Evaluating an Algorithm for the Generation of Multimodal Referring Expressions in a Virtual World: A Pilot Study