Indonesian RoBERTa Base IndoLEM Sentiment Classifier

Indonesian RoBERTa Base IndoLEM Sentiment Classifier is a sentiment-text-classification model based on the RoBERTa model. The model was originally the pre-trained Indonesian RoBERTa Base model, which is then fine-tuned on indolem 's Sentiment Analysis dataset consisting of Indonesian tweets and hotel reviews (Koto et al., 2020).

A 5-fold cross-validation experiment was performed, with splits provided by the original dataset authors. This model was trained on fold 0. You can find models trained on fold 0 , fold 1 , fold 2 , fold 3 , and fold 4 , in their respective links.

On fold 0 , the model achieved an F1 of 86.42% on dev/validation and 83.12% on test. On all 5 folds , the models achieved an average F1 of 84.14% on dev/validation and 84.64% on test.

Hugging Face's Trainer class from the 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

Evaluation Results

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

How to Use

As Text Classifier

from transformers import pipeline

pretrained_name = "w11wo/indonesian-roberta-base-indolem-sentiment-classifier-fold-0"

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

nlp("Pelayanan hotel ini sangat baik.")

Disclaimer

Do consider the biases which come from both the pre-trained RoBERTa model and IndoLEM 's Sentiment Analysis dataset that may be carried over into the results of this model.

Author

Indonesian RoBERTa Base IndoLEM Sentiment Classifier was trained and evaluated by Wilson Wongso . All computation and development are done on Google Colaboratory using their free GPU access.