Fine-tuned XLM-RoBERTa Base model for Named Entity Recognition (NER) on Indonesian news articles, with built-in rule-based post-processing for Indonesian text.
1from transformers import pipeline
2
3# Load the NER pipeline
4ner = pipeline("token-classification", model="tlabdev/ner-irish-roberta-base", aggregation_strategy="simple")
5
6# Run inference
7text = "Gubernur Jawa Barat meresmikan proyek senilai Rp 10 miliar."
8results = ner(text)
9
10# Display results
11for entity in results:
12 print(f"{entity['word']} -> {entity['entity_group']} (score: {entity['score']:.2f})")
1Learning rate: 3e-5
2Train batch size: 4
3Eval batch size: 8
4Gradient accumulation steps: 4
5Effective batch size: 16 (4 × 4)
6Epochs: 10
7Weight decay: 0.02
8LR scheduler: Linear
9Warmup ratio: 0.03
10Max gradient norm: 0.5
11Label smoothing: 0.05
12Mixed precision: FP16
13Optimizer: AdamW (default)
Base Model: xlm-roberta-base
Framework: Transformers 4.57.3, PyTorch 2.1.0+cu124
Language: Indonesian
Task: Token Classification (NER)
This fine-tuned model inherits the MIT License from the base model
XLM-RoBERTa.
1@misc{ner-irish-roberta-base,
2 author = {TLab Developer and Muhammad Faiz Khoiri},
3 title = {ner-irish-roberta-base},
4 year = {2025},
5 publisher = {HuggingFace},
6 howpublished = {\url{https://huggingface.co/tlabdev/ner-irish-roberta-base}}
7}
1@article{DBLP:journals/corr/abs-1911-02116,
2 author = {Alexis Conneau and
3 Kartikay Khandelwal and
4 Naman Goyal and
5 Vishrav Chaudhary and
6 Guillaume Wenzek and
7 Francisco Guzm{\'{a}}n and
8 Edouard Grave and
9 Myle Ott and
10 Luke Zettlemoyer and
11 Veselin Stoyanov},
12 title = {Unsupervised Cross-lingual Representation Learning at Scale},
13 journal = {CoRR},
14 volume = {abs/1911.02116},
15 year = {2019},
16 url = {http://arxiv.org/abs/1911.02116},
17 eprinttype = {arXiv},
18 eprint = {1911.02116}
19}
