chayuto/thai-job-ner-wangchanberta

Thai Job NER — Fine-tuned WangchanBERTa

Named Entity Recognition model for extracting structured HR data from informal Thai job postings (e.g., Facebook groups, Line chats). Fine-tuned from wangchanberta-base-att-spm-uncased (110M params).

Model Description

This model extracts 7 entity types from Thai job-related text:

Entity	Description	Example
HARD_SKILL	Skills or procedures	ดูแลผู้สูงอายุ, CPR, Python
PERSON	Names	คุณสมชาย, พี่แจน
LOCATION	Places	สีลม, ลาดพร้าว, บางนา
COMPENSATION	Pay amounts	18,000 บาท/เดือน
EMPLOYMENT_TERMS	Job structure	part-time, กะกลางวัน
CONTACT	Phone, Line, email	081-234-5678, @care123
DEMOGRAPHIC	Age, gender	อายุ 25-40, หญิง

Usage

from transformers import AutoTokenizer, AutoModelForTokenClassification, pipeline

model_name = "chayuto/thai-job-ner-wangchanberta"
ner = pipeline("ner", model=model_name, aggregation_strategy="simple")

text = "รับสมัครคนดูแลผู้สูงอายุ ย่านสีลม เงินเดือน 18,000 บาท โทร 081-234-5678"
results = ner(text)
for entity in results:
    print(f"{entity['entity_group']}: {entity['word']} ({entity['score']:.2%})")

Training

• Base model: airesearch/wangchanberta-base-att-spm-uncased (CamemBERT architecture, 110M params)
• Training data: 1,253 Thai job posts (synthetic silver labels from GPT-4o, fuzzy-aligned to IOB2) — Dataset on HuggingFace
• Hardware: Apple Silicon MPS backend, FP32
• Hyperparameters: LR=3e-5, warmup=0.1, batch=8, grad_accum=2, 15 epochs, class-weighted loss, label smoothing=0.05
• Training time: ~3 min 48 sec

Data Pipeline

Raw Thai text + GPT-4o entity extractions → fuzzy alignment with rapidfuzz + pythainlp TCC boundary snapping → subword token mapping via offset_mapping → IOB2-formatted HuggingFace Dataset.

Evaluation

Overall (Test Set, 126 examples)

Metric	Score
F1	0.897
Precision	0.850
Recall	0.949

Per-Entity F1

Entity	F1	Precision	Recall
CONTACT	0.962	0.942	0.983
LOCATION	0.959	0.928	0.991
EMPLOYMENT_TERMS	0.926	0.870	0.990
PERSON	0.907	0.861	0.958
HARD_SKILL	0.903	0.873	0.936
DEMOGRAPHIC	0.875	0.827	0.928
COMPENSATION	0.764	0.673	0.884

Links

• Model: chayuto/thai-job-ner-wangchanberta
• Dataset: chayuto/thai-job-ner-dataset
• Source Code: github.com/chayuto/thai-job-nlp-ner

Limitations

• Trained on synthetic data — may underperform on real-world posts with heavy emoji usage, OCR errors, or extreme colloquialism
• Thai-specific: limited English entity extraction capability
• 512 token max sequence length
• HARD_SKILL has the lowest F1 (0.761) due to open vocabulary and complex boundaries

Technical Notes

• FP16 is broken on MPS — always use FP32 for Apple Silicon training
• Uses offset_mapping to bypass WangchanBERTa's <_> space token misalignment in char_to_token()
• Thai Character Cluster (TCC) boundary snapping prevents Unicode grapheme splitting during alignment

License

MIT