Official implementation of our paper titled "Fine-Tuning Deep Learning Model for Detecting Indonesian Traffic Accident Tweet"
Jupyter Notebook
1
81 commits
updated Dec 14, 2023
This repository is the official implementation of the conference paper titled Fine-Tuning Deep Learning Model for Detecting Indonesian Traffic Accident Tweet. The paper was presented at the 6th International Conference of Computer and Informatics Engineering (IC2IE) in 2023.
Authors: Rizky Adi, Fumiyo Fukumoto, Bassamtiano Renaufalgi Irnawan
To install requirements:
pip install -r requirements.txt
To train the model(s) in the paper, run this command:
python trainer_finetune.py --model <pre-trained model name> --learning_rate <learning rate> --batch_size <batch size> --max_length <input token max length> --cnn <is using pre-trained model + cnn architecture>
You can select between these pre-trained model name [IndoBERT, IndoBERTweet, IndoRoBERTa_OSCAR, IndoRoBERTa_Wiki].
These are the pre-trained models we used in our experiments:
Big thanks to these people who provide the pre-trained models weight for free.
Our model achieves the following performance :
| Model | Accuracy | F1-Score |
|---|---|---|
| Vanilla IndoBERT | 92% | 88.73% |
| Vanilla IndoBERTweet | 91.5% | 88.11% |
| Vanilla Indonesian RoBERTa (OSCAR) | 92% | 88.73% |
| Vanilla RoBERTa Indonesia (Wiki) | 87.5% | 82.76% |
| IndoBERT + CNN | 91% | 82.76 |
| IndoBERTweet + CNN | 93% | 90.67% |
| Indonesian RoBERTa (OSCAR) + CNN | 88.5% | 84.56% |
| RoBERTa Indonesia (Wiki) + CNN | 90% | 85.92% |
Please refer to our paper for the details
Please cite the following paper if you use our work.
@INPROCEEDINGS{10331034,
author={Adi, Rizky and Fukumoto, Fumiyo and Irnawan, Bassamtiano Renaufalgi},
booktitle={2023 6th International Conference of Computer and Informatics Engineering (IC2IE)},
title={Fine-Tuning Deep Learning Model for Detecting Indonesian Traffic Accident Tweet},
year={2023},
volume={},
number={},
pages={19-24},
doi={10.1109/IC2IE60547.2023.10331034}}
Official implementation of our paper titled "Fine-Tuning Deep Learning Model for Detecting Indonesian Traffic Accident Tweet"
Jupyter Notebook
1
81 commits
updated Dec 14, 2023
This repository is the official implementation of the conference paper titled Fine-Tuning Deep Learning Model for Detecting Indonesian Traffic Accident Tweet. The paper was presented at the 6th International Conference of Computer and Informatics Engineering (IC2IE) in 2023.
Authors: Rizky Adi, Fumiyo Fukumoto, Bassamtiano Renaufalgi Irnawan
To install requirements:
pip install -r requirements.txt
To train the model(s) in the paper, run this command:
python trainer_finetune.py --model <pre-trained model name> --learning_rate <learning rate> --batch_size <batch size> --max_length <input token max length> --cnn <is using pre-trained model + cnn architecture>
You can select between these pre-trained model name [IndoBERT, IndoBERTweet, IndoRoBERTa_OSCAR, IndoRoBERTa_Wiki].
These are the pre-trained models we used in our experiments:
Big thanks to these people who provide the pre-trained models weight for free.
Our model achieves the following performance :
| Model | Accuracy | F1-Score |
|---|---|---|
| Vanilla IndoBERT | 92% | 88.73% |
| Vanilla IndoBERTweet | 91.5% | 88.11% |
| Vanilla Indonesian RoBERTa (OSCAR) | 92% | 88.73% |
| Vanilla RoBERTa Indonesia (Wiki) | 87.5% | 82.76% |
| IndoBERT + CNN | 91% | 82.76 |
| IndoBERTweet + CNN | 93% | 90.67% |
| Indonesian RoBERTa (OSCAR) + CNN | 88.5% | 84.56% |
| RoBERTa Indonesia (Wiki) + CNN | 90% | 85.92% |
Please refer to our paper for the details
Please cite the following paper if you use our work.
@INPROCEEDINGS{10331034,
author={Adi, Rizky and Fukumoto, Fumiyo and Irnawan, Bassamtiano Renaufalgi},
booktitle={2023 6th International Conference of Computer and Informatics Engineering (IC2IE)},
title={Fine-Tuning Deep Learning Model for Detecting Indonesian Traffic Accident Tweet},
year={2023},
volume={},
number={},
pages={19-24},
doi={10.1109/IC2IE60547.2023.10331034}}