| This thesis presents InOrder, a collaborative search interface for guided query refinement. Collaborative query reformulation is supported using an interactive web-mining process, encouraging selection of relevant keywords when appropriate search terminology is not known. A visual, stigmergetic interface is used to aggregate this collective activity into adaptive semantic reaps. These evolving semantic models represent search intent within a domain and help users identify and clarify search requirements through examination of past search strategies. Design considerations based on information relevance theory, cognitive factors and swarm systems are presented, illustrating how this framework has potential to create a self-organizing semantic system for search decision support. Finally, InOrder is shown to mitigate several issues present within standard search tools, and improve the efficiency of document retrieval within unfamiliar task-domains. |
