【作者】 朱松；

【导师】 徐建民；

【作者基本信息】 河北大学 ， 计算机应用技术， 2008， 硕士

【摘要】 传统的信息检索方法采用基于关键词匹配的方式进行资源检索,这种检索方法不能实现语义概念上的匹配,也不能准确得到用户所需的信息。当利用给定文档集合中所包含的术语间的关系时,能够提高信息检索系统的性能,而术语相似度和术语相关度就是信息检索领域中挖掘术语间关系的主要方法。这两种方法只是单一的挖掘术语间的相似关系或相关关系,虽然在检索性能上可以获得一定程度的提高,但是不能满足用户日益增长的准确、全面定位信息的需求。如果融合这两种方法以量化术语间的关系,是否会得到更好的检索效果成为一个值得研究的问题。为此,本文探讨了如何利用术语间的关系来提高信息检索系统的性能,分析了术语相似度和术语相关度融合的可行性。利用基于《知网》的术语相似度计算方法和术语相关度中的非对称性共现分析法,设计出一种衡量术语间关系的融合算法,并将这种量化的术语关系应用于两种检索模型中。实验结果表明所提方法比单一使用术语相似度或术语相关度的方法具有更好的检索效果,能更有效地解决术语间语义概念的匹配问题。更多还原

【Abstract】 The traditional information retrieval (IR) method based on the keyword matching can’t realize semantic retrieval and can’t find out the information of user needed exactly.The performance of IR system can be improved by using term relationships. In the field of IR, term similarity and term correlativity were used to mine term relationships respectively. The two methods are only used to mine single term relationships respectively, the retrieval performance has improved for a degree but doesn’t completely meet the needs of users. If the two methods were utilized together to quantify the relationship between terms, whether it will be a better search results as a problem worthy of study.So this paper discuses how to improve the retrieval efficiency by the means of term relationships and analysis the feasibility of merge term similarity and term correlativity. Using term similarity method based on HowNet and term correlativity method based on asymmetry co-occurrence analysis, this paper designs a merge method to measure the term relationship and applies this term relationship to two retrieval models. Experimental results which using this new method show that the retrieval efficiency can be improved more obviously, the matching problem of semantic concept between terms is resolved more effectively.更多还原