面向智慧民生领域的增量交互式数据集成方法

发布时间：2018-09-08 07:51

【摘要】：智慧民生作为智慧城市的重点领域,包含众多应用系统,积累了大量层次结构数据.为了形成城市范围完整数据集,需要集成并统一异构的数据模式,向用户提供统一的数据视图.针对智慧民生领域的领域知识宽泛、缺乏中文语义匹配支持、模式数量众多、元素标签缺失但实例数据丰富等几方面特点,提出了一种增量交互式模式集成方法.该方法采用增量迭代的方式逐步完成多模式集成任务,大幅降低集成计算量;在模式匹配阶段,综合利用模式信息和实例数据构造了多种适用于中文且能力互补的匹配器,并通过相似度熵来度量机器的决策置信度,适度引入人工干预;在中介模式生成阶段,处理模式间可能出现的各种冲突,最终输出全局统一的中介模式.利用从互联网爬取的多源二手房数据设计并完成实验,实验结果表明:此方法在人工干预程度足够小的前提下,具有较好的模式匹配准确性.
[Abstract]:As a key area of intelligent city, people's livelihood includes many application systems and accumulates a lot of hierarchical data. In order to form a complete urban data set, it is necessary to integrate and unify heterogeneous data patterns and provide users with a unified data view. An incremental interactive pattern integration method is proposed in view of the broad domain knowledge in the field of wisdom and livelihood, the lack of Chinese semantic matching support, the large number of patterns, the absence of element labels and the abundance of instance data. The method uses incremental iteration to accomplish the task of multi-pattern integration step by step and greatly reduces the amount of integration computation. In the stage of pattern matching, a variety of matching devices suitable for Chinese and complementary ability are constructed by synthesizing pattern information and instance data. The similarity entropy is used to measure the decision confidence of the machine, and the artificial intervention is introduced appropriately. In the stage of intermediary pattern generation, all kinds of conflicts that may occur between the patterns are dealt with, and the global uniform intermediary mode is finally output. The experimental results show that this method has good pattern matching accuracy on the premise that the degree of artificial intervention is small enough.
【作者单位】：高可信软件技术教育部重点实验室(北京大学);北京大学信息科学技术学院;软件工程国家工程研究中心(北京大学);
【基金】：国家“八六三”高技术研究发展计划基金项目(2013AA01A605) 质检公益性行业科研专项(201510209)~~
【分类号】：TP399-C2

【相似文献】