Automated scoring of originality using semantic representations

J. Isaiah HarbisonUniversity of Maryland
Henk HaarmannUniversity of Maryland

Abstract

Originality, a key aspect of creativity, is difficult to measure. We tested the relationship between originality and similarity in two semantic spaces: latent semantic analysis (LSA) and pointwise mutual information (PMI). Similarity in both spaces was negatively correlated with human judgments of originality of responses on a test of divergent thinking. PMI was correlated more strongly both with human judgments of similarity and human judgments of originality. In particular, the average PMI between two phrases was found to be the strongest predictor of phrase similarity and originality, even performing better than participants' self assessments of their originality.

Files

Automated scoring of originality using semantic representations (102 KB)



Back to Table of Contents