|
--- |
|
license: mit |
|
datasets: |
|
- oscar |
|
- DDSC/dagw_reddit_filtered_v1.0.0 |
|
- graelo/wikipedia |
|
language: |
|
- da |
|
widget: |
|
- text: Der var engang en [MASK] |
|
--- |
|
|
|
# What is this? |
|
|
|
A pre-trained BERT model (base version, ~110 M parameters) for Danish NLP. The model was not pre-trained from scratch but adapted from the English version with a tokenizer trained on Danish text. |
|
|
|
# How to use |
|
|
|
Test the model using the pipeline from the [🤗 Transformers](https://github.com/huggingface/transformers) library: |
|
|
|
```python |
|
from transformers import pipeline |
|
|
|
pipe = pipeline("fill-mask", model="KennethTM/bert-base-uncased-danish") |
|
|
|
pipe("Der var engang en [MASK]") |
|
``` |
|
|
|
Or load it using the Auto* classes: |
|
|
|
```python |
|
# Load model directly |
|
from transformers import AutoTokenizer, AutoModelForMaskedLM |
|
|
|
tokenizer = AutoTokenizer.from_pretrained("KennethTM/bert-base-uncased-danish") |
|
model = AutoModelForMaskedLM.from_pretrained("KennethTM/bert-base-uncased-danish") |
|
``` |
|
|
|
# Model training |
|
|
|
The model is trained using multiple Danish datasets and a context length of 512 tokens. |
|
|
|
The model weights are initialized from the English [bert-base-uncased model](https://huggingface.co/bert-base-uncased) with new word token embeddings created for Danish using [WECHSEL](https://github.com/CPJKU/wechsel). |
|
|
|
Initially, only the word token embeddings are trained using 1.000.000 samples. Finally, the whole model is trained for 8 epochs. |
|
|
|
|
|
# Evaluation |
|
|
|
The performance of the pretrained model was evaluated using [ScandEval](https://github.com/ScandEval/ScandEval). |
|
|
|
| task | dataset | summary | |
|
|:-------------------------|:-------------|:-------------------------------------------------------------------------------------------| |
|
| sentiment-classification | swerec | mcc = 63.02, mcc_se = 2.16, macro_f1 = 62.2, macro_f1_se = 3.61 | |
|
| sentiment-classification | angry-tweets | mcc = 47.21, mcc_se = 0.53, macro_f1 = 64.21, macro_f1_se = 0.53 | |
|
| sentiment-classification | norec | mcc = 42.23, mcc_se = 8.69, macro_f1 = 57.24, macro_f1_se = 7.67 | |
|
| named-entity-recognition | suc3 | micro_f1 = 50.03, micro_f1_se = 4.16, micro_f1_no_misc = 53.55, micro_f1_no_misc_se = 4.57 | |
|
| named-entity-recognition | dane | micro_f1 = 76.44, micro_f1_se = 1.36, micro_f1_no_misc = 80.61, micro_f1_no_misc_se = 1.11 | |
|
| named-entity-recognition | norne-nb | micro_f1 = 68.38, micro_f1_se = 1.72, micro_f1_no_misc = 73.08, micro_f1_no_misc_se = 1.66 | |
|
| named-entity-recognition | norne-nn | micro_f1 = 60.45, micro_f1_se = 1.71, micro_f1_no_misc = 64.39, micro_f1_no_misc_se = 1.8 | |
|
| linguistic-acceptability | scala-sv | mcc = 5.01, mcc_se = 5.41, macro_f1 = 49.46, macro_f1_se = 3.67 | |
|
| linguistic-acceptability | scala-da | mcc = 54.74, mcc_se = 12.22, macro_f1 = 76.25, macro_f1_se = 6.09 | |
|
| linguistic-acceptability | scala-nb | mcc = 19.18, mcc_se = 14.01, macro_f1 = 55.3, macro_f1_se = 8.85 | |
|
| linguistic-acceptability | scala-nn | mcc = 5.72, mcc_se = 5.91, macro_f1 = 49.56, macro_f1_se = 3.73 | |
|
| question-answering | scandiqa-da | em = 26.36, em_se = 1.17, f1 = 32.41, f1_se = 1.1 | |
|
| question-answering | scandiqa-no | em = 26.14, em_se = 1.59, f1 = 32.02, f1_se = 1.59 | |
|
| question-answering | scandiqa-sv | em = 26.38, em_se = 1.1, f1 = 32.33, f1_se = 1.05 | |
|
| speed | speed | speed = 4.55, speed_se = 0.0 | |
|
|
