|
--- |
|
language: lo |
|
tags: |
|
- lao-roberta-base-pos-tagger |
|
license: mit |
|
widget: |
|
- text: "ຮ້ອງ ມ່ວນ ແທ້ ສຽງດີ ອິຫຼີ" |
|
--- |
|
|
|
## Lao RoBERTa Base POS Tagger |
|
|
|
Lao RoBERTa Base POS Tagger is a part-of-speech token-classification model based on the [RoBERTa](https://arxiv.org/abs/1907.11692) model. The model was originally the pre-trained [Lao RoBERTa Base](https://huggingface.co/w11wo/lao-roberta-base) model, which is then fine-tuned on the [`Yunshan Cup 2020`](https://github.com/GKLMIP/Yunshan-Cup-2020) dataset consisting of tag-labelled Lao corpus. |
|
|
|
After training, the model achieved an evaluation accuracy of 83.14%. On the benchmark test set, the model achieved an accuracy of 83.30%. |
|
|
|
Hugging Face's `Trainer` class from the [Transformers](https://huggingface.co/transformers) library was used to train the model. PyTorch was used as the backend framework during training, but the model remains compatible with other frameworks nonetheless. |
|
|
|
## Model |
|
|
|
| Model | #params | Arch. | Training/Validation data (text) | |
|
| ----------------------------- | ------- | ------------ | ------------------------------- | |
|
| `lao-roberta-base-pos-tagger` | 124M | RoBERTa Base | `Yunshan Cup 2020` | |
|
|
|
## Evaluation Results |
|
|
|
The model was trained for 15 epochs, with a batch size of 8, a learning rate of 5e-5, with cosine annealing to 0. The best model was loaded at the end. |
|
|
|
| Epoch | Training Loss | Validation Loss | Accuracy | |
|
| ----- | ------------- | --------------- | -------- | |
|
| 1 | 1.026100 | 0.733780 | 0.746021 | |
|
| 2 | 0.646900 | 0.659625 | 0.775688 | |
|
| 3 | 0.500400 | 0.576214 | 0.798523 | |
|
| 4 | 0.385400 | 0.606503 | 0.805269 | |
|
| 5 | 0.288000 | 0.652493 | 0.809092 | |
|
| 6 | 0.204600 | 0.671678 | 0.815216 | |
|
| 7 | 0.145200 | 0.704693 | 0.818209 | |
|
| 8 | 0.098700 | 0.830561 | 0.816998 | |
|
| 9 | 0.066100 | 0.883329 | 0.825232 | |
|
| 10 | 0.043900 | 0.933347 | 0.825664 | |
|
| 11 | 0.027200 | 0.992055 | 0.828449 | |
|
| 12 | 0.017300 | 1.054874 | 0.830819 | |
|
| 13 | 0.011500 | 1.081638 | 0.830940 | |
|
| 14 | 0.008500 | 1.094252 | 0.831304 | |
|
| 15 | 0.007400 | 1.097428 | 0.831442 | |
|
|
|
## How to Use |
|
|
|
### As Token Classifier |
|
|
|
```python |
|
from transformers import pipeline |
|
|
|
pretrained_name = "w11wo/lao-roberta-base-pos-tagger" |
|
|
|
nlp = pipeline( |
|
"token-classification", |
|
model=pretrained_name, |
|
tokenizer=pretrained_name |
|
) |
|
|
|
nlp("ຮ້ອງ ມ່ວນ ແທ້ ສຽງດີ ອິຫຼີ") |
|
``` |
|
|
|
## Disclaimer |
|
|
|
Do consider the biases which come from both the pre-trained RoBERTa model and the `Yunshan Cup 2020` dataset that may be carried over into the results of this model. |
|
|
|
## Author |
|
|
|
Lao RoBERTa Base POS Tagger was trained and evaluated by [Wilson Wongso](https://w11wo.github.io/). All computation and development are done on Google Colaboratory using their free GPU access. |
|
|
