File size: 2,707 Bytes
85544c7 e8a9cb9 85544c7 e8a9cb9 85544c7 daf8cb6 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 |
---
language: id
tags:
- roberta
license: mit
datasets:
- indonlu
widget:
- text: "Hal-hal baik akan datang."
---
# Indo RoBERTa Emotion Classifier
Indo RoBERTa Emotion Classifier is emotion classifier based on [Indo-roberta](https://huggingface.co/flax-community/indonesian-roberta-base) model. It was trained on the trained on [IndoNLU EmoT](https://huggingface.co/datasets/indonlu) dataset. The model used was [Indo-roberta](https://huggingface.co/flax-community/indonesian-roberta-base) and was transfer-learned to an emotion classifier model. Based from the [IndoNLU bencmark](https://www.indobenchmark.com/), the model achieve an f1-macro of 72.05%, accuracy of 71.81%, precision of 72.47% and recall of 71.94%.
## Model
The model was trained on 7 epochs with learning rate 2e-5. Achieved different metrics as shown below.
| Epoch | Training Loss | Validation Loss | Accuracy | F1 | Precision | Recall |
|-------|---------------|-----------------|----------|----------|-----------|----------|
| 1 | 1.300700 | 1.005149 | 0.622727 | 0.601846 | 0.640845 | 0.611144 |
| 2 | 0.806300 | 0.841953 | 0.686364 | 0.694096 | 0.701984 | 0.696657 |
| 3 | 0.591900 | 0.796794 | 0.686364 | 0.696573 | 0.707520 | 0.691671 |
| 4 | 0.441200 | 0.782094 | 0.722727 | 0.724359 | 0.725985 | 0.730229 |
| 5 | 0.334700 | 0.809931 | 0.711364 | 0.720550 | 0.718318 | 0.724608 |
| 6 | 0.268400 | 0.812771 | 0.718182 | 0.724192 | 0.721222 | 0.729195 |
| 7 | 0.226000 | 0.828461 | 0.725000 | 0.733625 | 0.731709 | 0.735800 |
## How to Use
### As Text Classifier
```python
from transformers import pipeline
pretrained_name = "StevenLimcorn/indonesian-roberta-base-emotion-classifier"
nlp = pipeline(
"sentiment-analysis",
model=pretrained_name,
tokenizer=pretrained_name
)
nlp("Hal-hal baik akan datang.")
```
## Disclaimer
Do consider the biases which come from both the pre-trained RoBERTa model and the `EmoT` dataset that may be carried over into the results of this model.
## Author
Indonesian RoBERTa Base Emotion Classifier was trained and evaluated by [Steven Limcorn](https://github.com/stevenlimcorn). All computation and development are done on Google Colaboratory using their free GPU access.
If used, please cite
```bibtex
@misc {steven_limcorn_2023,
author = { {Steven Limcorn} },
title = { indonesian-roberta-base-emotion-classifier (Revision e8a9cb9) },
year = 2023,
url = { https://huggingface.co/StevenLimcorn/indonesian-roberta-base-emotion-classifier },
doi = { 10.57967/hf/0681 },
publisher = { Hugging Face }
}
``` |