Analyzing the Multimodal Behaviors of Users of a Speech-to-Speech Translation Device by using Concept Matching Scores.