Parametric Uncertain Database Management Systems

现实生活中的可用信息往往不完整、不精确、不确定和/或可能不一致。为了对真实的世界建模，我们需要能够表示和操作这些不完美的数据。该提案侧重于不确定性，即，可靠性有限的数据。直觉上，如果一个数据的真实性没有被确定，那么它就是不确定的。不确定性可以简单地建模为单位区间[0，1]中的真实的值或更复杂的对象，即，值或符号的集合。不确定性在人工智能和数据库研究中得到了广泛的研究，并取得了显著的成果。在1991年和2005年的数据库自我评估报告中，不确定性被确定为数据库研究的一个重要挑战：“进一步的研究[在不确定性]是必不可少的，因为我们必须学会不仅要科普有限的可靠性数据，但也这样做的效率，与大量的数据。“我们的建议是出于这种认识，并强调需要声明式编程和强大的工具和技术，以支持许多现有的和新兴的应用程序，处理不确定的数据。我们的目标是通过使用、采用和扩展现有的工具、技术和结果，为开发具有不确定性的所需数据库管理系统（DBMS）奠定理论和实践基础。最近，我们目睹了越来越多的研究活动，以建立这样的系统，例如，Orion（普渡大学）和Trio（斯坦福大学），它们将不确定数据视为一等公民。然而，在我们的建议中强调的是，这样的系统必须意识到数据中建模的不确定性的性质和种类，通过用户精心选择的所谓参数捕获，包括不确定性域和不确定性组合函数。这些参数将被认为是新的查询处理和优化技术，将在本提案中开发，以正确和有效地执行所需的操作。我们计划通过扩展PostgresSQL（一个面向对象的关系型开源数据库系统）来构建我们的参数化DBMS。

Available information in real-life is often incomplete, imprecise, uncertain, and/or possibly inconsistent. To model the real world, we need the ability to represent and manipulate such data with imperfection. This proposal focuses on uncertainty, i.e., data of limited reliability. Intuitively, a piece of data is uncertain if its truth is not established definitely. Uncertainty could modeled simply be a real value in the unit interval [0,1] or as more complex objects, i.e., a collection of values or symbols. Uncertainty has been studied extensively in artificial intelligence and database research, resulting in significant achievements. In the database self assessment reports in 1991 and 2005, uncertainty has been identified as an important challenge in database research: "Further research [in uncertainty] is essential as we must learn not only to cope with data of limited reliability, but also do so efficiency, with massive amounts of data." Our proposal is motivated by this recognition, and stress the need for declarative programming and powerful tools and techniques to support many existing and emerging applications that deal with uncertain data. Our goal is to contribute to theoretical and practical foundations towards the development of a desired database management system (DBMS) with uncertainty, by using, adopting, and extending existing tools, techniques, and results. Recently we have witnessed increased research activities in for building such systems, e.g., Orion (Purdue) and Trio (Stanford), which treat uncertain data as a first-class citizen. What is stressed in our proposal, however, is that such a system has to be made aware of the nature and the kind of uncertainty modeled in the data, captured through the so-called parameters that are carefully selected by a user and includes uncertainty domain and uncertainty combination functions. These parameters will be considered by new query processing and optimization techniques, to be developed in this proposal, in order to carry out the required operations correctly and efficiently. We plan to build our parametric DBMS by extending PostgresSQL -- an object-oriented relational open-source database system.