EstBERT
What's this?
The EstBERT model is a pretrained BERT<sub>Base</sub> model exclusively trained on Estonian cased corpus on both 128 and 512 sequence length of data.
How to use?
You can use the model transformer library both in tensorflow and pytorch version.
from transformers import AutoTokenizer, AutoModelForMaskedLM
tokenizer = AutoTokenizer.from_pretrained("tartuNLP/EstBERT")
model = AutoModelForMaskedLM.from_pretrained("tartuNLP/EstBERT")
You can also download the pretrained model from here, EstBERT_128 EstBERT_512
Dataset used to train the model
The EstBERT model is trained both on 128 and 512 sequence length of data. For training the EstBERT we used the Estonian National Corpus 2017, which was the largest Estonian language corpus available at the time. It consists of four sub-corpora: Estonian Reference Corpus 1990-2008, Estonian Web Corpus 2013, Estonian Web Corpus 2017 and Estonian Wikipedia Corpus 2017.
Reference to cite
Why would I use?
Overall EstBERT performs better in parts of speech (POS), name entity recognition (NER), rubric, and sentiment classification tasks compared to mBERT and XLM-RoBERTa. The comparative results can be found below;
Model | UPOS | XPOS | Morph | bf UPOS | bf XPOS | Morph |
---|---|---|---|---|---|---|
EstBERT | 97.89 | 98.40 | 96.93 | 97.84 | 98.43 | 96.80 |
mBERT | 97.42 | 98.06 | 96.24 | 97.43 | 98.13 | 96.13 |
XLM-RoBERTa | 97.78 | 98.36 | 96.53 | 97.80 | 98.40 | 96.69 |
Model | Rubric<sub>128</sub> | Sentiment<sub>128</sub> | Rubric<sub>128</sub> | Sentiment<sub>512</sub> |
---|---|---|---|---|
EstBERT | 81.70 | 74.36 | 80.96 | 74.50 |
mBERT | 75.67 | 70.23 | 74.94 | 69.52 |
XLM-RoBERTa | 80.34 | 74.50 | 78.62 | 76.07 |
Model | Precicion<sub>128</sub> | Recall<sub>128</sub> | F1-Score<sub>128</sub> | Precision<sub>512</sub> | Recall<sub>512</sub> | F1-Score<sub>512</sub> |
---|---|---|---|---|---|---|
EstBERT | 88.42 | 90.38 | 89.39 | 88.35 | 89.74 | 89.04 |
mBERT | 85.88 | 87.09 | 86.51 | 88.47 | 88.28 | 88.37 |
XLM-RoBERTa | 87.55 | 91.19 | 89.34 | 87.50 | 90.76 | 89.10 |