Investigating multimodal real-time patterns of joint attention in an hri word learning task