Live Visual Relevance Feedback for Query Formulation

Users browsing the Internet seem relatively satisfied with the performance of search engines. An optimistic explanation would be the high quality of search engines. A more pessimistic one would be that people just adapt easily to any new technology. A third explanation is people's ignorance about recall: as they simply don't know what relevant documents are missed, they can hardly be expected to worry about them. And so they easily conceive the result as the best they can get. To allow the user to better assess the quality of the search results, an algorithm was developed that computes a visual representation of the document space in the neighborhood of the user's query. The paper (1) outlines the algorithm, (2) shows how users can explore the neighborhood of a query, and (3) demonstrates how users can guess more judiciously whether they need to further elaborate their query to improve retrieval results