【作者】 王磊；

【导师】 黄广君；

【作者基本信息】 河南科技大学 ， 计算机应用技术， 2012， 硕士

【摘要】 随着互联网的发展及社会信息化的发展，信息呈爆炸式增长，人们对信息获取的要求越来越高，不断推动信息检索技术的发展。当前主流商用搜索引擎仍然采用基于关键词查找的方式。这种查找方式在给用户提供一种简便查询平台的同时，也给人们返回了大量与查询意愿无关的垃圾结果。因此，有学者提出使用查询扩展的方法来解决这个问题。查询扩展是在原查询词的基础上加入相似或相关的词以克服自然语言的“二义性”问题，改进查询意愿的描述。传统的查询扩展技术虽然在技术上有不断的改进，但是仍然以单个查询词为中心进行扩展，忽略了查询概念语义之间的关联扩展，因而没有充分表达和扩展出用户查询意图。近年来，语义概念查询扩展成为新的热点，就是在语义词典/领域本体的基础上构建概念语义空间，从概念语义空间中提取查询语义及其语义关联，实现语义概念扩展。这种扩展在一定意义上实现了语义层次的扩展，但过于依赖完备的语义体系，导致许多与用户查询意愿不相关的词的加入，从而容易出现查询漂移的问题。针对现有查询扩展存在扩展词质量不高的问题，本课题在前人研究成果的基础上，利用语义词典和文档集两种扩展源，在语义空间的基础上引入统计模型对查询词进行扩展。本文的研究工作主要包括以下几个方面：1.语义词典、领域本体等知识体系已经成为智能化信息检索不可或缺的支撑工具。在传统构建语义森林的基础上，提出了向上溯源查找最近公共祖先结点的方法，构建出覆盖面全、冗余度低，结构合理的概念语义空间；2.在对查询扩展进行范围控制的过程中，本课题提出了动态观察窗口加权模型，用于强化共现词之间的关联度。在大规模文档集中使用动态观察窗口加权模型对初始查询扩展词进行训练，动态设定显著性阈值，筛选出最终的查询扩展词；3.最后，本课题利用文本检索会议TREC（Text REtrieval Conference）提供的测试数据集设计并实现了实验系统，将实验结果转化为相应的算法测评指标MRR（平均倒数排名），通过对比可得，本课题提出的结合概念语义空间与动态观察窗口加权模型的扩展算法比传统伪相关反馈法的扩展效果有了较大的提高，从而提高了信息检索质量。本课题实验数据使用TREC（2005）会议的FR （Federal Register）部分，总计395M。这些数据中包括待检索文档56110篇，原始的查询文本50条与50条原始查询文本对应的目标文档号。所有实验数据均为来自TREC会议的标准数据集，从而保证了实验的客观性。更多还原

【Abstract】 In the development of Internet and information technology throughout thenational society, vast information emerges to us. The requirement of getting more andmore accurate information accelerates development of information retrievaltechnology. Current popular Search Engines still run with the method of searching keywords which provides us a simple search platform while getting a lot of uselessinformation that irrelevant to customers’ willing. Therefore, many academicians areusing a method called Query Expansion to resolve the problem of retrieval information.Query Expansion adds relevant words or phrases into original query in order toovercome ambiguous problem in natural language and describe customers’ willingmore clearly. Although traditional query expansion which only expands symbol andcharacter string taking query words as the center has been improved frequently, itignores semantic relationship in queries and fails to express customers’ willingcorrectly.In recent years, semantic concept query expansion has become to a new focus. Itmeans to construct concept semantic space on the basis of semantic dictionary orontology, and then extract meaning and relationship of query words. The way of queryexpansion achieves expansion in semantic aspect to a degree but results in querydrifting frequently because it depends on complete semantic construction and manyirrelevant expansion words are added into query collection.In view of current problems in query expansion, this task designs to queryexpansion in meaning making use of semantic dictionary and documents, introducingthe advantage of semantic analysis and statistical model on the basis of achievementsof predecessors. Therefore, the keys to this study are as follow: 1. Semantic dictionary and ontology have become necessary tools in intelligentinformation retrieval. On the basis of traditional method, we take advantage of findingcommon ancestor in order to structure overall, compact, efficient Concept SemanticSpace.2. In the process of controlling number of expansion terms, this task Importsdynamic observation windows weighting model into the algorithm, so that enhance thefunction of co-occurrence information in relevancy between every two words to getfinal expansion words by setting threshold about relational degree dynamically.3. Finally, authors develop an experiment system using testing data set providedby Text REtrieval Conference and evaluate the results of experiment by MRR. Theexperimental results show that the expansion algorithm of this task is much moreeffective than traditional pseudo-relevance feedback algorithm.All the experimental data are from Federal Register part of TREC which is395MB totally. They include56110chapters,50query topics and50target numbers ofchapters. These data are very standard from TREC, so that they ensure experimentalobjectivity.更多还原