    address = "Toronto, Canada",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2023.bionlp-1.26",
    pages = "292--305",
}