Web数据库近似查询结果自动排序方法

Automated Ranking of Approximate Query Results of Web Database

针对Web数据库近似查询产生的多查询结果问题,提出了一种近似查询结果自动排序方法,该方法利用KL距离(Kullback-Leibler distance),PIR(probabilistic information retrieval)模型和查询历史(query history)来构建元组排序打分函数;打分函数根据结果元组中被查询指定的属性值对初始查询的满足度和未被查询指定的属性值与用户偏好的相关度来评估元组的排序分值.实验证明,提出的排序方法能够较好地满足用户需求和偏好,并具有较高执行效率.

To solve the problem of over many results obtained from a Web database through an approximate query,an automated ranking method was first proposed for those results,taking advantages of the KL distance,PIR model and query history to formulate a tuple scoring function. According to the satisfiabilities of specified attribute values in result tuple to the initial query and the relevance of unspecified attribute values to users＇ preference,the scoring function evaluates the scores of ranked tuples. Experimental results demonstrated that the automated ranking method proposed can meet the users＇ requirements and their preference effectively with high efficiency.