Ancient Greek BERT finetuned for tagging and parsing PROIEL (UD)
This is a finetuned checkpoint of Ancient Greek BERT by Singh, Rutten and Lefever (2021), which has been trained on the UD version of PROIEL per 9 May 2023.
The code for training and using the model can be found on GitHub. The config file used is here: config.py.
If you use this model for something academic, feel free to cite the master's thesis that it sprung out of:
Bjerkeland, D. C. 2022. Tagging and Parsing Old Texts with New Techniques. University of Oslo. URL: http://urn.nb.no/URN:NBN:no-98954.
Performance
This is the performance on the test set of the UD version of PROIEL per 9 May 2023.
| Metric | Accuracy |
|---|---|
| UPOS | 0.9820489710079615 |
| XPOS | 0.9822742977317109 |
| feats | 0.926768814781433 |
| all tags | 0.916103349857293 |
| UAS | 0.8862100045065344 |
| LAS | 0.8496319663512093 |
| LA | 0.9076911521706474 |
Bibliography
- Bjerkeland, D. C. 2022. Tagging and Parsing Old Texts with New Techniques. University of Oslo. URL: http://urn.nb.no/URN:NBN:no-98954.
- Singh, P., Rutten, G., Lefever, E. 2021. A Pilot Study for BERT Language Modelling and Morphological Analysis for Ancient and Medieval Greek. Proceedings of LaTeCH-CLfL 2021, pp. 128–137. URL: https://doi.org/10.18653/v1/2021.latechclfl-1.15.
