III: Small: Collaborative Proposal: Towards Robust Uncertain Data Management

该项目的目标是开发一个系统框架，以便能够在各种应用程序领域中自然出现的数据不确定性的情况下进行“健壮”的查询处理。数据不确定性可能采取以下形式：丢失或不完整的数据、数据中的固有噪声、基于数据来源分配给数据的信任或声誉分数、或对使用自动化建模工具进行的预测的信心。输入的不确定性自然会导致对这类数据进行的任何查询或分析的结果都存在不确定性。为了能够对这种不确定的查询结果进行健壮和系统的推理，开发了有效的算法和实用工具来：(A)识别查询结果最敏感的输入不确定性，(B)决定如何使用诸如主题专家之类的稀缺资源来解决查询结果中的不确定性，以及(C)结合用户反馈来提高输入不确定性参数本身的稳健性。这些工具有可能使处理和分析不确定数据并从其中提取有用信息变得容易和直观，应用领域广泛，包括社交媒体分析、科学和生物数据管理、传感器数据管理、网络数据集成和信息提取。该项目为研究生和本科生提供研究机会，并与私人投资机构提供的几个高级研究生课程保持一致。该框架的原型实施、出版物和实验数据将通过项目网站：http://www.cs.umd.edu/~amol/RPrDB.进行传播

The goal of this project is to develop a systematic framework to enable "robust" query processing in presence of data uncertainties that arise naturally in a wide variety of application domains. Data uncertainties may take the form of missing or incomplete data, inherent noise in the data, trust or reputation scores assigned to data based on their sources of origin, or confidences in predictions made using automated modeling tools. The input uncertainties naturally lead to uncertainties in the results of any queries or analyses performed on such data. To enable robust and systematic reasoning over such uncertain query results, efficient algorithms and practical tools are developed to: (a) identify the input uncertainties to which query results are most sensitive, (b) decide how to use scarce resources like subject matter experts to resolve uncertainties in query results, and (c) incorporate user feedback to improve the robustness of the input uncertainty parameters themselves. The tools have the potential to make it easy and intuitive to process and analyze uncertain data and extract useful information from it in a wide range of real-world application domains including social media analysis, scientific and biological data management, sensor data management, web data integration, and information extraction. This project provides research opportunities for graduate and undergraduate students, and is aligned with several advanced graduate courses offered by the PIs. The prototype implementation of the framework, publications, and experimental data, will be disseminated via the project web site: http://www.cs.umd.edu/~amol/RPrDB.