Exploratory activity search

Abstract

Real world activities can be found all around the web. In this paper, we present our work on an exploratory activity search system that is utilizing common-sense data to enrich the search for activities in the web. Two algorithms are proposed. One is the weighted query expansion algorithm using ConceptNet, and the other is the ranking algorithm using query expansion result whose weights are tuned by the genetic algorithms. Finally, a detailed double-blinded user evaluation of the system follows, where we simultaneously evaluate the activity search system from the viewpoint of novelty, serendipity, usefulness and interest. There were 20 participants in this evaluation, and the proposed system has shown a 13.4% increase in novelty and 11.6% increase in serendipity when compared with Bing search.