网络环境下的领域知识挖掘

发布时间：2018-09-08 11:49

【摘要】： 当前海量异质、快速增长的网络资源带来了“数据过剩”和“知识贫乏”的矛盾,增大了人们及时获取有用知识的难度。本文以网络环境下的异质数据源为研究对象,探寻各种数据中知识发现的可行性,按照“模型提出—算法实现—数据验证”的思路,研究如何有效地利用和挖掘网络数据资源,获取潜在的、有价值的领域知识。 1.提出了网络环境下的领域知识挖掘模型。该模型是一个包括数据层、知识层和应用层的三层模型,指导从异构数据源中挖掘多维度知识以提供多种知识应用。基于该模型,论文以网络科技文献、博客日志和社会化标注为研究对象,进行了三种具体的领域知识挖掘实践研究。 2.提出了一种新的概率主题模型：Topic-Author模型。该模型对科技文献的文本信息和作者信息进行联合建模,深入对文献的分析。基于此模型构建了一个多维度文献知识挖掘框架,进行概念挖掘、专家发现、文献推荐,研究趋势分析、主题关系挖掘等领域知识的发现和应用。 3.提出了一个Blog知识挖掘框架,进行主题挖掘、观点分析和扩散研究。利用文本聚类和主题模型两种文本分析方法,挖掘Blog日志内容中的潜在概念,并对其进行观点分析。研究了社会化网络的扩散模型,总结了实现扩散最大化的方法,提出了一种改进的门槛扩散模型。 4.分析了基于社会化标注的集体智慧和Web环境下的知识组织分类法,进行了社会化标注的语义知识挖掘,提出了一种轻量级本体构建方法。该方法依据所提出的基于加权网络分割的社会化标签聚类算法,进行语义聚类和语义分层。研究结果表明,论文所提出的领域知识挖掘方法,能够发现大量有价值的、潜在的多维度知识,为用户提供多种知识应用服务,支持信息时代的知识获取与学习。
[Abstract]:At present, massive heterogeneous and rapidly growing network resources bring about the contradiction between "data surplus" and "knowledge poverty", which makes it more difficult for people to obtain useful knowledge in time. Taking heterogeneous data sources in the network environment as the research object, this paper explores the feasibility of knowledge discovery in various kinds of data, and studies how to effectively utilize and mine network data resources according to the idea of "model-algorithmic realization-data verification". Acquire potential, valuable domain knowledge. 1. A domain knowledge mining model in network environment is proposed. The model is a three-tiered model including data layer, knowledge layer and application layer, which can be used to mine multi-dimensional knowledge from heterogeneous data sources to provide a variety of knowledge applications. Based on this model, this paper takes the web science and technology literature, blog log and social tagging as the research objects, and carries out three kinds of practical research on domain knowledge mining. 2. In this paper, a new probabilistic topic model, the: Topic-Author model, is proposed. In this model, the text information and author information of scientific and technological documents are modeled jointly, and the literature is analyzed in depth. Based on this model, a framework of multi-dimensional literature knowledge mining is constructed, which includes concept mining, expert discovery, literature recommendation, research trend analysis, topic relationship mining, and so on. This paper presents a Blog knowledge mining framework for topic mining, viewpoint analysis and diffusion research. Text clustering and topic model are used to mine the potential concepts in Blog log content and analyze their viewpoints. In this paper, the diffusion model of social network is studied, the method of maximizing diffusion is summarized, and an improved threshold diffusion model is proposed. The collective wisdom based on social tagging and the classification of knowledge organization in Web environment are analyzed. The semantic knowledge mining of social tagging is carried out, and a lightweight ontology construction method is proposed. This method is based on the proposed social label clustering algorithm based on weighted network segmentation, and carries out semantic clustering and semantic stratification. The results show that the domain knowledge mining method proposed in this paper can discover a large number of valuable and potential multi-dimensional knowledge, provide users with a variety of knowledge application services, and support knowledge acquisition and learning in the information age.
【学位授予单位】：华东师范大学
【学位级别】：博士
【学位授予年份】：2010
【分类号】：G434

【相似文献】