seleznevchamp1/rubert-tiny2-NER-augmented

---

language: ru tags: - NER - BERT - augmentation - ruBERT license: mit datasets: - custom: hh.ru metrics: - f1

ruBERT-tiny2 (Augmented) for NER on Russian Job Postings

Compact ruBERT model fine-tuned with data augmentation to extract Skill and Knowledge entities.

Base model: cointegrated/ruBert-tiny2
Training: 250 manually annotated + 250 augmented samples (total 500).
F1: 0.52

Usage

from transformers import AutoTokenizer, AutoModelForTokenClassification, pipeline
model = AutoModelForTokenClassification.from_pretrained("seleznevchamp1/rubert-tiny2-NER-augmented")
tok = AutoTokenizer.from_pretrained("seleznevchamp1/rubert-tiny2-NER-augmented")
nlp = pipeline("ner", model=model, tokenizer=tok)
print(nlp("Опыт с Python и управление командой"))