Generating networked knowledge based on statistical modeling and data mining techniques and its applications

我们的目标是利用统计建模和数据挖掘技术从分散的信息片段中构建网络知识，并将这些知识应用于解决人类智力活动中的问题。为了实现这些目标，我们主要开展了以下研究工作：(1)基于主题模型从文本数据中提取关系结构的研究：我们从一些结构化的文本数据中对潜在主题进行统计估计，并将潜在主题用于一些应用，如实体之间或数据之间的关系结构提取。(2)从网络数据中推断节点簇及其应用的研究：我们从网络中对潜在社区进行统计估计，并将估计的社区应用于实际问题，如链接预测。(3)网络数据模式发现研究：我们开发了从复杂的结构化数据中发现特征模式的算法，例如其中节点或边与一组数字属性相关联的单个网络。

We aim to construct networked knowledge from scattered pieces of information using techniques of statistical modeling and data mining, and apply that knowledge to solve problems in human intellectual activities. To achieve these objectives, we mainly carried out the following research tasks:(1) Research on extracting relational structures from text data using topic models : We statistically estimated latent topics from some sorts of structured text data, and used the latent topics in some applications, such as extracting relational structures between entities or between data.(2) Research on inferring node clusters from network data and its applications : We statistically estimated latent communities from a network, and applied the estimated communities to real-world problems, such as link prediction.(3) Research on discovering patterns from network data : We developed algorithms to discover characteristic patterns from complex structured data, such as a single network in which nodes or edges are associated with a set of numerical attributes.