基于概念层次的网络挖掘技术

发布时间：2019-05-10 14:11

【摘要】： 概念层次(Concept Hierarchy)是指将大量的概念用层次化的方法组织起来,使得子概念所包含的意义比它的父亲更加特殊,可以被其父概念所概括。层次化的概念模型与普通的平面概念模型不同,在这样一个模型中,概念之间的距离是不均匀的。通过这种距离,我们可以衡量出各个概念之间的相似程度。这样一个模型可以构造出一种更接近于现实的标准,使得基于这个模型的分类、聚类、匹配等工作变得更加合理。概念层次的建立实现了一个层次化的标准。通过将其它的元素(如单词、查询请求、文档)映射到概念上,就可以在这些元素之间建立起联系,并且这些联系是包含语义信息的。层次化地组织概念是网络挖掘中的一个很常见的问题,许多的应用场景都是基于这样一个思想来完成。本课题的工作大致分为三个部分。我们首先从底层着手,致力从网络信息终挖掘出一个概念描述准确、层次结构划分合理的概念层次模型。我们以目前正在流行的社会化标注为基础,设计了一套从社会化标注数据中抽取概念并建立层次关系的方法。我们还以概念层次为基础,发掘其在网络挖掘中的应用。针对搜索引擎广告服务中涉及的关键词推荐问题,我们提出如何利用概念层次来提高推荐的覆盖面和准确度的方法。最后,考虑到概念层次本身的规模庞大,我们也希望应用一些可视化的技术,能将概念层次的全貌直观展现给用户。我们的方法体现在展现概念层次内在联系与本身结构上,都取得了不错的效果。
[Abstract]:Concept hierarchy (Concept Hierarchy) refers to the organization of a large number of concepts in a hierarchical way, so that the meaning of the child concept is more special than its father, and can be summarized by its father concept. The hierarchical conceptual model is different from the ordinary plane conceptual model. In such a model, the distance between concepts is not uniform. From this distance, we can measure the similarity between concepts. Such a model can construct a standard that is closer to reality, which makes the classification, clustering, matching and other work based on this model more reasonable. The establishment of conceptual hierarchy implements a hierarchical standard. By mapping other elements (such as words, query requests, documents) to concepts, you can establish connections between these elements, and these connections contain semantic information. The concept of hierarchical organization is a very common problem in network mining, and many application scenarios are based on such an idea. The work of this topic is roughly divided into three parts. Starting from the bottom layer, we try to mine a conceptual hierarchy model with accurate concept description and reasonable hierarchical division from the end of the network information. Based on the popular socialized tagging, we design a set of methods to extract concepts from socialized tagging data and establish hierarchical relationships. Based on the concept level, we also explore its application in network mining. In order to solve the problem of keyword recommendation in search engine advertising service, we propose a method to improve the coverage and accuracy of recommendation by using the conceptual level. Finally, considering the large scale of the concept level itself, we also hope to apply some visual technology to show the whole picture of the concept level to the user intuitively. Our method is embodied in showing the internal relationship at the conceptual level and its own structure, and has achieved good results.
【学位授予单位】：上海交通大学
【学位级别】：硕士
【学位授予年份】：2008
【分类号】：TP311.13

【共引文献】