ProMiNER Russian BioNNE-L Retriever

Dense Sentence Transformers retriever for Russian biomedical entity linking in the BioNNE-L task.

This model is part of ProMiNER, a Russian-track biomedical entity-linking system for BioNNE-L. The system links mentions from NEREL-BIO/BioNNE-L texts to UMLS concepts by combining dense retrieval and cross-encoder reranking.

• GitHub: https://github.com/bikingSolo/prominer
• Model type: sentence-transformers
• Base/initializer: andorei/BERGAMOT-multilingual-GAT

Training

Fine-tuned on Russian BioNNE-L mention-to-UMLS concept pairs with InfoNCE-style in-batch negatives and CUI-aware batching. This model is used to build candidate pools for the final reranker.

Hyperparameters:

• model_name: andorei/BERGAMOT-multilingual-GAT
• epochs: 10
• train_batch_size: 64
• learning_rate: 2e-05
• weight_decay: 0.01
• warmup_ratio: 0.1
• max_seq_length: 64
• num_train_pairs: 172593
• selection_metric: Acc@20

According to the Acc@20 on dev, the best epoch is 10.

Full local metadata exported from MLflow is included in prominer_metadata/.

Evaluation

Usage

from sentence_transformers import SentenceTransformer

model = SentenceTransformer("bikingSolo/prominer-ru-retriever")
mention_embeddings = model.encode(
    ["вестибулокохлеарный нерв"],
    normalize_embeddings=True,
)

Intended Use

This checkpoint is intended for research and reproducibility of the ProMiNER BioNNE-L Russian entity-linking pipeline. For the full system, use:

1. prominer-ru-retriever to retrieve candidate UMLS concepts.
2. prominer-ru-reranker to rerank those candidates with candidate-context profiles.

The dictionary-pretrained cross-encoder is primarily an intermediate checkpoint used to initialize the final reranker.

Data and Citation

Training and evaluation use BioNNE-L/NEREL-BIO resources and UMLS-derived terminology available in this repository's data layout. Cite the relevant NEREL-BIO and BioNNE-L papers when using this model.

Check https://github.com/bikingSolo/prominer for more info.