| There is a striking disconnect between the objects a user interacts with on a daily basis, and the world of online information. Object-centric tasks like researching a book, shopping for a DVD player, listening to a piece of music, and purchasing medication can all be enriched with access to the right information, at the right time. We built several information systems that address these disconnects by finding information relevant to these tasks automatically on behalf of the user. Each of these systems requires connections to multiple online information sources. Sources such as Web search engines, often used by these systems, are siloed information systems---each has their own network access point, query interface, and result presentation layer. Siloed systems inherently cannot interact with each other. Moreover, because our intelligent information systems are dynamic and opportunistic with respect to information retrieval, they cannot be limited to using a static set of queries and information sources. In order to synthesize information from multiple sources, these systems also need to enact chains of information retrieval using sources and queries selected at runtime.; This thesis is about building intelligent information systems using the Information Source Adapter Platform (or ISA Platform, for short), a set of practical enabling technologies for developing intelligent information systems. The ISA Platform provides an infrastructure for building lightweight interfaces to existing resources that can be composed automatically using means-ends analysis. The platform enables developers to rapidly create new and innovative applications that leverage existing resources, without requiring the transition to semantic web technologies be complete. We discuss several research systems that make use of our framework, as well as a commercial deployment that uses ISA as an integral platform technology. |
