... anearest-neighbor classifier to solve SAT verbal anal-ogy problems, yielding 47% accuracy. The same ap-proach is applied to classifying noun-modifier pairs:using the Diverse dataset of Nastase and Szpakow-icz ... material, measure, topic, type). For example,exam anxiety is classified as effect and therefore as CAUSALITY, and blue book is property and there-fore also PARTICIPANT.Some examples in the dataset ... we predicted its class, andif there were ties, we chose the class predicted by themajority of tied examples, if there was a majority.The results for the 30-class Diverse dataset areshown in...