A Preliminary Study on Inducing Lexical Concreteness from Dictionary Definitions

Abstract

While the distinction between concrete words and abstract words appears to be inherent, the measure of lexical concreteness relying on human ratings is more intuitive than objective. In this study, we aim at extending the concreteness distinction from the lexical level to the sense level, and inducing a numerical index of concreteness for individual senses and words from dictionary definitions. The high overall agreement between human ratings and definition-induced ratings is encouraging for us to further simulate the distinction from more language resources. Such a simulated index for concreteness is believed to inform not only lexicography but also natural language processing tasks like automatic word sense disambiguation.