Sentiment Analysis with RoBERTa Models

This repository contains code for training a sentiment analysis model using the Indonesian RoBERTa base model. The sentiment analysis is performed on an SmSA dataset obtained from the IndoNLU datasets.

Setup

1. Clone the repository:

   git clone https://github.com/agustyawan-arif/sentiment-roberta-id.git
   cd sentiment-roberta-id

2. Install dependencies:

   pip install -r requirements.txt

Inference

After training, you can use the trained model for sentiment analysis with the inference script. Run the script as follows:

python inference.py

This script allows you to interactively enter text for sentiment analysis, providing the predicted sentiment label and confidence score. Change model_name in line 22 to your model name if you have trained your own model

Configuration

The training and model configuration are specified in the config.conf file. Modify this file to adjust training parameters, output directories, and other settings.

Training

Run the training script to train the sentiment analysis model:

python src/trainer.py

The script uses the specified configuration file (config.conf) for training parameters.

Acknowledgments

• Hugging Face Model Cards
• Hugging Face Space
• Hugging Face Transformers
• IndoNLU Datasets