LingMess: Linguistically Informed Multi Expert Scorers for Coreference Resolution
LingMess is a linguistically motivated categorization of mention-pairs into 6 types of coreference decisions and learn a dedicated trainable scoring function for each category. This significantly improves the accuracy of the pairwise scorer as well as of the overall coreference performance on the English Ontonotes coreference corpus.
Please check the official repository for more details and updates.
Training on OntoNotes
We present the test results on OntoNotes 5.0 dataset.
| Model | Avg. F1 |
|---|---|
| SpanBERT-large + e2e | 79.6 |
| Longformer-large + s2e | 80.3 |
| Longformer-large + LingMess | 81.4 |
Citation
If you find LingMess useful for your work, please cite the following paper:
@misc{https://doi.org/10.48550/arxiv.2205.12644,
doi = {10.48550/ARXIV.2205.12644},
url = {https://arxiv.org/abs/2205.12644},
author = {Otmazgin, Shon and Cattan, Arie and Goldberg, Yoav},
keywords = {Computation and Language (cs.CL), FOS: Computer and information sciences, FOS: Computer and information sciences},
title = {LingMess: Linguistically Informed Multi Expert Scorers for Coreference Resolution},
publisher = {arXiv},
year = {2022},
copyright = {Creative Commons Attribution 4.0 International}
}