Document Type
Conference Proceeding
Publication Date
2016
Publication Title
The Fourth Annual Conference on Advances in Cognitive Systems (ACS-16)
Abstract
Intelligent systems that process natural language need representations of knowledge to support a human-like thought process, and they often use natural language words or phrases to name and represent concepts in a knowledge base. But some theories of cognition claim that language and thought are not the same thing, and that human thought processes occur at a deeper level of representation than words and phrases in language. In this paper we present results of a human subjects study of language-free primitive decomposition as a representation for commonsense knowledge. We found that our subjects could comprehend and use a primitive decomposition representation; they demonstrated a facile understanding of the physical primitives from Conceptual Dependency, matching them reliably to sentences in ways that agreed with our expectations. Our results also show that the set of conceptual primitives we used resemble real human conceptualizations of natural language in ways that were sharp and coherent. Because our human subjects were recruited using a crowdsourcing platform, we claim that crowdsourcing may provide a vast and inexpensive source of conceptual structures based on primitive decomposition.
Rights
© 2016 Cognitive Systems Foundation.
Recommended Citation
Macbeth, Jamie C. and Barionnette, Marydjina, "The Coherence of Conceptual Primitives" (2016). Computer Science: Faculty Publications, Smith College, Northampton, MA.
https://scholarworks.smith.edu/csc_facpubs/389
Comments
Archived as publsihed.
The Fourth Annual Conference on Advances in Cognitive Systems (ACS-16), Chicago, Illinois, June 23-26, 2016.