ConceptMap: A Conceptual Approach for Formulating User Preferences in Large Information Spaces

Abstract

In a large information space a user needs to iteratively investigate the data to formulate her preferences for IR systems. In recent years several visualization techniques have been proposed to help a user to better formulate her preferences. However, using these solutions a user needs to explicitly specify her preferences for IR systems in forms of keywords or phrases. In this paper we present ConceptMap, a system that takes the advantage of deep learning and a knowledge lake to provide a conceptual summary of the information space. ConceptMap allows a user to specify her preferences implicitly as a set of concepts without the need to iteratively investigate the information space. It provides a 2D Radial Map of concepts where a user can rank items relevant to her preferences through dragging and dropping. Our experiment results shows that ConceptMap can help users to better formulate their preferences when they need to retrieve varied and comprehensive list of information across a large amount of data.