Smart gesture selection with word embeddings applied to NAO robot

Abstract

Nowadays, Human-Robot Interaction (HRI) field is growing by the day, a fact which is evidenced by the increasing number of existing projects as well as the application of increasingly advanced techniques from different areas of knowledge and multi-disciplinary approaches. In a future where technology automatically controls services such as health care, pedagogy or construction, social interfaces would be one of the necessary pillars of HRI field. In this context, gesture plays an important role in the transmission of information and is one of fundamental mechanisms relevant to human-robot interaction. This work proposes a new methodology for gestural annotation in free text through a semantic similarity analysis using distributed representations based on word embeddings. The intention with this is to endow NAO robot with an intelligent mechanism for gesture allocation.