Maverick Coreference Resolution
Collection
Efficient and Accurate Coreference Resolution models.
•
3 items
•
Updated
•
8
Official weights for Maverick-mes trained on PreCo and based on DeBERTa-large. This model achieves 87.4 Avg CoNLL-F1 on PreCo coreference resolution dataset.
Other available models at SapienzaNLP huggingface hub:
hf_model_name | training dataset | Score | Singletons |
---|---|---|---|
"sapienzanlp/maverick-mes-ontonotes" | OntoNotes | 83.6 | No |
"sapienzanlp/maverick-mes-litbank" | LitBank | 78.0 | Yes |
"sapienzanlp/maverick-mes-preco" | PreCo | 87.4 | Yes |
N.B. Each dataset has different annotation guidelines, choose your model according to your use case.
@inproceedings{martinelli-etal-2024-maverick,
title = "Maverick: Efficient and Accurate Coreference Resolution Defying Recent Trends",
author = "Martinelli, Giuliano and
Barba, Edoardo and
Navigli, Roberto",
booktitle = "Proceedings of the Annual Meeting of the Association for Computational Linguistics (ACL 2024)",
year = "2024",
address = "Bangkok, Thailand",
publisher = "Association for Computational Linguistics",
}