Quantifying count/mass elasticity

In this paper we investigate the mass/count distinction in the nominal domain from a large-scale distributional perspective. In particular, we address the elasticity of this contrast -- the fact that many nouns appear in both mass and count contexts (many hopes/much hope). We show that not only is the mass/count distinction not binary, it is not even a matter of degree, and that there is little correlation between the rate of mass and of count occurrences on a by noun basis. In addition, to investigate meaning shifts in the conversion of mass to count nouns we built a vector-space meaning model for plural (count) and singular (potentially non-count) uses of nouns. We show that the semantic distance between plural and singular nouns is much higher for those nouns that appear frequently in mass contexts than for those that appear frequently in count contexts.