Applications of Top-k Representative Queries

Abstract

Top-k queries find a list of k objects that have the largest scores based on some user-provided relevance function. In practice, objects in the top-k list are often similar to each other and are thus informationally redundant. This possibility of information redundancy has ignited much interest in the problem of diversity-aware top-k queries. In diversity-aware queries, the goal is to maximize the diversity of the answer set, which in turn enhances the overall information content by representing a wider spectrum of the high scoring objects. In this paper, we demonstrate that the information content in the top-k answer set can be further enhanced by favoring the representative power of the answer set instead of diversity.