Scholarlink

<p>Named Entity Recognition (NER) plays a central role in modern data science and natural language processing pipelines, supporting tasks such as information retrieval, question answering, biomedical text mining, and knowledge graph construction. However, most existing NER systems struggle to achieve both high accuracy and computational efficiency across diverse domains. Large transformer-based models yield strong results but are often too resource-intensive for real-time or large-scale deployments, while lightweight alternatives frequently suffer from boundary detection errors and reduced generalization to complex or domain-specific entities. To address these challenges, we propose ECHANT (Efficient Context Hierarchical Attention Network), a novel deep learning architecture designed to balance recognition performance with computational efficiency. ECHANT employs a hierarchical attention mechanism spanning character, subword, word, and sentence levels, along with selective context integration modules that embed linguistic and domain-specific knowledge. Additionally, parameter sharing and low-precision computation reduce model complexity without compromising accuracy. Experimental results demonstrate the effectiveness of ECHANT, achieving F1-scores of 93.5% on CoNLL-2003 and 91.0% on OntoNotes 5.0, while reducing inference time by 48% compared to the BERT-Large-CRF baseline. These results highlight ECHANT’s potential as a scalable and efficient solution for NER within data-intensive and real-world analytics environments.</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

ECHANT: An efficient context hierarchical attention network for named entity recognition

  • 期刊论文

  • 发表日期：

    2026年3月14日

查看全文
  • Abdalrahman Hussein,
  • Aryender Singh,
  • Meerja Akhil Jabbar,
  • Chetansinh R. Vaghela,
  • T. Narmadha,
  • Manoranjan Parhi,
  • J. Albert Mayan,
  • Pradeep Kumar Mishra,
  • Vikas Wasson

摘要

Named Entity Recognition (NER) plays a central role in modern data science and natural language processing pipelines, supporting tasks such as information retrieval, question answering, biomedical text mining, and knowledge graph construction. However, most existing NER systems struggle to achieve both high accuracy and computational efficiency across diverse domains. Large transformer-based models yield strong results but are often too resource-intensive for real-time or large-scale deployments, while lightweight alternatives frequently suffer from boundary detection errors and reduced generalization to complex or domain-specific entities. To address these challenges, we propose ECHANT (Efficient Context Hierarchical Attention Network), a novel deep learning architecture designed to balance recognition performance with computational efficiency. ECHANT employs a hierarchical attention mechanism spanning character, subword, word, and sentence levels, along with selective context integration modules that embed linguistic and domain-specific knowledge. Additionally, parameter sharing and low-precision computation reduce model complexity without compromising accuracy. Experimental results demonstrate the effectiveness of ECHANT, achieving F1-scores of 93.5% on CoNLL-2003 and 91.0% on OntoNotes 5.0, while reducing inference time by 48% compared to the BERT-Large-CRF baseline. These results highlight ECHANT’s potential as a scalable and efficient solution for NER within data-intensive and real-world analytics environments.