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-cross-encoder
Base/initializer: bikingSolo/prominer-ru-pretrained-cross-encoder

Training

Fine-tuned on candidate lists produced by the ProMiNER dense retriever, initialized from the dictionary-pretrained cross-encoder, and optimized with LambdaLoss. This is the best final model in the repository.

Selected MLflow parameters:

reranker_model_name_or_path: bikingSolo/prominer-ru-pretrained-cross-encoder
retriever_model_name_or_path: bikingSolo/prominer-ru-retriever
loss_name: lambdaloss
lambdaloss_weighting_scheme: ndcg2pp
epochs: 5
train_batch_size: 32
learning_rate: 1e-05
lr_scheduler_type: linear
weight_decay: 0.01
warmup_ratio: 0.1
max_seq_length: 384
train_candidate_pool_size: 20
dev_candidate_pool_size: 20
test_candidate_pool_size: 20
num_train_lists: 21547
num_train_pairs: 416508
selection_metric: Acc@1

According to the Acc@1 on dev, the best epoch is 3.

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

Evaluation

Metrics below are copied from the local MLflow run artifacts.

split	Acc@1	Acc@5	Acc@10	Acc@20	MRR
dev	0.7188498402555911	0.8293016887266088	0.8580556823368325	0.8667275216795983	0.7684439989494755
test	0.7339609483960948	0.8425732217573222			0.7795647373314744

Usage

python

from sentence_transformers import CrossEncoder

model = CrossEncoder("bikingSolo/prominer-ru-reranker", num_labels=1)
scores = model.predict([
    (
        "вестибулокохлеарный нерв",
        "слуховой нерв; вестибулокохлеарный нерв; nervus vestibulocochlearis [viii]",
    )
])

Intended Use

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

prominer-ru-retriever to retrieve candidate UMLS concepts.
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.