[1]董文永,梁智学,周孟强,等.融合时间编码和时间分布注意力的推理模型[J].计算机技术与发展,2025,(06):182-188.[doi:10.20165/j.cnki.ISSN1673-629X.2025.0014]
　DONG Wen-yong,LIANG Zhi-xue,ZHOU Meng-qiang,et al.A Reasoning Model Integrating Time Encoding and Time Distribution Attention[J].,2025,(06):182-188.[doi:10.20165/j.cnki.ISSN1673-629X.2025.0014]
点击复制

融合时间编码和时间分布注意力的推理模型()

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
期数:
2025年06期
页码:
182-188
栏目:
人工智能
出版日期:
2025-06-10

文章信息/Info

Title:
A Reasoning Model Integrating Time Encoding and Time Distribution Attention
文章编号:
1673-629X(2025)06-0182-07
作者:
董文永1; 2; 梁智学2; 周孟强2; 贾亚洁3
1. 新疆政法学院 信息网络安全学院,新疆 图木舒克 843900;
2. 武汉大学 计算机学院,湖北 武汉 430072;
3. 武汉大学 国家网络安全学院,湖北 武汉 430072
Author(s):
DONG Wen-yong1; 2; LIANG Zhi-xue2; ZHOU Meng-qiang2; JIA Ya-jie3
1. School of Information Network Security,Xinjiang University of Political Science and Law,Tumushuke 843900,China;
2. School of Computer Science,Wuhan University,Wuhan 430072,China;
3. School of Cyber Science and Engineering,Wuhan University,Wuhan 430072,China
关键词:
时序知识图谱; 图卷积神经网络; 注意力机制; 时间编码; 知识推理
Keywords:
temporal knowledge graphs; graph convolutional neural networks; attention mechanism; time encoding; knowledge reasoning
分类号:
TP391.1
DOI:
10.20165/j.cnki.ISSN1673-629X.2025.0014
摘要:
为了充分利用时序知识图谱数据的时间维度特征,解决时序知识图谱时间信息挖掘不充分的问题,提出了一种融合时间编码和时间分布注意力的推理模型 TETDM,该模型结合长期历史信息和短期历史信息来捕捉不同模式的历史事件对推理目标的影响。 首先,模型将较长一段历史信息中对应实体或关系出现的频率作为推理模型的长期约束条件。 其次,模型通过时间感知编码和时间分布注意力机制对短期历史信息进行编码,深入挖掘时序知识图谱数据的时间维度特征。 其中,时间感知编码器能够精准捕捉时间信息对实体和关系的影响,从而生成带有时间特征的实体和关系表示。 时间分布注意力机制通过在不同子图中建模每个重复事实的注意力,而不是仅仅学习它们的表示,来学习历史重复事件的变量分布。 通过编码时间特征对实体和关系的影响以及编码不同时间戳下历史信息的贡献程度,可以获得在时间维度上更为精准的实体和关系的嵌入表示,从而提高模型的推理能力。 在 ICEWS14、ICEWS05-15、ICEWS18 以及 GDELT 数据集上的结果进一步证实,TETDM 提出的时间编码模块和时间分布注意力模块均可以提升模型的推理性能。
Abstract:
To fully leverage the temporal characteristics of temporal knowledge graph (TKG) data and address the issue of insufficient u-tilization of temporal information, we propose a reasoning model named TETDM ( Temporal Encoding and Temporal Distribution Attention-based Reasoning Model),which integrates long-term and short-term historical information to capture the impact of different historical event patterns on inference targets. Specifically,the model incorporates the frequency of entities or relations over an extended historical period as a long-term constraint for inference. Furthermore,it encodes short-term historical information using a temporal-aware encoder and a temporal distribution attention mechanism to deeply explore the temporal characteristics of TKG data. The temporal-aware encoder accurately captures the influence of temporal information on entities and relations,generating temporal feature-enriched representations of entities and relations. The temporal distribution attention mechanism, instead of merely learning representations of repeated facts,models the attention of each repeated fact within different subgraphs to learn the variable distribution of historical repetitive events. By encoding the influence of temporal features on entities and relations,as well as the contribution of historical information across different timestamps,the model produces more precise temporal embeddings of entities and relations,thereby enhancing its reasoning ca-pability. Experimental results on ICEWS14,ICEWS05 -15,ICEWS18,and GDELT datasets further demonstrate that both the temporal encoding module and the temporal distribution attention module proposed by TETDM can improve the reasoning performance of the model.

相似文献/References:

[1]刘 臣,李 秋,郝宇辰.基于图卷积神经网络的在线社区行为预测[J].计算机技术与发展,2021,31(04):28.[doi:10. 3969 / j. issn. 1673-629X. 2021. 04. 005]
　LIU Chen,LI Qiu,HAO Yu-chen.Prediction of Collaborative Behavior of Users in Online KnowledgeCommunities Based on Graph Convolutional Neural Network[J].,2021,31(06):28.[doi:10. 3969 / j. issn. 1673-629X. 2021. 04. 005]
[2]李社蕾,杨博雄,陆娇娇.谱图傅里叶变换与谱图小波变换基分析研究[J].计算机技术与发展,2021,31(05):85.[doi:10. 3969 / j. issn. 1673-629X. 2021. 05. 015]
　,,et al.ＡｎａｌｙｓｉｓａｎｄＳｔｕｄｙｏｆＳｐｅｃｔｒａｌＦｏｕｒｉｅｒＴｒａｎｓｆｏｒｍ ａｎｄＳｐｅｃｔｒａｌＷａｖｅｌｅｔＴｒａｎｓｆｏｒｍ Ｂａｓｉｓ[J].,2021,31(06):85.[doi:10. 3969 / j. issn. 1673-629X. 2021. 05. 015]
[3]建兵,杨 超*,刘方方,等.基于图卷积神经网络和 RoBERTa 的物流订单分类[J].计算机技术与发展,2023,33(10):195.[doi:10. 3969 / j. issn. 1673-629X. 2023. 10. 030]
　WANG Jian-bing,YANG Chao*,LIU Fang-fang,et al.A Logistics Order Classification Method Based on Graph Convolution Network and RoBERTa[J].,2023,33(06):195.[doi:10. 3969 / j. issn. 1673-629X. 2023. 10. 030]
[4]张 艳,肖文琛,张 博.基于双流骨架信息的人体动作识别方法[J].计算机技术与发展,2024,34(01):158.[doi:10. 3969 / j. issn. 1673-629X. 2024. 01. 023]
　ZHANG Yan,XIAO Wen-chen,ZHANG Bo.Human Action Recognition Method Based on Two-flow Skeleton Information[J].,2024,34(06):158.[doi:10. 3969 / j. issn. 1673-629X. 2024. 01. 023]
[5]张茜茜,陈 平.基于图卷积的离子液体 CO2 溶解度可解释性预测[J].计算机技术与发展,2024,34(02):134.[doi:10. 3969 / j. issn. 1673-629X. 2024. 02. 020]
　ZHANG Qian-qian,CHEN Ping.Interpretable Prediction of CO2 Solubility of Ionic Liquids Based on Graph Convolution Neural Network[J].,2024,34(06):134.[doi:10. 3969 / j. issn. 1673-629X. 2024. 02. 020]
[6]朱 越,黄海于,罗学义.基于实时视频流的 3D 人体姿势和形状估计[J].计算机技术与发展,2024,34(04):42.[doi:10. 3969 / j. issn. 1673-629X. 2024. 04. 007]
　ZHU Yue,HUANG Hai-yu,LUO Xue-yi.3D Human Pose and Shape Estimation Based on Live Video Streams[J].,2024,34(06):42.[doi:10. 3969 / j. issn. 1673-629X. 2024. 04. 007]
[7]胡翼,于海,郭鑫,等.融合句法信息的实体关系联合抽取[J].计算机技术与发展,2024,34(08):93.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0149]
　HU Yi,YU Hai,GUO Xin,et al.Syntactic Information Fused Joint Entity Relation Extraction[J].,2024,34(06):93.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0149]
[8]董云春,晏峻峰*.融合知识图谱与图卷积神经网络中医证型分类模型[J].计算机技术与发展,2025,(03):140.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0341]
　DONG Yun-chun,YAN Jun-feng*.A Classification Model of TCM Syndrome Based on Integration of Knowledge Graph and Graph Convolutional Network[J].,2025,(06):140.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0341]
[9]曹茂俊,王瑞芳,张国良.基于图神经网络的测井领域知识图谱实体对齐方法[J].计算机技术与发展,2025,(04):121.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0348]
　CAO Mao-jun,WANG Rui-fang,ZHANG Guo-liang.Entity Alignment Method for Logging Domain Data Asset Knowledge Graph Based on Graph Neural Networks[J].,2025,(06):121.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0348]

更新日期/Last Update: 2025-06-10