结合查询相关性的关键词查询排序方法  

Ranking approach combined with search relevance for keyword search

作　　者：杨书新[1] 徐慧琴[1] 谭伟[1] 

机构地区：[1]江西理工大学信息工程学院,江西赣州341000

出　　处：《计算机工程与设计》2013年第9期3136-3140,共5页Computer Engineering and Design

基　　金：江西省科技厅青年科学基金项目(20122BAB211035);江西省自然基金项目(20122BAB201045);江西省教育厅科技基金项目(GJJ12349;GJJ12345;GJJ12347)

摘　　要：针对关系数据库关键词查询系统中的结果排序问题,提出了一种新的排序方法。该方法结合了查询相关性和结构权重,将单个元组看作是一个虚拟文档,通过对元组引入信息检索(information retrieval,IR)式评分方式,采用标准化词频和标准化逆文档频率说明元组与查询条件之间的相关性程度,对整个结果采用结构权重来反应结果的语义强度。相比于以往只考虑结构权重的排序方法,该方法能更有效的将与查询高度相关的结果排在前面。实验结果表明,结合查询相关性的排序方法可以有效的对结果进行排序。To address the problem of result ranking in keyword search based on relational database, a new ranking approach which combines structure weigth with search relevance is proposed. A single tuple is considered as a virtual document. IR style scoring method which adds normalized term frequency and normalized inverse document frequency in this approach to indicate the correlation degree between a single tuple and given query condition. Structure weight for result tree is adopted to explain the se- mantic strength among tuples. Compared to the other well-known methods which only considered about structured weight, this method can obtain highly related results more effectively. The effectiveness of the proposed method is verified through a series of experiments.

关 键 词：关键词查询 结构权重 虚拟文档 相关性 结果排序 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]