Research and business is currently moving from centralized databases towards information systems integrating distributed, autonomous data sources. With it, research focus has shifted from traditional query optimization to the field of query planning. Query planning is the problem of finding query execution plans across distributed, heterogeneous, overlapping, and autonomous data sources. We argue that for such data sources the main discriminator for different query execution strategies is no longer response time, as it is for database queries, but ‑ more generally ‑ the information quality (IQ) of the result. This thesis investigates the usage of iq-criteria to improve the answering of user queries against integrated information systems. We discuss what kind of iq-metadata is necessary, how it can be acquired, and ‑ most importantly ‑ how it can be used to improve the quality of query results and the performance of query planning algorithms. A simple application for these research issues is a meta-search engine that uses existing search engines as its distributed data sources. Other examples include stock information systems, travel guides, and distributed molecular biology databases.
Research and business is currently moving from centralized databases towards information systems integrating distributed, autonomous data sources. With it, research focus has shifted from traditional query optimization to the field of query planning. Query planning is the problem of finding query execution plans across distributed, heterogeneous, overlapping, and autonomous data sources. We argue that for such data sources the main discriminator for different query execution strategies is no longer response time, as it is for database queries, but ‑ more generally ‑ the information quality (IQ) of the result. This thesis investigates the usage of iq-criteria to improve the answering of user queries against integrated information systems. We discuss what kind of iq-metadata is necessary, how it can be acquired, and ‑ most importantly ‑ how it can be used to improve the quality of query results and the performance of query planning algorithms. A simple application for these research issues is a meta-search engine that uses existing search engines as its distributed data sources. Other examples include stock information systems, travel guides, and distributed molecular biology databases.
The thesis has three main parts. Part I lays the foundation for the problem of querying Web data sources and shows why iq-reasoning is helpful. We describe the mediator‑wrapper architecture and show how to describe sources and user queries using the concept of the universal relation. Several application examples serve as rationale throughout the thesis. Part II introduces our model of information quality.We present a comprehensive set of iq-criteria together with score assessment methods. Each data source is rated by a set of iq-criteria, such as completeness, understandability, or accuracy. To compare data sources and query plans qualitatively using multiple criteria, we present appropriate ranking methods, which aggregate iq-criterion scores to an overall quality value. Part III puts information quality to work by combining query planning with iq-reasoning. We revise the conventional query planning goal of finding all plans for a query: The new goal is to find the best N plans and use a quality model to quantify the term best’. We present two algorithms to solve this problem. The first acts as an add-on to any given query planning algorithm, the second explicitly integrates iq-reasoning into the planning process, thereby speeding up query planning itself. Next, we part from the conventional query planning paradigm of finding different plans for a query, each with a different result. The usage of new outerjoin-type merge operators to combine sources enables a reduction of the paradigm to finding a single, best plan. We concentrate on the completeness criterion describing the amount of data returned by a plan and present two families of optimization algorithms for different real world situations. All algorithms are evaluated using a simulation testbed. The main contribution of the thesis is the comprehensive integration of information quality reasoning and query planning. Research has recognized the importance of quality reasoning, but, to the best of our knowledge, iq-reasoning for query planning has not been adequately addressed before
Keine Kommunikationsobjekte vorhanden.
