空间文本数据的量质融合与推送

发布时间：2019-01-25 10:39

【摘要】：随着智能手机等移动设备的普及以及移动互联网络的快速发展,基于地理位置的服务得到了广泛的关注与应用。这些服务产生了大量空间文本数据,既包含空间地理位置,同时又具有相关的文本信息。能否更好地利用这些数据方便我们的日常生活,有两个方面的因素尤其值得考虑:第一,数据质量,高质量的数据能够提升用户体验;第二,用户获得信息的途径,用户需要高效便捷的获得感兴趣数据的方式。本文主要针对这两方面展开研究:一方面对空间文本数据进行质量优化,包括提高数据准确性,降低数据冗余;另一方面研究如何应对不同用户的需求,将空间文本数据信息高效地推送给用户。本文的研究内容与贡献具体包括:(1)基于众包的空间文本数据提纯:为保证数据正确性,用众包的方法帮助筛选数据中的正确关键词。本文提出了基于众包的任务执行框架。一方面,将空间文本数据作为任务发布在众包平台上并收集工人答案。为能够根据工人答案筛选正确关键词,结合工人质量等因素构造了一个有效的推断模型。另一方面,为进一步提高推断准确率,研究了能够最大化推断准确率提升的众包任务分配算法。(2)空间文本数据的top-k融合:为降低数据冗余,对数据进行top-k融合。传统方法不考虑空间文本数据的结合,为解决该问题,本文提出了一种基于空间文本签名的过滤验证算法框架,通过签名过滤不相关的数据对。接着发现签名的访问顺序对算法效率有着巨大的影响,于是提出了一种最佳优先算法,以签名上界从大到小的顺序逐一访问签名。这种方式能够快速地找到top-k结果。此外论文优化了空间文本签名,提出了扩展的空间文本签名进一步加强了算法的过滤能力。(3)空间文本数据的推送:为满足用户不同的推送需求,本文提出并研究了参数化的空间文本信息推送问题。为解决该问题,论文提出了一种空间前缀以及对应的过滤验证算法。又结合经典的空间层次索引结构提出了区域前缀,区域前缀能有效减少前缀大小并支持区域过滤。在此基础上结合多关键词过滤技术提出了空间文本前缀,进一步提升了算法的过滤能力。
[Abstract]:With the popularity of mobile devices such as smart phones and the rapid development of mobile Internet, geographically based services have been widely concerned and applied. These services produce a large amount of spatial text data, which not only contain spatial location, but also have relevant text information. Whether we can make better use of these data to facilitate our daily life, there are two factors worth considering in particular: first, data quality, high quality data can improve the user experience; Second, the way users get information, users need efficient and convenient access to data of interest. This paper focuses on these two aspects: on the one hand, optimize the quality of spatial text data, including improving the accuracy of data, reducing data redundancy; On the other hand, how to deal with the needs of different users and efficiently push spatial text data to users. The contents and contributions of this paper are as follows: (1) Spatial text data purification based on crowdsourcing: in order to ensure the correctness of the data, crowdsourcing method is used to help filter the correct keywords in the data. This paper presents a task execution framework based on crowdsourcing. On the one hand, the spatial text data is published on the crowdsourcing platform as a task and the workers' answers are collected. In order to select the correct keywords according to the workers' answers and combine the factors such as workers' quality, an effective inference model is constructed. On the other hand, in order to further improve the accuracy of inference, the crowdsourcing task assignment algorithm which can maximize the accuracy of inference is studied. (2) top-k fusion of spatial text data: in order to reduce data redundancy, the data is fused by top-k. The traditional method does not consider the combination of spatial text data. In order to solve this problem, this paper proposes a filtering algorithm framework based on spatial text signature, which filters irrelevant data pairs by signature. Then it is found that the access order of the signature has a great influence on the efficiency of the algorithm, so an optimal priority algorithm is proposed to access the signature one by one with the upper bound from large to small. This way you can quickly find top-k results. In addition, the paper optimizes the spatial text signature, and proposes an extended spatial text signature to further enhance the filtering ability of the algorithm. (3) push the spatial text data: to meet the user's different push requirements, This paper presents and studies the parameterized spatial text information push problem. In order to solve this problem, a spatial prefix and a corresponding filtering algorithm are proposed. Based on the classical spatial hierarchical index structure, a region prefix is proposed, which can effectively reduce the prefix size and support region filtering. On this basis, the spatial text prefixes are proposed in combination with multi-keyword filtering technology, which further improves the filtering ability of the algorithm.
【学位授予单位】：清华大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP391.1

【相似文献】