Exploring, understanding and recording analysis steps in information-management applications

Large-scale Web-based information management tasks require coordinated execution of a great number of information retrieval, analysis and visualization steps. Although previous research has been done in this area, it is currently impractical for users to properly manage and execute such a large number of steps, and to quickly adapt and reuse them for similar tasks and on different system environments. By developing and testing a Web-based information management system that helps users perform large-scale tasks, we observed seven key characteristics of users' information management activities and constraints of the Web-based system environment: incremental task development, recurrent task execution, evolving user understanding of their requirements, shared analysis patterns, collection-based information processing, dynamic aspects of Web resources, and a large number of component services. Based on these characteristics, we hypothesize that we can enable users to rapidly perform large-scale information management tasks by providing them with mechanisms to explore and test the use of various information sources and services based on their requirements, and to capture and reuse sets of high-level information management steps without knowing the technical details. These enable users to focus on overall information flows. To test the hypothesis that these mechanisms can facilitate much larger information management tasks, we have developed a middle layer for a Web-based information management system. The middle layer includes a taxonomy-based component description framework that allows users to explore various options to accomplish their tasks, an active document collection template composition mechanism that enables users to understand and capture sets of high-level information management steps, and a dataflow-based service coordination mechanism that allows users to test options by reconfiguring an application and by running them in the application. Using these mechanisms, we have observed an order of magnitude increase in the number of information management steps that users can handle. We have also observed a high rate of application reuse, and rapid adaptation of applications to local environments in practical application domains.