A minimally-intrusive approach for query-driven data integration systems

Abstract

Integrating data from very large, dynamic, heterogeneous and autonomous data sources is a key requirement to satisfy growing information needs. In order to allow for ad-hoc answering of analytical questions, necessary up-front integration effort must be minimized and data integration systems must be adapted to the expectations and requirements of their users. While existing approaches offer support for incremental data integration, they require significant usage effort. They usually rely on explicit user feedback or invade the proven analysis workflows of their users. We propose a minimally-intrusive approach for query-driven data integration systems that allows for ad-hoc analysis of different data sources and minimizes the alteration of established analysis workflows. We suggest mechanisms for data source discovery and incremental data integration that primarily rely on implicit user feedback collected by query log analysis. Our proposed system architecture supports different types of data sources and integrates with existing data analysis tools.