Scholarlink

Disambiguation is a crucial task in natural language processing, especially for address segmentation, where accurately parsing complex address formats is essential. Traditional methods like CRF or CNN alone have limitations—CRF excels at sequence labeling but lacks deep feature extraction, while CNN captures local patterns but struggles with long-range dependencies. To address this, we propose a hybrid CRF-CNN model that combines their strengths. Experiments show it achieves 95.8% accuracy, outperforming standalone CRF (95.5%) and CNN (94.0%), demonstrating stronger robustness in handling diverse address structures. This hybrid approach offers a more effective solution for real-world disambiguation tasks.

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

Conditional Random Fields and Convolutional Neural Networks Fusion Approach to Address Disambiguation

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Jingyi Xiao,
  • Zhitao Wei,
  • Qing Shen

摘要

Disambiguation is a crucial task in natural language processing, especially for address segmentation, where accurately parsing complex address formats is essential. Traditional methods like CRF or CNN alone have limitations—CRF excels at sequence labeling but lacks deep feature extraction, while CNN captures local patterns but struggles with long-range dependencies. To address this, we propose a hybrid CRF-CNN model that combines their strengths. Experiments show it achieves 95.8% accuracy, outperforming standalone CRF (95.5%) and CNN (94.0%), demonstrating stronger robustness in handling diverse address structures. This hybrid approach offers a more effective solution for real-world disambiguation tasks.