vipubmed-deberta-base-en-vimedner-ner-vi

ViPubMedDeBERTa-base-En-ViMedNER-Vi

Fine-tuned manhtt-079/vipubmed-deberta-base for Vietnamese biomedical Named Entity Recognition with UMLS semantic-type labels, trained on the Vietnamese split of En-ViMedNER.

• Input: word-segmented Vietnamese biomedical text
• Output: BIO tags over 21 UMLS semantic types
• Details: read our paper at ACL Anthology or arXiv

Preprocessing Requirement: Word Segmentation

Vietnamese input should be word-segmented with VnCoreNLP (RDRSegmenter) before inference, since the model was fine-tuned on word-segmented text.

How to Use

from transformers import AutoTokenizer, AutoModelForTokenClassification
from transformers import pipeline

model_id = "phuong-tk-nguyen/vipubmed-deberta-base-en-vimedner-ner-vi"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForTokenClassification.from_pretrained(model_id)

ner = pipeline("token-classification", model=model, tokenizer=tokenizer,
               aggregation_strategy="simple")

segmented_text = "Nhóm chẩn_đoán phổ_biến nhất dẫn đến nhập_viện là các bệnh_lý tim_mạch."

print(ner(segmented_text))
# [{'entity_group': 'T033', 'score': np.float32(0.7057605), 'word': 'chẩn_đoán', 'start': 4, 'end': 14}, {'entity_group': 'T058', 'score': np.float32(0.8302413), 'word': '', 'start': 36, 'end': 37}, {'entity_group': 'T058', 'score': np.float32(0.4629524), 'word': 'nhập_viện', 'start': 37, 'end': 46}, {'entity_group': 'T038', 'score': np.float32(0.7310076), 'word': 'các bệnh_lý tim', 'start': 49, 'end': 65}]

Related

Another model variation: phuong-tk-nguyen/vipubmed-deberta-xsmall-en-vimedner-ner-vi