伴随语义Web的发展，RDF数据不断增长。建立不同RDF数据集间的链接是语义Web研究中的核心问题，对于实现知识之网具有重要意义。大规模RDF数据集表现出异构性强、高维稀疏等特点，导致传统基于名称、结构等特征的RDF数据链接方法效果不佳。表示学习技术能将RDF数据集嵌入一个低维语义向量空间，消除不同数据集间的符号语义鸿沟，因此基于表示学习的RDF数据链接方法颇具潜力。但现有表示学习模型主要面向单个RDF数据集，少数基于表示学习的链接方法较为简单，也未涉及除元素链接以外的复杂链接。本项目力图在针对RDF数据链接的表示学习方法、基于表示学习的元素链接方法、基于表示学习的RDF数据复杂链接方法等方面有所创新，还拟研发一个新型的基于表示学习的查询驱动型RDF数据链接系统。形成的新方法、技术及系统将提高RDF数据链接的质量和性能，推动语义Web发展，对于促进智能软件等相关领域的技术进步也有积极意义。

Along with the rapid development of Semantic Web, the amount of RDF data is continuously increasing. Establishing linkage between different RDF datasets is a fundamental problem in Semantic Web research and important to the realization of the Web of knowledge. Large-scale RDF datasets are often strongly-heterogeneous, highly-dimensional and very sparse, causing that traditional RDF data linkage methods using features like name and structure cannot achieve satisfactory results. Representation learning techniques are capable of embedding an RDF dataset into a low-dimensional semantic vector space, which resolve the symbolic gap between different RDF datasets, thus representation learning based RDF data linkage methods have great potential. However, current representation learning models mainly focus on a single RDF dataset, a few existing linkage methods based on representation learning are relatively simple and have not considered complex forms of linkage other than RDF element linkage yet. In this project, we look forward to making contributions to representation learning methods for RDF data linkage, representation learning based RDF element linkage methods, representation learning based RDF data complex linkage methods, etc. Additionally, we plan to develop a new query-driven RDF data linkage system based on representation learning. The outcome innovative methods, techniques and system will improve the quality and performance of RDF data linkage, benefit the development of Semantic Web, and also promote the technical progress of intelligent software and other related areas.