---
language: su
tags:
  - sundanese-roberta-base-emotion-classifier
license: mit
widget:
  - text: "Wah, éta gélo, keren pisan!"
---

## Sundanese RoBERTa Base Emotion Classifier

Sundanese RoBERTa Base Emotion Classifier is an emotion-text-classification model based on the [RoBERTa](https://arxiv.org/abs/1907.11692) model. The model was originally the pre-trained [Sundanese RoBERTa Base](https://hf.co/w11wo/sundanese-roberta-base) model, which is then fine-tuned on the [Sundanese Twitter dataset](https://github.com/virgantara/sundanese-twitter-dataset), consisting of Sundanese tweets.

10% of the dataset is kept for evaluation purposes. After training, the model achieved an evaluation accuracy of 98.41% and F1-macro of 98.43%.

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) |
| ------------------------------------------- | ------- | ------------ | ------------------------------- |
| `sundanese-roberta-base-emotion-classifier` | 124M    | RoBERTa Base | Sundanese Twitter dataset       |

## Evaluation Results

The model was trained for 10 epochs and the best model was loaded at the end.

| Epoch | Training Loss | Validation Loss | Accuracy | F1       | Precision | Recall   |
| ----- | ------------- | --------------- | -------- | -------- | --------- | -------- |
| 1     | 0.801800      | 0.293695        | 0.900794 | 0.899048 | 0.903466  | 0.900406 |
| 2     | 0.208700      | 0.185291        | 0.936508 | 0.935520 | 0.939460  | 0.935540 |
| 3     | 0.089700      | 0.150287        | 0.956349 | 0.956569 | 0.956500  | 0.958612 |
| 4     | 0.025600      | 0.130889        | 0.972222 | 0.972865 | 0.973029  | 0.973184 |
| 5     | 0.002200      | 0.100031        | 0.980159 | 0.980430 | 0.980430  | 0.980430 |
| 6     | 0.001300      | 0.104971        | 0.980159 | 0.980430 | 0.980430  | 0.980430 |
| 7     | 0.000600      | 0.107744        | 0.980159 | 0.980174 | 0.980814  | 0.979743 |
| 8     | 0.000500      | 0.102327        | 0.980159 | 0.980171 | 0.979970  | 0.980430 |
| 9     | 0.000500      | 0.101935        | 0.984127 | 0.984376 | 0.984073  | 0.984741 |
| 10    | 0.000400      | 0.105965        | 0.984127 | 0.984142 | 0.983720  | 0.984741 |

## How to Use

### As Text Classifier

```python
from transformers import pipeline

pretrained_name = "sundanese-roberta-base-emotion-classifier"

nlp = pipeline(
    "sentiment-analysis",
    model=pretrained_name,
    tokenizer=pretrained_name
)

nlp("Wah, éta gélo, keren pisan!")
```

## Disclaimer

Do consider the biases which come from both the pre-trained RoBERTa model and the Sundanese Twitter dataset that may be carried over into the results of this model.

## Author

Sundanese RoBERTa Base Emotion Classifier 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.

## Citation Information

```bib
@article{rs-907893,
    author   = {Wongso, Wilson
                and Lucky, Henry
                and Suhartono, Derwin},
    journal  = {Journal of Big Data},
    year     = {2022},
    month    = {Feb},
    day      = {26},
    abstract = {The Sundanese language has over 32 million speakers worldwide, but the language has reaped little to no benefits from the recent advances in natural language understanding. Like other low-resource languages, the only alternative is to fine-tune existing multilingual models. In this paper, we pre-trained three monolingual Transformer-based language models on Sundanese data. When evaluated on a downstream text classification task, we found that most of our monolingual models outperformed larger multilingual models despite the smaller overall pre-training data. In the subsequent analyses, our models benefited strongly from the Sundanese pre-training corpus size and do not exhibit socially biased behavior. We released our models for other researchers and practitioners to use.},
    issn     = {2693-5015},
    doi      = {10.21203/rs.3.rs-907893/v1},
    url      = {https://doi.org/10.21203/rs.3.rs-907893/v1}
}
```