Scholarlink

<p>Drug-drug interactions (DDIs) play a critical role in several biomedical applications, particularly in pharmacovigilance. While neural networks have shown promise in DDI extraction, existing methods often overlook salient information. To address this challenge, we propose a two-channel DDI relationship extraction model based on cross attention - DCCA-IE. This model consists of a sequence channel using BioBERT-CNN and a graph channel integrating BioBERT-LSTM-GAT. The dual channel architecture is crucial to capture both sequential and relational information present in biomedical texts. Furthermore, by employing a cross attention mechanism, we effectively integrate information from the dual channels, resulting in a comprehensive representation for DDI relationship extraction. Experimental evaluation on a benchmark dataset shows that our method outperforms state-of-the-art methods in DDI extraction.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Dual channel drug-drug interactions extraction based on cross attention

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月23日

查看全文
  • Yaxun Jia,
  • Zhu Yuan,
  • Yunchao Gong,
  • Haixiang Yang,
  • Zuo-lin Xiang

摘要

Drug-drug interactions (DDIs) play a critical role in several biomedical applications, particularly in pharmacovigilance. While neural networks have shown promise in DDI extraction, existing methods often overlook salient information. To address this challenge, we propose a two-channel DDI relationship extraction model based on cross attention - DCCA-IE. This model consists of a sequence channel using BioBERT-CNN and a graph channel integrating BioBERT-LSTM-GAT. The dual channel architecture is crucial to capture both sequential and relational information present in biomedical texts. Furthermore, by employing a cross attention mechanism, we effectively integrate information from the dual channels, resulting in a comprehensive representation for DDI relationship extraction. Experimental evaluation on a benchmark dataset shows that our method outperforms state-of-the-art methods in DDI extraction.