Information exploration in search computing
More Info
expand_more
Abstract
Search computing queries typically address search tasks that go beyond a single interaction. In this paper, we show a query paradigm that supports multi-step, exploratory search over multiple Web data sources. Our paradigm requires users to be aware of searching over "interconnected objects" with given semantics, but each exploration step is simplified as much as possible, by presenting to users at each step simple interfaces, offering some choices that can be supported by the system; choices include moving "forward", by adding new objects to the search, or "backward", by excluding some objects from the search; and the selection and de-selection of displayed results in order to dynamically manipulate the result set. For supporting exploration, we designed a new architectural element, called query orchestrator, which connects the user interface module with the execution engine; the orchestrator maintains the history of the query session and caches query results for reuse at subsequent interactions.