III: Small: Query and Goal Driven Entity Resolution Framework

数据清理技术传统上旨在提高后端数据仓库中的数据质量，现在正迅速崛起，成为实时信息访问的重要组成部分。随着Web朝着支持交互式分析的方向发展，并且基本搜索从简单的关键字检索转移到基于从网页中提取的语义更丰富的概念(例如实体)的检索，对能够帮助缓解数据质量挑战的“即时”清理技术的需求正在迅速增加。该项目探索了三项新的创新，这些创新将有助于推动数据清理，使其成为一项能够实现实时信息访问的嵌入式技术。第一个创新是“查询感知数据清理”，它基于这样一种观察，即可以显著地利用诸如查询之类的实时任务的特殊性来为数据清理过程带来新的优化。第二项创新是数据清理框架，它从当今系统的“尽力而为”的临时设置转变为一种原则性的方法，它暴露和利用了清理成本和所取得结果的质量之间的根本权衡。最后，由于清理结果需要提供给最终用户或分析代码，本提案假定并解决了如何将通过数据清理代码处理的结果提交给最终接受者的方法。主要贡献是隐藏数据中的不确定性并确定结果的机制，同时最大化最终应用程序目标。拟议的研究旨在通过促进实时数据清理和提高数据质量，对网上互动分析和搜索带来变革性的改进。该项目还旨在通过将作为这项研究的一部分而开发的机制纳入Web People Search Technology(WEST)，使WEST成为实时的动态网络人搜索工具，从而使研究界受益。其目标是支持West作为一个即插即用系统，其他研究人员可以在其中嵌入和测试他们的数据清理算法和工具。最后，计划中的研究、系统开发和教育活动将显著增强学生的教育体验，为他们在当今知识驱动的社会中更光明的未来做好准备。有关更多信息，请参阅该项目的网站：http://sherlock.ics.uci.edu

Data cleaning technologies, traditionally designed to improve quality of data in back-end data warehouses, are fast emerging as a vital component of real-time information access. As the Web evolves towards supporting interactive analytics and basic search migrates from simple keyword retrieval to retrieval based on semantically richer concepts (e.g., entities) extracted from web pages, the need for "on-the-fly" cleaning techniques that can help alleviate data quality challenges is rapidly increasing. This project explores three new innovations that will help advance data cleaning towards becoming an embedded enabling technology for real-time information access. The first innovation is "query-aware data cleaning" which is based on the observation that the specificity of the real-time task such as a query can be exploited significantly to bring new optimizations to the data cleaning process. The second innovation is a data cleaning framework that migrates from the "best-effort" adhoc setup of today's systems into a principled approach that exposes and exploits a fundamental tradeoff between the cost of cleaning and quality of results achieved. Finally, since results of cleaning need to be fed to the end-user or analysis code, the proposal postulates and addresses approaches towards how results processed through data cleaning code can be presented to the end-recipient. The primary contribution is mechanisms to hide the uncertainty in the data and determinize the results while maximizing the end application goals. The proposed research is intended to bring transformative improvements in interactive analytics and search on the web by facilitating real-time data cleaning and data quality enhancements. The project also aims to benefit the research community by incorporating mechanisms developed as part of this research into the Web People Search Technology (WEST), enabling WEST to become a real-time on-the-fly web people search tool. The goal is to support WEST as a plug-and-play system wherein other researchers could embed and test their data cleaning algorithms and tools. Finally, the planned research, system development, and educational activities are going to significantly enhance the educational experience of students, preparing them for a brighter future in the today's knowledge driven society.For further information see the project web site at the URL: http://sherlock.ics.uci.edu