Using sentence context and implicit contrast to learn sensor-grounded meanings for relational and deictic words: the twig system

摘要: This thesis describes a novel system that allows robot to infer the meanings of new words from their usage in context. TWIG (Transportable Word Intension Generator) can parse simple sentences, determine reference any unknown objects or people environment through sentence context, and over time what are by building "definition trees" imply word structure. The was originally built learn pronouns, category has previously been unmodeled robotic learning literature, but is general enough some other categories, including prepositions transitive verbs. implemented on physical equipped with face detectors, vision systems, sensor network for object localization. succeeded "I" "you" refer speaker addressee; "he" must person neither these; "this" "that" proximal distal non-person objects; "above" "below" relative height; "am" "are" identity relation. be used production as well comprehension, found produce more correct sentences fewer incorrect about its than similar systems lacked system's extension inference definition tree capabilities. work contains several approaches area learning, also interpreted computational model how human infants use contrast meaning.