Incorporating Temporal and Semantic Information with Eye Gaze for Automatic Word Acquisition in Multimodal Conversational Systems