Adaptive Referring Expression Generation in Spoken Dialogue Systems: Evaluation with Real Users