HPI-DHC @ BioASQ DisTEMSIT

Preparation

• Download DisTEMIST data (v5.0) from Zenodo (https://zenodo.org/record/6532684) and unzip content into data/distemist
• Download trained models and dictionaries from Zenodo (https://zenodo.org/record/6642064) and extract into dicts and models
• Install Python dependencies:
  • pip install -r requirements.txt
  • python -m spacy download es_core_news_md

Subtrack 1: Named Entity Recognition

Model Training

The Hydra config file ner_config.yaml contains all hyperparameters to reproduce the NER training results.

To train the model on CUDA device n, run:

CUDA_VISIBLE_DEVICES=<n> python scripts/run_ner_training.py

We performed a hyperparameter grid search over the parameters listed in ner_hyperparamter_sweep.sh.

Prediction

See: notebooks/01_Entities.ipynb

Subtrack 2: Entity Linking

See: notebooks/02_Linking.ipynb