A Computational Exploration
of the Creative Aspect of Language
In her recent project funded by the Ministry of Science, Nurit Melnik investigates linguistic creativity by combining traditional linguistics with computational methods. Her research explores how speakers innovate within language systems, examining the balance between novelty and comprehensibility that makes certain expressions engaging to listeners.
Among several linguistic phenomena analyzed, Melnik and her team documented patterns in which speakers modify conventional expressions, like "thinking outside the box", with unexpected variations (e.g., “building confidence to swim outside the box”, regarding acclimation to open water swimming). This effort incorporated computational models to measure similarity between conventional expressions and their innovative variants, providing quantitative evidence for theories regarding optimal innovations. By applying vector space models to linguistic data, Melnik et al. demonstrate that successful creative expressions maintain an intermediate similarity to their source — different enough to be noticed, but similar enough to be understood.