Graph databases are becoming widely successful as data models that allow to effectively represent and process complex relationships among various types of data. As with any other type of data repository, graph databases may suffer from errors and discrepancies with respect to the real-world data they intend to represent. In this work we explore the notion of probabilistic unclean graph databases, previously proposed for relational databases, in order to capture the idea that the observed (unclean) graph database is actually the noisy version of a clean one that correctly models the world but that we know partially. As the factors that may be involved in the observation can be many, e.g, all different types of clerical errors or unintended transformations of the data, we assume a probabilistic model that describes the distribution over all possible ways in which the clean (uncertain) database could have been polluted. Based on this model we define two computational problems: data cleaning and probabilistic query answering and study for both of them their corresponding complexity when considering that the transformation of the database can be caused by either removing (subset) or adding (superset) nodes and edges.
翻译:图表数据库作为数据模型正在变得广泛成功,能够有效地反映和处理各种类型数据之间的复杂关系。与任何其他类型的数据储存库一样,图表数据库在它们打算代表的真实世界数据方面可能存在错误和差异。在这项工作中,我们探索了先前为相关数据库提议的概率性不洁图形数据库的概念,以便捕捉以下想法:观测到的(非清洁)图形数据库实际上是清洁数据库的噪音版本,它能正确模拟世界,但我们部分地知道。由于观测中可能涉及的因素可能很多,例如所有不同类型的文书错误或数据意外变换,因此我们假定一种概率模型,用以描述清洁(不确定)数据库可能受到污染的所有可能方式的分布情况。基于这一模型,我们定义了两个计算问题:数据清理和概率性查询回答,并在考虑数据库的变换代(子集)或添加(超级设置)节点和边缘可以引起数据库的改变。