Graded structure in adjective categories
Simon De Deyne, Wouter Voorspoels, Steven Verheyen, Danielle Navarro, Gert Storms · eScholarship (California Digital Library) · 2011
In contrast to nouns, little is known about the graded structure of adjective categories. In this study we investigate if adjectives categories show a similar graded structure and explain it using a similarity-based account. The results show a reliable graded structure which is adequately explained by two formal models based on prototypes and exemplars, therefore generalizing the model performance from concrete nouns to adjectives. Finally, we show the attention weights of these models deal with the additional challenge of antonymy in adjectives and discuss the findings in the light of alternative accounts that do not rely on item-similarity. Keywords: concepts; adjectives, typicality, antonymy, GCM, prototypes. Research on natural language concepts generally focuses on object categories; that is, words, most often nouns, that refer to concrete objects such as cats, bat, apples and coconuts. Category members vary in typicality, with cats being more typical mammals than bats for example, and one of the most robust effects across a wide range of tasks is a processing advantage for typical items. This effect is thought to reflect a