Harnessing Causal Structure Alignment for Enhanced Cross-Domain Named Entity Recognition

被引:0
作者
Liu, Xiaoming [1 ,2 ]
Cao, Mengyuan [1 ,3 ]
Yang, Guan [1 ,4 ]
Liu, Jie [2 ,5 ]
Liu, Yang [6 ]
Wang, Hang [1 ,3 ]
机构
[1] Zhongyuan Univ Technol, Sch Comp Sci, Zhengzhou 450007, Peoples R China
[2] China Language Intelligence Res Ctr, Beijing 100089, Peoples R China
[3] Henan Key Lab Publ Opin Intelligent Anal, Zhengzhou 450007, Peoples R China
[4] Zhengzhou Key Lab Text Proc & Image Understanding, Zhengzhou 450007, Peoples R China
[5] North China Univ Technol, Sch Informat Sci, Beijing 100144, Peoples R China
[6] Xidian Univ, Sch Telecommun Engn, Xian 710071, Peoples R China
关键词
cross-domain named entity recognition; transfer learning; causal inference; feature interactions; causally invariant knowledge;
D O I
10.3390/electronics13010067
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Cross-domain named entity recognition (NER) is a crucial task in various practical applications, particularly when faced with the challenge of limited data availability in target domains. Existing methodologies primarily depend on feature representation or model parameter sharing mechanisms to enable the transfer of entity recognition capabilities across domains. However, these approaches often ignore the latent causal relationships inherent in invariant features. To address this limitation, we propose a novel framework, the Causal Structure Alignment-based Cross-Domain Named Entity Recognition (CSA-NER) framework, designed to harness the causally invariant features within causal structures to enhance the cross-domain transfer of entity recognition competence. Initially, CSA-NER constructs a causal feature graph utilizing causal discovery to ascertain causal relationships between entities and contextual features across source and target domains. Subsequently, it performs graph structure alignment to extract causal invariant knowledge across domains via the graph optimal transport (GOT) method. Finally, the acquired causal invariant knowledge is refined and utilized through the integration of Gated Attention Units (GAUs). Comprehensive experiments conducted on five English datasets and a specific CD-NER dataset exhibit a notable improvement in the average performance of the CSA-NER model in comparison to existing cross-domain methods. These findings underscore the significance of unearthing and employing latent causal invariant knowledge to effectively augment the entity recognition capabilities in target domains, thereby contributing a robust methodology to the broader realm of cross-domain natural language processing.
引用
收藏
页数:18
相关论文
共 33 条
[21]  
Ng I, 2019, Arxiv, DOI arXiv:1911.07420
[22]  
Kingma DP, 2014, Arxiv, DOI [arXiv:1312.6114, 10.48550/arXiv.1312.6114]
[23]  
Paszke A, 2019, ADV NEUR IN, V32
[24]  
Rojas-Carulla M, 2018, J MACH LEARN RES, V19
[25]   Named Entity Recognition for Few-Shot Power Dispatch Based on Multi-Task [J].
Tan, Zhixiang ;
Chen, Yan ;
Liang, Zengfu ;
Meng, Qi ;
Lin, Dezhao .
ELECTRONICS, 2023, 12 (16)
[26]  
Tong M., 2021, Learning from Miscellaneous Other-Class Words for Few-Shot Named Entity Recognition, P6236
[27]   Stylometric Fake News Detection Based on Natural Language Processing Using Named Entity Recognition: In-Domain and Cross-Domain Analysis [J].
Tsai, Chih-Ming .
ELECTRONICS, 2023, 12 (17)
[28]  
Van Lint J. H., 2001, A Course in Combinatorics
[29]  
Wei D., 2020, Advances in Neural Information Processing Systems, V33, P3895
[30]   Learning Causal Representations for Robust Domain Adaptation [J].
Yang, Shuai ;
Yu, Kui ;
Cao, Fuyuan ;
Liu, Lin ;
Wang, Hao ;
Li, Jiuyong .
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING, 2023, 35 (03) :2750-2764