This model is part of the Master Thesis "Domain-adversarial moral foundation prediction for long EU legal documents". All related code is published on GitHub.
The models are finetuned for classifying moral foundations. We use domain-adversarial training to transfer knowledge from a labeled source domain to an unlabeled target domain.
This specific model is a finetuned google-bert/bert-base-uncased model that identifies the harm foundation, uses the MFRC, MFNC and MFNC as source domain and EU law documents as target domain.
Refer to JakobLindscheid/moral_foundations_ensemble-bert-DANN-eu_docs, which wraps all single label models in an ensemble model.