| 基于深度学习的领域知识对齐模型研究:知识网络视角 |
| |
| 引用本文： | 余传明,李浩男,安璐.基于深度学习的领域知识对齐模型研究:知识网络视角[J].情报学报,2020,39(5):521-533. |
| |
| 作者姓名： | 余传明 李浩男 安璐 |
| |
| 作者单位： | 中南财经政法大学信息与安全工程学院,武汉 430073;中南财经政法大学统计与数学学院,武汉 430073;武汉大学信息管理学院,武汉 430072 |
| |
| 基金项目： | 国家自然科学基金面上项目“大数据环境下基于领域知识获取与对齐的观点检索研究”(71373286);国家自然科学基金重大课题“国家安全大数据综合信息集成与分析方法”(71790612)。 |
| |
| 摘 要： | 随着大数据的迅速发展,知识网络在不同语言、不同领域和不同模态等情境下呈现高度多样性和复杂性,如何对齐与整合多源情境下的异构知识网络,成为研究者所面临的严峻挑战。本文在知识网络深度表示学习的基础上,提出一种由知识网络构建、跨语言网络表示学习和统计机器学习三个模块构成的知识网络对齐(knowledge network alignment,KNA)模型。为验证模型的有效性,在中英文双语知识网络数据集上开展实证研究,借助于网络表示学习算法将异构知识网络表征到同一空间,利用已知的对齐链接来训练统计机器学习模型,并通过模型来预测未知的节点对齐链接。KNA模型在跨语言共词网络对齐任务中取得Precision@1值为0.7731,高于基线方法 (0.6806),验证了KNA模型在跨语言知识网络对齐上的有效性。研究结果对于改进知识网络的节点对齐效果,促进多源情境下的异构知识网络融合具有重要意义。
|
| 关 键 词： | 领域知识对齐 知识网络 深度学习 网络表示学习 统计机器学习 |
Research on Domain Knowledge Alignment Based on Deep Learning:Knowledge Network Perspective |
| |
| Yu Chuanming,Li Haonan,An Lu.Research on Domain Knowledge Alignment Based on Deep Learning:Knowledge Network Perspective[J].Journal of the China Society for Scientific andTechnical Information,2020,39(5):521-533. |
| |
| Authors: | Yu Chuanming Li Haonan An Lu |
| |
| Institution: | (School of Information and Safety Engineering,Zhongnan University of Economics and Law,Wuhan 430073;School of Statistics and Mathematics,Zhongnan University of Economics and Law,Wuhan 430073;School of Information Management,Wuhan University,Wuhan 430072) |
| |
| Abstract: | With the rapid development of big data,knowledge networks have become highly diverse and complex among different languages,domains,and modalities. How to align and integrate heterogeneous knowledge networks under multisource context is becoming a considerable challenge. Based on the deep representation learning of knowledge networks,this paper proposes a knowledge network alignment(KNA) model,which consists of three modules-the knowledge network construction module,the cross-lingual network representation module,and the statistical machine learning module.To verify the validity of the model,we conducted an empirical study on Chinese and English bilingual knowledge networks and projected heterogeneous knowledge networks into the same space. Based on this process,the statistical machine learning model was designed by using known alignment links between knowledge nodes among different networks,and unknown alignment links were predicted by the model. The KNA model obtained a Precision@1 value(0.7731) in the crosslingual word co-occurrence network alignment task,which is higher than that of the baseline method(0.6806),which verifies the validity of the KNA model in cross-lingual knowledge network alignment. The research results are of great significance for improving the accuracy of knowledge node alignment among different knowledge networks and for promoting the integration of heterogeneous knowledge networks under multi-source context. |
| |
| Keywords: | domain knowledge alignment knowledge network deep learning network representation learning statistical machine learning |
| 本文献已被 维普 万方数据 等数据库收录！ |
|
