Team Fusion@ SU@ BC8 SympTEMIST赛道：基于Transformer的症状识别与链接方法
计算与语言
2026-04-09 v1 人工智能
摘要
本文提出了一种基于Transformer的方法来解决SympTEMIST命名实体识别（NER）和实体链接（EL）任务。对于NER，我们在增强的训练集上微调了一个基于RoBERTa的（1）token级分类器，并附加BiLSTM和CRF层。实体链接通过使用跨语言的SapBERT XLMR-Large（2）生成候选实体，并通过与知识库计算余弦相似度来完成。知识库的选择被证明对模型准确率具有最高影响。
引用
@article{arxiv.2604.06424,
title = {Team Fusion@ SU@ BC8 SympTEMIST track: transformer-based approach for symptom recognition and linking},
author = {Georgi Grazhdanski and Sylvia Vassileva and Ivan Koychev and Svetla Boytcheva},
journal= {arXiv preprint arXiv:2604.06424},
year = {2026}
}
备注
6 pages, 3 tables, Proceedings of the BioCreative VIII Challenge and Workshop: Curation and Evaluation in the era of Generative Models, American Medical Informatics Association 2023 Annual Symposium