Dynamic database integration for knowledge sharing

组织淹没在数据中。最基本的问题是，我们如何管理所有这些数据，并将其转化为有价值的信息？ 数据以不同的格式分散在不同的系统中，并由不同的组织控制。它就像一个巨大的拼图。 为了回答问题，你需要来自各个地方的碎片，只有当碎片放在一起时，才能产生有意义的结果。 将数据转化为信息的关键是工具，这些工具允许您从单个片段构建整个画面。 该提案旨在开发算法和相关工具，以帮助所有用户查找、查询和集成来自多个数据源的数据。 该方法适用于医学信息、商业数据的集成，以及海量科学数据的传播与集成。 数据库集成是将来自多个数据源的数据组合到一个统一的信息系统中的过程。 虽然在集成架构和模式匹配算法方面取得了进展，但数据库集成仍然是一个具有挑战性和人力密集型的问题。 主要挑战之一是构建一个集成视图，将所有数据源中的信息组合在一起。 建议的研究是开发技术，支持快速，动态集成的数据库系统。 本文的研究主要集中在两个方面：集成系统的查询处理与优化和基于本体的模式匹配与合并。 在分布式查询处理中，对联接进行评估和排序对于实现高性能至关重要。 将研究如何改进基于散列的早期连接算法，并将其与特定于集成环境的优化器一起使用。第二个研究领域是使用本体（适合计算机处理的领域表示）来提高模式匹配的准确性。 目标是支持迭代模式合并和细化，以生成集成视图。 该研究将使用先前构建的Unity集成架构进行连接算法，查询优化算法和模式合并方法的实验和评估。

Organizations are drowning in data. The basic question is how can we manage all this data and turn it into valuable information?  Data is scattered around different systems, in different formats, and controlled by various organizations. It is like a giant jigsaw puzzle.  In order to answer questions, you need pieces from various places, and only when the pieces are put together can meaningful results be produced.  The key to turning data into information are the tools that allow you to construct the whole picture from the individual pieces.   This proposal aims to develop the algorithms and associated tools to help all users find, query, and integrate data from many data sources.  The approaches are applicable to the integration of medical information, business data, and the dissemination and integration of the vast amounts of scientific data.   Database integration is the process of combining data from multiple data sources into a single unified information system.  Although advances have been made including integration architectures and schema matching algorithms, database integration remains a challenging and human-intensive problem.  One of the major challenges is the construction of an integrated view combining the information in all data sources.    The proposed research is to develop techniques for supporting rapid, dynamic integration of database systems.  The research is in two areas: query processing and optimization for integration systems and ontology-based schema matching and merging.  Evaluating and ordering joins in distributed query processing is critical to achieving high performance.  Research will be performed on how to improve the hash-based early join algorithms and utilize them with optimizers specific for the integration environment. The second area of study is using ontologies (domain representations suitable for computer processing) to improve the accuracy of matching schemas.  The goal is to support iterative schema merging and refinement to produce an integrated view.  The research will use the previously constructed Unity integration architecture for experimentation and evaluation of join algorithms, query optimization heuristics, and schema merging approaches.