Developmental stages of perception and language acquisition in a perceptually grounded robot

The objective of this research is to develop a system for language learning based on a “minimum” of pre-wired language-specific functionality, that is compatible with observations of perceptual and language capabilities in the human developmental trajectory. In the proposed system, meaning (in terms of descriptions of events and spatial relations) is extracted from video images based on detection of position, motion, physical contact and their parameters. Meaning extraction requires attentional mechanisms that are implemented from low-level perceptual primitives. Mapping of sentence form to meaning is performed by learning grammatical constructions, i.e., sentence to meaning mappings as defined by Goldberg [Goldberg, A. (1995). Constructions. Chicago and London: Univ. of Chicago Press]. These are stored and retrieved from a “construction inventory” based on the constellation of grammatical function words uniquely identifying the target sentence structure. The resulting system displays robust acquisition behavior that reproduces certain observations from developmental studies, with very modest “innate” language specificity.