Page 263 - 《软件学报》2025年第12期

软件学报 ISSN 1000-9825, CODEN RUXUEW E-mail: jos@iscas.ac.cn
2025,36(12):5644−5673 [doi: 10.13328/j.cnki.jos.007402] [CSTR: 32375.14.jos.007402] http://www.jos.org.cn
©中国科学院软件研究所版权所有. Tel: +86-10-62562563

*
多语义视图驱动的 OWL 知识图谱表示学习方法

杨建喜 1 , 谢江村 1 , 李 韧 1 , 杨小霞 2 , 肖 桥 2 , 蒋仕新 1 , 贺丽荣 1

1
(重庆交通大学 信息科学与工程学院, 重庆 400074)
2
(重庆交通大学 交通运输学院, 重庆 400074)
通信作者: 李韧, E-mail: renli@cqjtu.edu.cn

摘 要: 针对当前 OWL 知识表示学习方法存在的概念层和实例层复杂语义信息联合表征能力不足等问题, 提出一
种概念-属性-实例多语义视图驱动的 OWL 图谱知识表示学习方法 (MSV-KRL). 该方法采用“多语义视图划分、语
义感知自监督进阶训练、多任务联合表示学习”的 3 阶段架构. 首先, MSV-KRL 在 OWL2Vec*的基础上, 优化
OWL 到 RDF 图结构的映射策略, 提出 5 类细粒度语义视图划分策略. 其次, 通过语义视图内随机游走和标注属性
替换策略, 生成序列化进阶训练数据, 并开展预训练模型的自监督进阶训练, 以提升其面向多语义视图上下文的适
配能力. 最后, 在多任务学习框架下, 通过多语义视图预测任务联合优化损失, 实现对 OWL 知识图谱中概念、属性
和实例复杂语义有效表示学习. 实验结果表明, MSV-KRL 在多个基准数据集上的表现优于现有先进的知识表示学
习方法, 且能适配于多种语言模型, 有效提升 OWL 复杂语义的知识表示能力.
关键词: 知识表示学习; OWL 知识图谱; 多语义视图; 进阶训练; 多任务学习
中图法分类号: TP182

中文引用格式: 杨建喜, 谢江村, 李韧, 杨小霞, 肖桥, 蒋仕新, 贺丽荣. 多语义视图驱动的OWL知识图谱表示学习方法. 软件学报,
2025, 36(12): 5644–5673. http://www.jos.org.cn/1000-9825/7402.htm
英文引用格式: Yang JX, Xie JC, Li R, Yang XX, Xiao Q, Jiang SX, He LR. OWL Knowledge Graph Representation Learning Method
via Multi-semantic View. Ruan Jian Xue Bao/Journal of Software, 2025, 36(12): 5644–5673 (in Chinese). http://www.jos.org.cn/1000-
9825/7402.htm

OWL Knowledge Graph Representation Learning Method via Multi-semantic View
2
1
1
1
1
2
YANG Jian-Xi , XIE Jiang-Cun , LI Ren , YANG Xiao-Xia , XIAO Qiao , JIANG Shi-Xin , HE Li-Rong 1
1
(School of Information Science and Engineering, Chongqing Jiaotong University, Chongqing 400074, China)
2
(College of Traffic and Transportation, Chongqing Jiaotong University, Chongqing 400074, China)
Abstract: To address the issues in current OWL representation learning methods, which lack the ability to jointly represent complex
semantic information across both the concept layer and the instance layer, an OWL representation learning approach using multi-semantic
views of concepts, properties, and instances is proposed. The proposed method adopts a three-stage architecture including multi-semantic
views partitioning, semantic-aware self-supervised post-training, and joint multi-task representation learning. First, MSV-KRL optimizes the
mapping strategy from OWL to RDF graphs based on OWL2Vec*, and five fine-grained semantic view partitioning strategies are
proposed. Subsequently, serialized post-training data is generated through the random walk and annotated attribute replacement strategy.
The self-supervised post-training of the pre-trained model is then carried out to enhance adaptability to multi-semantic views. Finally, by
employing a multi-task learning strategy, the complex semantic representation learning of concepts, properties, and instances in OWL
graphs is achieved through joint optimization loss of multi-semantic view prediction tasks. Experimental results demonstrate that MSV-
KRL outperforms baseline representation learning methods on multiple benchmarks. MSV-KRL can be adapted to multiple language
models, significantly improving the knowledge representation capability of OWL’s complex semantics.

* 基金项目: 国家自然科学基金 (62003063); 重庆市教委科学技术研究重大项目 (KJZD-M202300703); 重庆市自然科学基金面上项目 (CSTB
2023NSCQ-MSX0145); 重庆市教委科学技术研究青年项目 (KJQN202200720)
收稿时间: 2024-07-03; 修改时间: 2024-10-31, 2024-12-30; 采用时间: 2025-02-11; jos 在线出版时间: 2025-07-09
CNKI 网络首发时间: 2025-07-10