面向主题的多Agent搜索系统

【作者】 孙荣霜；

【导师】 沈洁；

【作者基本信息】 扬州大学 ， 计算机应用技术， 2008， 硕士

【摘要】 随着网络容量的不断增长,信息更新的不断加快,人们想快速地检索到符合自己要求的信息变得越来越困难。信息冗余、主题参杂等问题都影响了人们检索信息的效率。在解决这些问题的过程中,利用面向主题的技术提高信息检索的效率,是人们一直在研究的热点。本论文从面向主题信息检索的角度出发,研究了在信息检索过程中利用本体、知识库以及多Agent等技术来提高人们检索信息的效率。任何信息检索的过程都包含了用户根据自己的主题对信息进行分类的思想,如何满足用户的这种主题性检索才是信息检索亟需解决的问题。除了我们通常利用各种技术、算法来提高搜索引擎的查准率和查全率以外,我们还可以借助许多其他的方法来实现这个目的。本论文就介绍了一种结合了本体、知识库及多Agent技术的面向主题的检索方法。本论文的主要工作包括三个方面的内容。(1)找出符合用户检索意图的主题。在面向主题检索的过程中,弄清楚什么是主题是首要的问题。本论文中主题的涵义是指从用户角度出发的、对信息具有语义上的鉴别能力的概念。正是出于对语义的理解和利用,我们才考虑借助本体及知识库来解决对主题的描述及划分工作。搜索引擎不但要明白用户的主题,更要能通过一定的学习以达到对主题的“自我识别”。(2)把用户的主题有效地收集起来形成知识库。知识库在解决我们的面向主题的信息检索的问题起到了很大的作用。我们依据知识库对一个主题的“定性”来判断如何对用户提供的主题进行适当的“分解”,由多Agent系统来执行分解后的信息检索过程以及检索后的结果整理过程。(3)协作式搜索,提高检索的效率。为了更好的利用现有的搜索引擎的搜索能力以及它们对网页的分析能力,我们利用多Agent系统来进行同步检索,对同一个任务进行不同的检索,再去除其中的冗余信息,将这样的结果提供给用户,给用户带来便利。本论文中最终是为了实现在手机领域的具有主题性的信息检索。为了实现这样一个过程,我们设计了一个三层结构的搜索模型,分别用来处理用户检索词、主题分析以及协作式搜索。主题分析的工作主要依靠本体知识库来完成。为了构建一个适合本系统的、简单的本体知识库,我们借助网页分类相关技术来确定我们的关键词列表,并以这些关键词为初始原语组织了一个简单的手机本体。多Agent在本论文中的应用只是传统的多Agent系统的一个很小的部分,因为在本论文中我们只是借助利用多Agent系统来实现对一个检索任务的多引擎搜索和结果的无冗余汇总。更多还原

【Abstract】 As the consistently growing capacity of network and the speeding-up update of information, it’s more and more difficult for people to find out the wanted information quickly. Problems such as redundant information, mixed-up topic and so on affect the searching efficiency. During the process of dealing with the problems, a hot topic is take advantage of the topic oriented technology to improve information searching efficiency. This thesis from the aspect of topic oriented information searching area, studies how to take advantage of self, knowledge base and multi-Agent to improve the efficiency. Whatever information searching, it contains the process of user classifies the information based on his/her topic, thus how to meet the requirement of user’s topic searching is the critical problem of information searching. Besides the normal technology and algorithm which improves the efficiency of search engine’s accuracy and coverage, we can use other methods to approve this objective. This thesis introduces a method combined with self, knowledge base and multi-Agent to search information based on topic.Find out the topic which meets the user’s searching purpose. During the process of topic oriented information searching, find out what’s the proper topic is the critical problem. The topic mentioned in this thesis is the concept which from user’s aspect and with the semantic identification capacity of information. It’s just from the semantic understanding, that we consider take advantage of self and knowledge base to describe and classify topic. Search engine should not only understand user’s topic, but also automatically identify the topic through self-study.Collect user’s topics effectively and form knowledge base. Knowledge plays an important role when solving problems of topic oriented information searching. We justify a problem according to knowledge, and then separate the problem, and then multi- Agent system will execute and searching the organized result.Cooperative search and improve the efficiency of searching. To take better advantage of current search engine’s searching capability and the web analysis capability; we use multi-Agent system to search simultaneously. A task is searched from different aspect, filtered out the redundant information, and the result is convenient for end users.The purpose of this thesis is to implement topic oriented information searching on mobile area. To achieve the goal we design a three-tier searching model for user searching word, topic analysis and cooperative search. Topic analysis depends on self knowledge base. To construct a proper and simple knowledge base suitable for this system, we us web classification related technology to set up a list of key words, and use these key words to form a simple mobile self. Multi-Agent used in this thesis is a small part of traditional one, because we just use it to implement a multi-Agent search engine for searching tasks and a summary without redundant information.更多还原