Evaluation of co-speech gestures grounded in word-distributed representation

Area of Science:

• Artificial Intelligence
• Cognitive Science
• Robotics

Background:

• Perceivable intentions in artificial agents are linked to resolving the symbol grounding problem.
• Symbol grounding involves endowing agent language with quantitative meaning from the physical world.

Purpose of the Study:

• To develop a method for characterizing robot gestures with quantitative meaning derived from text data.
• To enable artificial agents to exhibit perceivable intentions through grounded language and gestures.

Main Methods:

• Utilized word-distributed representations from a large text corpus.
• Generated "size images" of words by defining axes in vector space.
• Converted size images into physical robot gestures (movement and posture).

Main Results:

• Human ratings on gesture "naturalness" partially validated the proposed method.
• Demonstrated a link between quantitative word meaning and perceived communicative intention in robots.

Conclusions:

• The proposed method offers a pathway to achieving human-like symbolic grounding in artificial agents.
• Advanced artifacts with enhanced symbolic grounding and perceivable intentions are a future possibility.