Learning perception lattices to compare generative explanations of human-language activity stories

Abstract

Recent research in building large semantic networks of commonsense human-knowledge, such as ConceptNet, has been driven by a small number of hand-programmed rules for recognizing and placing value on specific numbers and types of semantic relationships. We propose statistical algorithms for learning these semantic relationships from partially ordered sets… (More)

Topics

1 Figure or Table

Cite this paper

@inproceedings{Morgan2006LearningPL, title={Learning perception lattices to compare generative explanations of human-language activity stories}, author={Bo Morgan}, year={2006} }