Ancient Greek BERT
The first and only available Ancient Greek sub-word BERT model!
State-of-the-art post fine-tuning on Part-of-Speech Tagging and Morphological Analysis.
Pre-trained weights are made available for a standard 12 layer, 768d BERT-base model.
You can also use the model directly on the HuggingFace Model Hub here.
Please refer to our paper titled: "A Pilot Study for BERT Language Modelling and Morphological Analysis for Ancient and Medieval Greek". In Proceedings of The 5th Joint SIGHUM Workshop on Computational Linguistics for Cultural Heritage, Social Sciences, Humanities and Literature (LaTeCH-CLfL 2021).
How to use
Requirements:
pip install transformers
pip install unicodedata
pip install flair
Can be directly used from the HuggingFace Model Hub with:
from transformers import AutoTokenizer, AutoModel
tokeniser = AutoTokenizer.from_pretrained("pranaydeeps/Ancient-Greek-BERT")
model = AutoModel.from_pretrained("pranaydeeps/Ancient-Greek-BERT")
Fine-tuning for POS/Morphological Analysis
- finetune_pos.py can be used to finetune the BERT model for POS tagging on your own data. We provide sample files from the Gold standard treebanks, however the full treebanks can't be made available at this time. Please contact the authors for more details.
- Even though the full treebanks aren't made available, we provide a pre-trained POS Tagging model in the directory SuperPeitho-FLAIR-v2, which can directly be used for inference and has an accuracy of ~90 percent on the 3 treebanks available. You can import the pre-trained model in FLAIR with:
from flair.models import SequenceTagger
tagger = SequenceTagger.load('SuperPeitho-FLAIR-v2/final-model.pt')
Training data
The model was initialised from AUEB NLP Group's Greek BERT and subsequently trained on monolingual data from the First1KGreek Project, Perseus Digital Library, PROIEL Treebank and Gorman's Treebank
Training and Eval details
Standard de-accentuating and lower-casing for Greek as suggested in AUEB NLP Group's Greek BERT. The model was trained on 4 NVIDIA Tesla V100 16GB GPUs for 80 epochs, with a max-seq-len of 512 and results in a perplexity of 4.8 on the held out test set. It also gives state-of-the-art results when fine-tuned for PoS Tagging and Morphological Analysis on all 3 treebanks averaging >90% accuracy. Please consult our paper or contact me for further questions!
Cite
If you end up using Ancient-Greek-BERT in your research, please cite the paper:
@inproceedings{ancient-greek-bert,
author = {Singh, Pranaydeep and Rutten, Gorik and Lefever, Els},
title = {A Pilot Study for BERT Language Modelling and Morphological Analysis for Ancient and Medieval Greek},
year = {2021},
booktitle = {The 5th Joint SIGHUM Workshop on Computational Linguistics for Cultural Heritage, Social Sciences, Humanities and Literature (LaTeCH-CLfL 2021)}
}
65 Dec 26, 2022
124 Dec 27, 2022
109 Dec 14, 2022
27 Jun 28, 2022
16 Sep 12, 2022
19 Dec 10, 2022
210 Dec 18, 2022
385 Jan 6, 2023
239 Jan 2, 2023
87 Jan 08, 2023
35 Dec 22, 2022
5 Dec 02, 2022
3.9k Dec 27, 2022
14 Apr 05, 2022
45 Jul 22, 2022
121 Nov 05, 2022
2 Jun 09, 2022
6 Oct 08, 2022
690 Jan 02, 2023
3 Feb 21, 2022
150 Dec 23, 2022
670 Jan 02, 2023
22 May 28, 2022
1 Feb 08, 2022
52 Dec 22, 2022
1 Jan 13, 2022
19 Nov 05, 2022
231 Nov 15, 2022
65 Dec 22, 2022