Making the Constraint Model of Concept Combination More Familiar and Efficient

Constraint theory and the C model, which is the computational-level implementation of the Constraint theory, explain how novel word combinations may be understood satisfying three constraints of diagnosticity, plausibility, and informativeness. Two limitations of the C model are less efficient computation procedures that operate over a large space of possibilities in order to identify the best interpretations, and the system’s inability to deal with familiarity effects. In this paper we explore the issue of familiarity in the comprehension of noun-noun compounds and we present PUNC a prototype system for Producing and Understanding NounNoun Compounds. PUNC tries to overcome algorithmic limitations of the C model in being more efficient in its computational complexity. It also deals with a wider span of empirical phenomena, such as dimensions of word familiarity.