联系我们

论文查询结果	返回搜索

论文编号：	1007
作者编号：	2120072355
上传时间：	2009/5/26 8:13:21
中文题目：	元搜索引擎提问融合方法研究
英文题目：	Research of Query Fusion Metho
指导老师：	李培
中文关键字：	元搜索引擎;用户查询日志;相关反
英文关键字：	Meta Search Engine;User Query
中文摘要：	本文将信息融合理论与元搜索引擎的研究相结合，在相关理论的基础上，对提问融合方法做一些探讨和研究。本文介绍了信息融合基本含义、原理、层次结构以及技术、方法，总结了查询扩展的三类方法，并着重介绍和分析了相关反馈的定义、类型、技术与算法，详述了语言模型的建模过程与方法。 本文的研究内容包括以下几个方面： 首先，提出了对用户的查询行为进行信息挖掘，给出了记录用户信息需求的三种方式，并分析用户的查询行为，对用户的信息需求进行分类。本文试图利用用户查询日志来探寻用户的信息搜索行为，提出了基于用户查询日志的两种提问融合策略：基于用户查询模糊聚类的提问融合策略和基于用户点击文档用词的提问融合策略，从而利用日志进行伪相关反馈。 第二，基于用户查询模糊聚类的提问融合策略是利用系统聚类法对用户查询日志中的用户查询进行聚类分析，用户查询是聚类的样本，每条用户查询所对应的用户点击的文献是聚类样本的特征。 第三，基于用户点击文档用词的提问融合策略将用户查询中使用的词或短语与文档中出现的相应词或短语以条件概率的形式连接，利用贝叶斯公式挑选出文档中与该查询关联最紧密的词加入原查询，以达到扩展优化的目的。 最后，本文总结了Top k反馈程序和自动选取前 篇文献的两种策略，提出了使用语言模型为工具计算相关度系数，据此改进了多查询与伪相关反馈的融合，提出了基于Top k~反馈的提问融合算法。 本文的创新之处是提出了基于用户点击文档用词的提问融合策略和基于Top k~反馈的提问融合算法。
英文摘要：	This article will be the combination of information fusion theory and the study of meta-search engine, based on the relevant theory, to explore and do some research on the query fusion. This paper introduces the basic theory of information fusion including the signification, principle, hierarchy, as well as technology and methods, summarizes the three types of query expansion methods, with an emphasis on relevance feedback of the definition, type, technology and algorithms, details in the language model to build mode of the process and methods. The contents we study in the paper are several facets below. First of all, we propose that utilize the user's query logs for information mining, memorize the record of the user's information needs in three ways, analyze the user's query behavior, and classify the information needs of the user. This paper attempts to make use of the user query log to explore the user's information search behavior, proposes two types of query fusion strategies to use the user query log, which are the query fusion strategy of user query fuzzy Clustering and the query fusion strategy of the words of user-clicked-documents strategy, in order to use the user query log to do pseudo-relevance feedback. Secondly, the query fusion strategy of user query fuzzy clustering is to use hierarchical clustering method to cluster user queries of the user query log. The user queries are the clustering samples, and the user-clicked-documents corresponding to each user query is the characters of clustering samples. Thirdly, the query fusion strategy of the words of user-clicked-documents strategy connects the word or phrase in the user query with the word or phrase appearing in the corresponding documents with the conditional probability form, using Bayesian formula for selecting and adding the words closest to the original query in the documents, in order to achieve the purpose of optimizing the expansion. Finally, this article summarizes the Top k feedback process and two strategies of automatically selecting the top k documents, propose the use of language model for calculating the correlation coefficient, by which to improve the algorithm of multiple query and pseudo relevance feedback, and propose the query fusion algorithm of Top k~ relevance feedback. The innovations in the paper include the query fusion strategy of the words of user-clicked-documents strategy and the query fusion algorithm of Top k~ relevance feedback.
查看全文：	预览  下载(下载需要进行登录)