Migrate model card from transformers-repo
Browse filesRead announcement at https://discuss.huggingface.co/t/announcement-all-model-cards-will-be-migrated-to-hf-co-model-repos/2755
Original file history: https://github.com/huggingface/transformers/commits/master/model_cards/tartuNLP/EstBERT/README.md
README.md
ADDED
|
@@ -0,0 +1,41 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language: et
|
| 3 |
+
---
|
| 4 |
+
# EstBERT
|
| 5 |
+
|
| 6 |
+
|
| 7 |
+
### What's this?
|
| 8 |
+
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.
|
| 9 |
+
|
| 10 |
+
### How to use?
|
| 11 |
+
You can use the model transformer library both in tensorflow and pytorch version.
|
| 12 |
+
```
|
| 13 |
+
from transformers import AutoTokenizer, AutoModelForMaskedLM
|
| 14 |
+
tokenizer = AutoTokenizer.from_pretrained("tartuNLP/EstBERT")
|
| 15 |
+
model = AutoModelForMaskedLM.from_pretrained("tartuNLP/EstBERT")
|
| 16 |
+
```
|
| 17 |
+
You can also download the pretrained model from here, [EstBERT_128]() [EstBERT_512]()
|
| 18 |
+
#### Dataset used to train the model
|
| 19 |
+
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](https://metashare.ut.ee/repository/browse/estonian-national-corpus-2017/b616ceda30ce11e8a6e4005056b40024880158b577154c01bd3d3fcfc9b762b3/), 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.
|
| 20 |
+
|
| 21 |
+
### Why would I use?
|
| 22 |
+
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;
|
| 23 |
+
|
| 24 |
+
|Model |UPOS |XPOS |Morph |bf UPOS |bf XPOS |Morph |
|
| 25 |
+
|--------------|----------------------------|-------------|-------------|-------------|----------------------------|----------------------------|
|
| 26 |
+
| EstBERT | **_97.89_** | **98.40** | **96.93** | **97.84** | **_98.43_** | **_96.80_** |
|
| 27 |
+
| mBERT | 97.42 | 98.06 | 96.24 | 97.43 | 98.13 | 96.13 |
|
| 28 |
+
| XLM-RoBERTa | 97.78 | 98.36 | 96.53 | 97.80 | 98.40 | 96.69 |
|
| 29 |
+
|
| 30 |
+
|
| 31 |
+
|Model|Rubric<sub>128</sub> |Sentiment<sub>128</sub> | Rubric<sub>128</sub> |Sentiment<sub>512</sub> |
|
| 32 |
+
|-------------------|----------------------------|--------------------|-----------------------------------------------|----------------------------|
|
| 33 |
+
| EstBERT | **_81.70_** | 74.36 | **80.96** | 74.50 |
|
| 34 |
+
| mBERT | 75.67 | 70.23 | 74.94 | 69.52 |
|
| 35 |
+
| XLM\-RoBERTa | 80.34 | **74.50** | 78.62 | **_76.07_**|
|
| 36 |
+
|
| 37 |
+
|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> |
|
| 38 |
+
|--------------|----------------|----------------------------|----------------------------|----------------------------|-------------|----------------|
|
| 39 |
+
| EstBERT | **88.42** | 90.38 |**_89.39_** | 88.35 | 89.74 | 89.04 |
|
| 40 |
+
| mBERT | 85.88 | 87.09 | 86.51 |**_88.47_** | 88.28 | 88.37 |
|
| 41 |
+
| XLM\-RoBERTa | 87.55 |**_91.19_** | 89.34 | 87.50 | **90.76** | **89.10** |
|