Zero-shot relation extraction model via multi-view structural-semantic alignment

doi:10.11772/j.issn.1001-9081.2025121600

Journal of Computer Applications

Received:2026-01-08 Revised:2026-03-13 Online:2026-03-27 Published:2026-03-27
Contact: Guan-Feng LI

基于多视角结构语义对齐的零样本关系抽取模型

肖毓航¹,李贯峰²,陈昱胤¹,秦晶¹

1. Ningxia University
2. 宁夏大学

通讯作者: 李贯峰
基金资助:
不确定RDF知识图谱数据查询关键技术研究;融合逻辑规则的不确定知识图谱表示学习方法研究;基于图神经网络的知识图谱多跳推理方法研究

Abstract

Abstract: To address relation confusion in zero-shot relation extraction caused by missing labeled samples for unseen relations and semantic-structural mismatch between instances and relation descriptions, a zero-shot relation extraction model via multi-view structural-semantic alignment DVBE-ZSRE(Dual-View Boundary-Enhanced Zero-shot Relation Extraction) was proposed. First, instance representations were constructed from three perspectives, including global semantics, contextual semantics, and entity-boundary structure. Then, structured representations of relation descriptions were obtained by introducing virtual head and tail entities. Next, a structural-view matching mechanism was designed to fuse structural and semantic similarities for top-k candidate relation retrieval. Finally, a bidirectional multi-view contrastive learning objective was formulated to align instance-side and description-side representations, and fine-grained classification was conducted with joint encoding within the candidate set. Experimental results demonstrate that the proposed model achieves F1 scores of 91.91% and 95.46% on Wiki-ZSL and FewRel at m=5, respectively, and 79.92% and 86.28% at m=15, respectively; on FewRel at m=15, an improvement of 6.18 percentage points over EMMA (Efficient Multi-grained Matching Approach) is obtained. an additional gain of 1.49 percentage points over the best-performing compared method under the same setting is achieved. These results indicate that structural–semantic alignment and cross-space consistency constraints effectively alleviate confusion among similar relations and improve generalization to unseen relations.

Key words: Keywords: relation extraction, multi-view Encoding, structural-view matching, contrastive learning, structural–semantic

摘要： 针对零样本关系抽取中目标关系缺乏标注样本、且实例文本与关系描述在语义与结构层面不一致导致关系混淆的问题，提出一种基于多视角结构语义对齐的零样本关系抽取模型DVBE-ZSRE(Dual-View Boundary-Enhanced Zero-shot Relation Extraction)。首先，从全局语义、上下文语义与实体边界结构三个视角构建实例表示；其次，引入虚拟头尾实体得到关系描述的结构化表示；再次，设计结构视角匹配机制融合结构与语义相似度以召回top-k候选关系；最后，构建双向多视角对比学习约束实例侧与描述侧表征对齐，并在候选集合内通过联合编码完成细粒度分类。实验结果表明，在Wiki-ZSL与FewRel数据集上，本模型在m=5设置下的F1分别达到91.91%与95.46%，在m=15设置下分别达到79.92%与86.28%；其中在FewRel的m=15设置下，相比强基线EMMA(Efficient Multi-grained Matching Approach)模型F1提升6.18个百分点，并相较同设置下F1最高的对比方法提升1.49个百分点。上述结果表明，所提模型能够通过结构语义对齐与跨空间一致性约束缓解相似关系混淆，并在未见关系数量增大时保持更稳定的泛化性能。

关键词: 零样本关系抽取, 多视角编码, 结构视角匹配, 对比学习, 结构语义对齐

CLC Number:

TP391.1

肖毓航李贯峰陈昱胤秦晶. 基于多视角结构语义对齐的零样本关系抽取模型[J]. 《计算机应用》唯一官方网站, DOI: 10.11772/j.issn.1001-9081.2025121600.

[1]	Dirui ZHANG, Jiayu LIN, Zuhong LIANG. Supervised contrastive generative sentiment analysis method with uncertainty-aware unlikelihood learning [J]. Journal of Computer Applications, 2026, 46(5): 1416-1423.
[2]	Kun FU, Haoyu WEI, Weijing LIU, Xing DANG, Zezheng LIU, Jianwei LI. Graph neural network framework for topology semantic dual-domain collaboration [J]. Journal of Computer Applications, 2026, 46(5): 1378-1387.
[3]	Haihua ZHAO, Yijun HU, Rui TANG, Xian MO. Multimodal recommendation method based on semantic fusion and contrast enhancement [J]. Journal of Computer Applications, 2026, 46(4): 1058-1068.
[4]	Xiaoxia LIU, Liqun KUANG, Song WANG, Shichao JIAO, Huiyan HAN, Fengguang XIONG. Multi-scale spatio-temporal decoupling for contrastive learning of skeleton action recognition [J]. Journal of Computer Applications, 2026, 46(3): 767-774.
[5]	Yuhang XIAO, Guanfeng LI, Yuyin CHEN, Jing QIN. Few-shot relation extraction model with graph-based multi-view contrastive learning [J]. Journal of Computer Applications, 2026, 46(3): 732-740.
[6]	Hu LUO, Mingshu ZHANG. Rumor detection method based on cross-modal attention mechanism and contrastive learning [J]. Journal of Computer Applications, 2026, 46(2): 361-367.
[7]	Limei DONG, Yanzi LI, Jiayin LI, Li XU. Neighborhood-enhanced unsupervised graph anomaly detection [J]. Journal of Computer Applications, 2026, 46(2): 458-466.
[8]	Ziyang CHENG, Ruizhang HUANG, Jingjing XUE. Deep evolutionary topic clustering model [J]. Journal of Computer Applications, 2026, 46(1): 85-94.
[9]	Wen LI, Kairong LI, Kai YANG. Subgraph-aware contrastive learning with data augmentation [J]. Journal of Computer Applications, 2026, 46(1): 1-9.
[10]	Xingyao YANG, Zheng QI, Jiong YU, Zulian ZHANG, Shuai MA, Hongtao SHEN. Session-based recommendation model based on time-aware and space-enhanced dual channel graph neural network [J]. Journal of Computer Applications, 2026, 46(1): 104-112.
[11]	Zhixiong XU, Bo LI, Xiaoyong BIAN, Qiren HU. Adversarial sample embedded attention U-Net for 3D medical image segmentation [J]. Journal of Computer Applications, 2025, 45(9): 3011-3016.
[12]	Chao LIU, Yanhua YU. Knowledge-aware recommendation model combining denoising strategy and multi-view contrastive learning [J]. Journal of Computer Applications, 2025, 45(9): 2827-2837.
[13]	Zhiyuan WANG, Tao PENG, Jie YANG. Integrating internal and external data for out-of-distribution detection training and testing [J]. Journal of Computer Applications, 2025, 45(8): 2497-2506.
[14]	Jin XIE, Surong CHU, Yan QIANG, Juanjuan ZHAO, Hua ZHANG, Yong GAO. Dual-branch distribution consistency contrastive learning model for hard negative sample identification in chest X-rays [J]. Journal of Computer Applications, 2025, 45(7): 2369-2377.
[15]	Zhenzhou WANG, Fangfang GUO, Jingfang SU, He SU, Jianchao WANG. Robustness optimization method of visual model for intelligent inspection [J]. Journal of Computer Applications, 2025, 45(7): 2361-2368.