| 摘要: |
| 不一致数据无法正确反映现实世界,其上的查询结果内含错误或矛盾,而现有的很多不一致数据查询处理相关研究都存在信息丢失的问题. AQA(annotation based query answer)针对这一问题采用信任标签在属性级别上区分一致和不一致数据,避免了信息丢失.但AQA 假设记录在依赖左边属性上的分量可信,且只针对函数依赖一种约束,具有应用局限性.在综合约束(函数依赖、包含依赖和域约束)范围内、不确定属性任意的情况下扩展了AQA,重新审视了AQA 的数据模型及其上的查询代数,讨论了任意约束在查询结果上的蕴含约束计算问题.实验结果表明,扩展后的AQA非连接类查询的性能和普通的SQL基本相同,连接查询经优化后性能接近普通SQL查询,但AQA不丢失信息,与部分同类研究相比有很大优势. |
| 关键词: 不确定数据 数据质量 一致的查询回答 完整性约束 数据清洗 |
| DOI：10.3724/SP.J.1001.2012.04079 |
| 分类号: |
| 基金项目:上海海事大学校基金(20110042) |
|
| Query Answer over Inconsistent Database with Credible Annotations |
|
WU Ai-Hua1, TAN Zi-Jing2, WANG Wei2
|
|
1.College of Information Engineering, Shanghai Maritime University, Shanghai 201306, China;2.School of Computer Science, Fudan University, Shanghai 200433, China
|
| Abstract: |
| Inconsistent data is confusing and conflicting. Computing credible query answers over such data is significant. However, previous related works lose information. The approach of annotation based query answer (AQA) introduces confidence annotation to differ consistently and inconsistently in attribute value. Thus, a credible query answer can be computed and information loss can also be avoided. This is limited, however, in functional dependencies. This paper extends the approach to applications where multi constraints are involved, and no attribute is definitely credible. This paper redefines its representing model and query algebra, discusses the rules for calculating valid implied constraints of the above types on query result for any query algebra, proposes a cost based heuristic algorithm to repair, and annotates the initial database. The experiments show that time performance of extended AQA is almost similar to that of SQL for any query without join, and close to SQL for join queries after optimization, but it doesn’t loss information. |
| Key words: uncertain data data quality consistent query answer integrity constraints data cleaning |
