【作者】 杨艳； 何天宇；

【Author】 Yang Yan and He Tianyu (School of Computer Science and Technology,Hei Longjiang University,Harbin 150080) (The Key Laboratory of Database and Parallel Computing,Hei Longjiang University,Harbin 150080)

【机构】 黑龙江大学计算机科学技术学院； 黑龙江大学数据库与并行计算重点实验室；

【摘要】 关系数据库关键词查询已经成为当前的研究热点.对于用户输入的关键词序列,现有系统以单个查询词为查询基本单位,没有考虑查询词之间的语义关系,而查询词之间的语义关系对于查询处理具有重要的意义.用户输入的关键词序列经常包含很多短语,一个短语代表了用户的一个完整的语义需求.提出基于单词级倒排索引的短语识别算法,可有效识别出用户查询中包含的短语.为了使查询结果更准确,又提出了短语反馈机制.理论分析和实验结果表明,基于短语的查询可以显著提高查询效率和准确性.更多还原

【Abstract】 Keyword search over relational databases has become a critical issue.For the query keywords the user input,existing systems take single keyword as the basic unit,none of them consider the relations between the keywords.The relations between keywords is very important for query processing.The keywords sequence in a user query often contains many phrases and each phrase represents a complete semantic.In this paper,we propose a phrase recognition algorithm based on the word-level inverted index,which can effectively recognize the phrases contained in the user query.To improve the query precision,a phrase feedback mechanism is proposed also.Theoretical analyses and experiments show that the effciency and the query precision arc both greatly improved by our phrase recognition and feedback methods.更多还原