随着多媒体数据的爆发式增长,云端数据呈现出大规模多模态混合并存的特性.服务于数据分析的传统存储系统因为缺乏数据的语义管理而面临读取延时超长的挑战.针对图像和文本2种模态数据,在传统存储系统之上提出一种跨模态图文数据内容筛...随着多媒体数据的爆发式增长,云端数据呈现出大规模多模态混合并存的特性.服务于数据分析的传统存储系统因为缺乏数据的语义管理而面临读取延时超长的挑战.针对图像和文本2种模态数据,在传统存储系统之上提出一种跨模态图文数据内容筛选存储机制(cross-modal image and text content sifting storage,CITCSS),用于提供大规模在线相似性内容筛选服务,从存储系统层面缓解数据分析时必须从存储中读出所有数据的读带宽压力.机制分为离线与在线2个阶段.离线阶段中,引入基于自监督的生成对抗式Hash方法,系统利用这种方法生成语义元数据.然后,将元数据注入独立的元数据空间.最后,根据相似性Hash码间汉明距离能够度量语义距离的特点,利用Neo4j图数据库构建Hash元数据图谱,并在语义图谱中建立Hash码与存储路径之间的映射.在线阶段中,用户发送与分析相关的图像或文本,存储系统首先转化数据为Hash码.然后,在筛选半径内通过Hash元数据图谱寻找相似节点,进而找到相似文件的底层存储路径返回筛选数据.实验结果表明,与传统语义存储系统相比,CITCSS在召回率超过98%的性能下,读取延迟相对降低了99.07%~99.77%.展开更多
An ontology and metadata for online learning resource repository management is constructed. First, based on the analysis of the use-case diagram, the upper ontology is illustrated which includes resource library ontol...An ontology and metadata for online learning resource repository management is constructed. First, based on the analysis of the use-case diagram, the upper ontology is illustrated which includes resource library ontology and user ontology, and evaluated from its function and implementation; then the corresponding class diagram, resource description framework （RDF） schema and extensible markup language （XML） schema are given. Secondly, the metadata for online learning resource repository management is proposed based on the Dublin Core Metadata Initiative and the IEEE Learning Technologies Standards Committee Learning Object Metadata Working Group. Finally, the inference instance is shown, which proves the validity of ontology and metadata in online learning resource repository management.展开更多
Because of the importance of graphics and information within the domain of architecture, engineering and construction （AEC）, an appropriate combination of visualization technology and information management technolo...Because of the importance of graphics and information within the domain of architecture, engineering and construction （AEC）, an appropriate combination of visualization technology and information management technology is of utter importance in the development of appropriately supporting design and construction applications. Virtual environments, however, tend not to make this information available. The sparse number of applications that present additional information furthermore tend to limit their scope to pure construction information and do not incorporate information from loosely related knowledge domains, such as cultural heritage or architectural history information. We therefore started an investigation of two of the newest developments in these domains, namely game engine technology and semantic web technology. This paper documents part of this research, containing a review and comparison of the most prominent game engines and documenting our architectural semantic web. A short test-case illustrates how both can be combined to enhance information visualization for architectural design and construction.展开更多
文摘随着多媒体数据的爆发式增长,云端数据呈现出大规模多模态混合并存的特性.服务于数据分析的传统存储系统因为缺乏数据的语义管理而面临读取延时超长的挑战.针对图像和文本2种模态数据,在传统存储系统之上提出一种跨模态图文数据内容筛选存储机制(cross-modal image and text content sifting storage,CITCSS),用于提供大规模在线相似性内容筛选服务,从存储系统层面缓解数据分析时必须从存储中读出所有数据的读带宽压力.机制分为离线与在线2个阶段.离线阶段中,引入基于自监督的生成对抗式Hash方法,系统利用这种方法生成语义元数据.然后,将元数据注入独立的元数据空间.最后,根据相似性Hash码间汉明距离能够度量语义距离的特点,利用Neo4j图数据库构建Hash元数据图谱,并在语义图谱中建立Hash码与存储路径之间的映射.在线阶段中,用户发送与分析相关的图像或文本,存储系统首先转化数据为Hash码.然后,在筛选半径内通过Hash元数据图谱寻找相似节点,进而找到相似文件的底层存储路径返回筛选数据.实验结果表明,与传统语义存储系统相比,CITCSS在召回率超过98%的性能下,读取延迟相对降低了99.07%~99.77%.
基金The Advanced University Action Plan of the Minis-try of Education of China (2004XD-03).
文摘An ontology and metadata for online learning resource repository management is constructed. First, based on the analysis of the use-case diagram, the upper ontology is illustrated which includes resource library ontology and user ontology, and evaluated from its function and implementation; then the corresponding class diagram, resource description framework （RDF） schema and extensible markup language （XML） schema are given. Secondly, the metadata for online learning resource repository management is proposed based on the Dublin Core Metadata Initiative and the IEEE Learning Technologies Standards Committee Learning Object Metadata Working Group. Finally, the inference instance is shown, which proves the validity of ontology and metadata in online learning resource repository management.
文摘Because of the importance of graphics and information within the domain of architecture, engineering and construction （AEC）, an appropriate combination of visualization technology and information management technology is of utter importance in the development of appropriately supporting design and construction applications. Virtual environments, however, tend not to make this information available. The sparse number of applications that present additional information furthermore tend to limit their scope to pure construction information and do not incorporate information from loosely related knowledge domains, such as cultural heritage or architectural history information. We therefore started an investigation of two of the newest developments in these domains, namely game engine technology and semantic web technology. This paper documents part of this research, containing a review and comparison of the most prominent game engines and documenting our architectural semantic web. A short test-case illustrates how both can be combined to enhance information visualization for architectural design and construction.
