基于模糊逻辑的关系数据库信息检索方法研究

发布时间：2018-05-22 11:50

本文选题：关系数据库 + 对象级别检索　；参考：《大连海事大学》2013年硕士论文

【摘要】：关系数据库关键词检索技术使得用户不需要了解任何SQL语言和底层数据库模式,就可以方便地检索到数据库中存储的信息,就像使用搜索引擎一样简单、便捷,因此关键词检索技术成为数据库查询领域的研究重点。数据库的规范化设计使得检索结果以元组级别的形式展现,这将导致检索结果不完整,语义难以理解,因此提出对象级别的信息检索技术。对象级别的信息检索技术可以直接地表达检索结果的语义,返回的结果也更加完整。但当用户输入的是具有模糊性的检索关键词时,其检索效果并不好。而模糊数学知识的引入则可以很好地解决数值型关键词的模糊检索问题。在对数值型的关键词进行操作时,首先对数据库进行分析,然后有针对性地提出隶属函数,利用隶属函数和模糊化算子对数值型关键词进行操作。在相关性排序时,通过利用模糊逻辑中的模糊推理方法来计算对象结点的权重。在对关键词检索结果进行排序时,不仅要考虑信息检索技术本身的特点,还要考虑数据库特性,主要有元组重要性,属性重要性以及属性上关键词的IR分数。数据库中各个元组／属性被检索到的次数是不同的,这就表明不同的元组／属性对用户的重要性是不一样的。在计算属性上关键词的IR分数时,采用传统的TF/IDF可以很好地得到结果,但这只限于对非数值属性的关键词,对于数值属性的关键词则很不实用。因此需要利用隶属函数对数值型关键词进行操作,进而实现对数值属性关键词的模糊检索,并有效地进行相关性排序。本文采用上述方法设计并实现了一个基于模糊逻辑的关系数据库对象级别信息检索原型系统。利用DBLP数据集对该原型系统进行了实验验证,并采用P@K和MAP两个评价指标对实验进行评价。最终的实验结果表明本文的方法能有效改善检索结果的排序效果。
[Abstract]:Relational database keyword retrieval technology makes it easy for users to retrieve the information stored in the database without knowing any SQL language and underlying database schema, which is as simple and convenient as using search engine. Therefore, keyword retrieval technology has become the research focus in database query field. The standardized design of the database makes the retrieval result display in the form of tuple level, which will lead to incomplete retrieval results and difficult semantic understanding. Therefore, an object-level information retrieval technique is proposed. Object level information retrieval technology can directly express the semantics of retrieval results, and the returned results are more complete. However, when users input fuzzy search keywords, the retrieval effect is not good. The introduction of fuzzy mathematics knowledge can solve the problem of fuzzy retrieval of numerical keywords. When the key words of numerical type are operated, the database is analyzed first, then the membership function is put forward, and the membership function and fuzzy operator are used to operate the key words of numerical type. The weights of object nodes are calculated by using fuzzy reasoning method in fuzzy logic. When sorting the results of keyword retrieval, we should consider not only the characteristics of information retrieval technology, but also the characteristics of database, including tuple importance, attribute importance and IR score of attribute keywords. The number of times each tuple / attribute is retrieved in the database is different, which indicates that the importance of different tuple / attribute to the user is different. When calculating the IR score of the keywords on the attributes, the traditional TF/IDF can get the results well, but this is limited to the keywords of the non-numeric attributes, but it is not practical for the keywords of the numerical attributes. Therefore, it is necessary to use the membership function to operate the numeric keywords, and then to realize the fuzzy retrieval of the numeric attribute keywords, and to sort the correlation effectively. This paper designs and implements a prototype system of object level information retrieval in relational database based on fuzzy logic. The prototype system is verified by using DBLP data set, and the experiment is evaluated by two evaluation indexes: P@ K and MAP. The final experimental results show that the proposed method can effectively improve the ranking effect of retrieval results.
【学位授予单位】：大连海事大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP311.132.3;TP391.3

【相似文献】