From Adjective Glosses to Attribute Concepts: Learning Different Aspects That an Adjective Can Describe

Adjectives are one of the major contributors to conveying subjective meaning to sentences. Understanding the semantics of adjectives is essential for understanding natural language sentences. In this paper we propose a novel approach for learning different properties that an adjective can describe, corresponding to the ‘attribute concepts’, which is not currently available in existing linguistic resources. We accomplish this by reading adjective glosses and bootstrapping attribute concepts from a seed of adjectives. We show that we can learn new attribute concepts using adjective glosses of WordNet adjectives with high accuracy as compared with human annotation on a test set.