BioCreative VII-Track 1: A BERT-based System for Relation Extraction in Biomedical Text
BioCreative VII-Track 1: A BERT-based System for Relation Extraction in Biomedical Text
复制标题
DOI：
--
复制
发表时间：
2022
期刊：
影响因子：
--
通讯作者：
D. Mahendran;Sudhanshu Ranjan;Jia-Hong Tang;Mai H Nguyen;Bridget T. McInnes
中科院分区：
文献类型：
--
作者：
D. Mahendran;Sudhanshu Ranjan;Jia-Hong Tang;Mai H Nguyen;Bridget T. McInnes
-这篇文章描述了我们的团队参与了BioCreative VII挑战的第一个轨道，以自动检测化合物/药物和基因/蛋白质之间的关系。在这里，我们讨论了三个具有不同输入表示的基于上下文的语言模型：两个基于Transformers(BERT)的通用双向编码器表示模型和一个基于BioBERT的模型。对于这项任务，我们的最佳模型实现了总体精确度为0.55，召回率为0.52，测试集上的F1分数为0.54。
— This paper describes our team's participation in Track 1 of the BioCreative VII challenge to automatically detect relations between chemical compounds/drugs and genes/proteins. Here, we discuss the three contextualized language-based models with different input representations: two general Bidirectional Encoder Representations from Transformers (BERT)-based models and a BioBERT-based model. Our best model for this task achieved an overall Precision of 0.55, Recall of 0.52, and an F 1 score of 0.54 on the test set.